sit.c

   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 *	IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
   4 *	Linux INET6 implementation
   5 *
   6 *	Authors:
   7 *	Pedro Roque		<roque@di.fc.ul.pt>
   8 *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
   9 *
  10 *	Changes:
  11 * Roger Venning <r.venning@telstra.com>:	6to4 support
  12 * Nate Thompson <nate@thebog.net>:		6to4 support
  13 * Fred Templin <fred.l.templin@boeing.com>:	isatap support
  14 */
  15
  16#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  17
  18#include <linux/module.h>
  19#include <linux/capability.h>
  20#include <linux/errno.h>
  21#include <linux/types.h>
  22#include <linux/socket.h>
  23#include <linux/sockios.h>
  24#include <linux/net.h>
  25#include <linux/in6.h>
  26#include <linux/netdevice.h>
  27#include <linux/if_arp.h>
  28#include <linux/icmp.h>
  29#include <linux/slab.h>
  30#include <linux/uaccess.h>
  31#include <linux/init.h>
  32#include <linux/netfilter_ipv4.h>
  33#include <linux/if_ether.h>
  34
  35#include <net/sock.h>
  36#include <net/snmp.h>
  37
  38#include <net/ipv6.h>
  39#include <net/protocol.h>
  40#include <net/transp_v6.h>
  41#include <net/ip6_fib.h>
  42#include <net/ip6_route.h>
  43#include <net/ndisc.h>
  44#include <net/addrconf.h>
  45#include <net/ip.h>
  46#include <net/udp.h>
  47#include <net/icmp.h>
  48#include <net/ip_tunnels.h>
  49#include <net/inet_ecn.h>
  50#include <net/xfrm.h>
  51#include <net/dsfield.h>
  52#include <net/net_namespace.h>
  53#include <net/netns/generic.h>
  54#include <net/netdev_lock.h>
  55#include <net/inet_dscp.h>
  56
  57/*
  58   This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  59
  60   For comments look at net/ipv4/ip_gre.c --ANK
  61 */
  62
  63#define IP6_SIT_HASH_SIZE  16
  64#define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  65
  66static bool log_ecn_error = true;
  67module_param(log_ecn_error, bool, 0644);
  68MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  69
  70static int ipip6_tunnel_init(struct net_device *dev);
  71static void ipip6_tunnel_setup(struct net_device *dev);
  72static void ipip6_dev_free(struct net_device *dev);
  73static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
  74		      __be32 *v4dst);
  75static struct rtnl_link_ops sit_link_ops __read_mostly;
  76
  77static unsigned int sit_net_id __read_mostly;
  78struct sit_net {
  79	struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
  80	struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
  81	struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE];
  82	struct ip_tunnel __rcu *tunnels_wc[1];
  83	struct ip_tunnel __rcu **tunnels[4];
  84
  85	struct net_device *fb_tunnel_dev;
  86};
  87
  88static inline struct sit_net *dev_to_sit_net(struct net_device *dev)
  89{
  90	struct ip_tunnel *t = netdev_priv(dev);
  91
  92	return net_generic(t->net, sit_net_id);
  93}
  94
  95/*
  96 * Must be invoked with rcu_read_lock
  97 */
  98static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
  99					     struct net_device *dev,
 100					     __be32 remote, __be32 local,
 101					     int sifindex)
 102{
 103	unsigned int h0 = HASH(remote);
 104	unsigned int h1 = HASH(local);
 105	struct ip_tunnel *t;
 106	struct sit_net *sitn = net_generic(net, sit_net_id);
 107	int ifindex = dev ? dev->ifindex : 0;
 108
 109	for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 110		if (local == t->parms.iph.saddr &&
 111		    remote == t->parms.iph.daddr &&
 112		    (!dev || !t->parms.link || ifindex == t->parms.link ||
 113		     sifindex == t->parms.link) &&
 114		    (t->dev->flags & IFF_UP))
 115			return t;
 116	}
 117	for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
 118		if (remote == t->parms.iph.daddr &&
 119		    (!dev || !t->parms.link || ifindex == t->parms.link ||
 120		     sifindex == t->parms.link) &&
 121		    (t->dev->flags & IFF_UP))
 122			return t;
 123	}
 124	for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 125		if (local == t->parms.iph.saddr &&
 126		    (!dev || !t->parms.link || ifindex == t->parms.link ||
 127		     sifindex == t->parms.link) &&
 128		    (t->dev->flags & IFF_UP))
 129			return t;
 130	}
 131	t = rcu_dereference(sitn->tunnels_wc[0]);
 132	if (t && (t->dev->flags & IFF_UP))
 133		return t;
 134	return NULL;
 135}
 136
 137static struct ip_tunnel __rcu **
 138__ipip6_bucket(struct sit_net *sitn, struct ip_tunnel_parm_kern *parms)
 139{
 140	__be32 remote = parms->iph.daddr;
 141	__be32 local = parms->iph.saddr;
 142	unsigned int h = 0;
 143	int prio = 0;
 144
 145	if (remote) {
 146		prio |= 2;
 147		h ^= HASH(remote);
 148	}
 149	if (local) {
 150		prio |= 1;
 151		h ^= HASH(local);
 152	}
 153	return &sitn->tunnels[prio][h];
 154}
 155
 156static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 157		struct ip_tunnel *t)
 158{
 159	return __ipip6_bucket(sitn, &t->parms);
 160}
 161
 162static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 163{
 164	struct ip_tunnel __rcu **tp;
 165	struct ip_tunnel *iter;
 166
 167	for (tp = ipip6_bucket(sitn, t);
 168	     (iter = rtnl_dereference(*tp)) != NULL;
 169	     tp = &iter->next) {
 170		if (t == iter) {
 171			rcu_assign_pointer(*tp, t->next);
 172			break;
 173		}
 174	}
 175}
 176
 177static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 178{
 179	struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 180
 181	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 182	rcu_assign_pointer(*tp, t);
 183}
 184
 185static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 186{
 187#ifdef CONFIG_IPV6_SIT_6RD
 188	struct ip_tunnel *t = netdev_priv(dev);
 189
 190	if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
 191		ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 192		t->ip6rd.relay_prefix = 0;
 193		t->ip6rd.prefixlen = 16;
 194		t->ip6rd.relay_prefixlen = 0;
 195	} else {
 196		struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 197		memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 198	}
 199#endif
 200}
 201
 202static int ipip6_tunnel_create(struct net_device *dev)
 203{
 204	struct ip_tunnel *t = netdev_priv(dev);
 205	struct sit_net *sitn = net_generic(t->net, sit_net_id);
 206	int err;
 207
 208	__dev_addr_set(dev, &t->parms.iph.saddr, 4);
 209	memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
 210
 211	if (test_bit(IP_TUNNEL_SIT_ISATAP_BIT, t->parms.i_flags))
 212		dev->priv_flags |= IFF_ISATAP;
 213
 214	dev->rtnl_link_ops = &sit_link_ops;
 215
 216	err = register_netdevice(dev);
 217	if (err < 0)
 218		goto out;
 219
 220	ipip6_tunnel_clone_6rd(dev, sitn);
 221
 222	ipip6_tunnel_link(sitn, t);
 223	return 0;
 224
 225out:
 226	return err;
 227}
 228
 229static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 230					     struct ip_tunnel_parm_kern *parms,
 231					     int create)
 232{
 233	__be32 remote = parms->iph.daddr;
 234	__be32 local = parms->iph.saddr;
 235	struct ip_tunnel *t, *nt;
 236	struct ip_tunnel __rcu **tp;
 237	struct net_device *dev;
 238	char name[IFNAMSIZ];
 239	struct sit_net *sitn = net_generic(net, sit_net_id);
 240
 241	for (tp = __ipip6_bucket(sitn, parms);
 242	    (t = rtnl_dereference(*tp)) != NULL;
 243	     tp = &t->next) {
 244		if (local == t->parms.iph.saddr &&
 245		    remote == t->parms.iph.daddr &&
 246		    parms->link == t->parms.link) {
 247			if (create)
 248				return NULL;
 249			else
 250				return t;
 251		}
 252	}
 253	if (!create)
 254		goto failed;
 255
 256	if (parms->name[0]) {
 257		if (!dev_valid_name(parms->name))
 258			goto failed;
 259		strscpy(name, parms->name, IFNAMSIZ);
 260	} else {
 261		strcpy(name, "sit%d");
 262	}
 263	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 264			   ipip6_tunnel_setup);
 265	if (!dev)
 266		return NULL;
 267
 268	dev_net_set(dev, net);
 269
 270	nt = netdev_priv(dev);
 271
 272	nt->net = net;
 273	nt->parms = *parms;
 274	if (ipip6_tunnel_create(dev) < 0)
 275		goto failed_free;
 276
 277	if (!parms->name[0])
 278		strcpy(parms->name, dev->name);
 279
 280	return nt;
 281
 282failed_free:
 283	free_netdev(dev);
 284failed:
 285	return NULL;
 286}
 287
 288#define for_each_prl_rcu(start)			\
 289	for (prl = rcu_dereference(start);	\
 290	     prl;				\
 291	     prl = rcu_dereference(prl->next))
 292
 293static struct ip_tunnel_prl_entry *
 294__ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 295{
 296	struct ip_tunnel_prl_entry *prl;
 297
 298	for_each_prl_rcu(t->prl)
 299		if (prl->addr == addr)
 300			break;
 301	return prl;
 302
 303}
 304
 305static int ipip6_tunnel_get_prl(struct net_device *dev, struct ip_tunnel_prl __user *a)
 306{
 307	struct ip_tunnel *t = netdev_priv(dev);
 308	struct ip_tunnel_prl kprl, *kp;
 309	struct ip_tunnel_prl_entry *prl;
 310	unsigned int cmax, c = 0, ca, len;
 311	int ret = 0;
 312
 313	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
 314		return -EINVAL;
 315
 316	if (copy_from_user(&kprl, a, sizeof(kprl)))
 317		return -EFAULT;
 318	cmax = kprl.datalen / sizeof(kprl);
 319	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 320		cmax = 1;
 321
 322	/* For simple GET or for root users,
 323	 * we try harder to allocate.
 324	 */
 325	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 326		kcalloc(cmax, sizeof(*kp), GFP_KERNEL_ACCOUNT | __GFP_NOWARN) :
 327		NULL;
 328
 329	ca = min(t->prl_count, cmax);
 330
 331	if (!kp) {
 332		/* We don't try hard to allocate much memory for
 333		 * non-root users.
 334		 * For root users, retry allocating enough memory for
 335		 * the answer.
 336		 */
 337		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC | __GFP_ACCOUNT |
 338					      __GFP_NOWARN);
 339		if (!kp) {
 340			ret = -ENOMEM;
 341			goto out;
 342		}
 343	}
 344
 345	rcu_read_lock();
 346	for_each_prl_rcu(t->prl) {
 347		if (c >= cmax)
 348			break;
 349		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 350			continue;
 351		kp[c].addr = prl->addr;
 352		kp[c].flags = prl->flags;
 353		c++;
 354		if (kprl.addr != htonl(INADDR_ANY))
 355			break;
 356	}
 357
 358	rcu_read_unlock();
 359
 360	len = sizeof(*kp) * c;
 361	ret = 0;
 362	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 363		ret = -EFAULT;
 364
 365	kfree(kp);
 366out:
 367	return ret;
 368}
 369
 370static int
 371ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 372{
 373	struct ip_tunnel_prl_entry *p;
 374	int err = 0;
 375
 376	if (a->addr == htonl(INADDR_ANY))
 377		return -EINVAL;
 378
 379	ASSERT_RTNL();
 380
 381	for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 382		if (p->addr == a->addr) {
 383			if (chg) {
 384				p->flags = a->flags;
 385				goto out;
 386			}
 387			err = -EEXIST;
 388			goto out;
 389		}
 390	}
 391
 392	if (chg) {
 393		err = -ENXIO;
 394		goto out;
 395	}
 396
 397	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 398	if (!p) {
 399		err = -ENOBUFS;
 400		goto out;
 401	}
 402
 403	p->next = t->prl;
 404	p->addr = a->addr;
 405	p->flags = a->flags;
 406	t->prl_count++;
 407	rcu_assign_pointer(t->prl, p);
 408out:
 409	return err;
 410}
 411
 412static void prl_list_destroy_rcu(struct rcu_head *head)
 413{
 414	struct ip_tunnel_prl_entry *p, *n;
 415
 416	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 417	do {
 418		n = rcu_dereference_protected(p->next, 1);
 419		kfree(p);
 420		p = n;
 421	} while (p);
 422}
 423
 424static int
 425ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 426{
 427	struct ip_tunnel_prl_entry *x;
 428	struct ip_tunnel_prl_entry __rcu **p;
 429	int err = 0;
 430
 431	ASSERT_RTNL();
 432
 433	if (a && a->addr != htonl(INADDR_ANY)) {
 434		for (p = &t->prl;
 435		     (x = rtnl_dereference(*p)) != NULL;
 436		     p = &x->next) {
 437			if (x->addr == a->addr) {
 438				*p = x->next;
 439				kfree_rcu(x, rcu_head);
 440				t->prl_count--;
 441				goto out;
 442			}
 443		}
 444		err = -ENXIO;
 445	} else {
 446		x = rtnl_dereference(t->prl);
 447		if (x) {
 448			t->prl_count = 0;
 449			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 450			t->prl = NULL;
 451		}
 452	}
 453out:
 454	return err;
 455}
 456
 457static int ipip6_tunnel_prl_ctl(struct net_device *dev,
 458				struct ip_tunnel_prl __user *data, int cmd)
 459{
 460	struct ip_tunnel *t = netdev_priv(dev);
 461	struct ip_tunnel_prl prl;
 462	int err;
 463
 464	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
 465		return -EPERM;
 466	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
 467		return -EINVAL;
 468
 469	if (copy_from_user(&prl, data, sizeof(prl)))
 470		return -EFAULT;
 471
 472	switch (cmd) {
 473	case SIOCDELPRL:
 474		err = ipip6_tunnel_del_prl(t, &prl);
 475		break;
 476	case SIOCADDPRL:
 477	case SIOCCHGPRL:
 478		err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 479		break;
 480	}
 481	dst_cache_reset(&t->dst_cache);
 482	netdev_state_change(dev);
 483	return err;
 484}
 485
 486static int
 487isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 488{
 489	struct ip_tunnel_prl_entry *p;
 490	int ok = 1;
 491
 492	rcu_read_lock();
 493	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 494	if (p) {
 495		if (p->flags & PRL_DEFAULT)
 496			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 497		else
 498			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 499	} else {
 500		const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 501
 502		if (ipv6_addr_is_isatap(addr6) &&
 503		    (addr6->s6_addr32[3] == iph->saddr) &&
 504		    ipv6_chk_prefix(addr6, t->dev))
 505			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 506		else
 507			ok = 0;
 508	}
 509	rcu_read_unlock();
 510	return ok;
 511}
 512
 513static void ipip6_tunnel_uninit(struct net_device *dev)
 514{
 515	struct ip_tunnel *tunnel = netdev_priv(dev);
 516	struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
 517
 518	if (dev == sitn->fb_tunnel_dev) {
 519		RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 520	} else {
 521		ipip6_tunnel_unlink(sitn, tunnel);
 522		ipip6_tunnel_del_prl(tunnel, NULL);
 523	}
 524	dst_cache_reset(&tunnel->dst_cache);
 525	netdev_put(dev, &tunnel->dev_tracker);
 526}
 527
 528static int ipip6_err(struct sk_buff *skb, u32 info)
 529{
 530	const struct iphdr *iph = (const struct iphdr *)skb->data;
 531	const int type = icmp_hdr(skb)->type;
 532	const int code = icmp_hdr(skb)->code;
 533	unsigned int data_len = 0;
 534	struct ip_tunnel *t;
 535	int sifindex;
 536	int err;
 537
 538	switch (type) {
 539	default:
 540	case ICMP_PARAMETERPROB:
 541		return 0;
 542
 543	case ICMP_DEST_UNREACH:
 544		switch (code) {
 545		case ICMP_SR_FAILED:
 546			/* Impossible event. */
 547			return 0;
 548		default:
 549			/* All others are translated to HOST_UNREACH.
 550			   rfc2003 contains "deep thoughts" about NET_UNREACH,
 551			   I believe they are just ether pollution. --ANK
 552			 */
 553			break;
 554		}
 555		break;
 556	case ICMP_TIME_EXCEEDED:
 557		if (code != ICMP_EXC_TTL)
 558			return 0;
 559		data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
 560		break;
 561	case ICMP_REDIRECT:
 562		break;
 563	}
 564
 565	err = -ENOENT;
 566
 567	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 568	t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 569				iph->daddr, iph->saddr, sifindex);
 570	if (!t)
 571		goto out;
 572
 573	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 574		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 575				 t->parms.link, iph->protocol);
 576		err = 0;
 577		goto out;
 578	}
 579	if (type == ICMP_REDIRECT) {
 580		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
 581			      iph->protocol);
 582		err = 0;
 583		goto out;
 584	}
 585
 586	err = 0;
 587	if (__in6_dev_get(skb->dev) &&
 588	    !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
 589		goto out;
 590
 591	if (t->parms.iph.daddr == 0)
 592		goto out;
 593
 594	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 595		goto out;
 596
 597	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 598		t->err_count++;
 599	else
 600		t->err_count = 1;
 601	t->err_time = jiffies;
 602out:
 603	return err;
 604}
 605
 606static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 607				  const struct in6_addr *v6addr)
 608{
 609	__be32 v4embed = 0;
 610	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 611		return true;
 612	return false;
 613}
 614
 615/* Checks if an address matches an address on the tunnel interface.
 616 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 617 * Long story:
 618 * This function is called after we considered the packet as spoofed
 619 * in is_spoofed_6rd.
 620 * We may have a router that is doing NAT for proto 41 packets
 621 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 622 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 623 * function will return true, dropping the packet.
 624 * But, we can still check if is spoofed against the IP
 625 * addresses associated with the interface.
 626 */
 627static bool only_dnatted(const struct ip_tunnel *tunnel,
 628	const struct in6_addr *v6dst)
 629{
 630	int prefix_len;
 631
 632#ifdef CONFIG_IPV6_SIT_6RD
 633	prefix_len = tunnel->ip6rd.prefixlen + 32
 634		- tunnel->ip6rd.relay_prefixlen;
 635#else
 636	prefix_len = 48;
 637#endif
 638	return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 639}
 640
 641/* Returns true if a packet is spoofed */
 642static bool packet_is_spoofed(struct sk_buff *skb,
 643			      const struct iphdr *iph,
 644			      struct ip_tunnel *tunnel)
 645{
 646	const struct ipv6hdr *ipv6h;
 647
 648	if (tunnel->dev->priv_flags & IFF_ISATAP) {
 649		if (!isatap_chksrc(skb, iph, tunnel))
 650			return true;
 651
 652		return false;
 653	}
 654
 655	if (tunnel->dev->flags & IFF_POINTOPOINT)
 656		return false;
 657
 658	ipv6h = ipv6_hdr(skb);
 659
 660	if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 661		net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 662				     &iph->saddr, &ipv6h->saddr,
 663				     &iph->daddr, &ipv6h->daddr);
 664		return true;
 665	}
 666
 667	if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 668		return false;
 669
 670	if (only_dnatted(tunnel, &ipv6h->daddr))
 671		return false;
 672
 673	net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 674			     &iph->saddr, &ipv6h->saddr,
 675			     &iph->daddr, &ipv6h->daddr);
 676	return true;
 677}
 678
 679static int ipip6_rcv(struct sk_buff *skb)
 680{
 681	const struct iphdr *iph = ip_hdr(skb);
 682	struct ip_tunnel *tunnel;
 683	int sifindex;
 684	int err;
 685
 686	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 687	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 688				     iph->saddr, iph->daddr, sifindex);
 689	if (tunnel) {
 690		if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 691		    tunnel->parms.iph.protocol != 0)
 692			goto out;
 693
 694		skb->mac_header = skb->network_header;
 695		skb_reset_network_header(skb);
 696		IPCB(skb)->flags = 0;
 697		skb->dev = tunnel->dev;
 698
 699		if (packet_is_spoofed(skb, iph, tunnel)) {
 700			DEV_STATS_INC(tunnel->dev, rx_errors);
 701			goto out;
 702		}
 703
 704		if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
 705		    !net_eq(tunnel->net, dev_net(tunnel->dev))))
 706			goto out;
 707
 708		/* skb can be uncloned in iptunnel_pull_header, so
 709		 * old iph is no longer valid
 710		 */
 711		iph = (const struct iphdr *)skb_mac_header(skb);
 712		skb_reset_mac_header(skb);
 713
 714		err = IP_ECN_decapsulate(iph, skb);
 715		if (unlikely(err)) {
 716			if (log_ecn_error)
 717				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 718						     &iph->saddr, iph->tos);
 719			if (err > 1) {
 720				DEV_STATS_INC(tunnel->dev, rx_frame_errors);
 721				DEV_STATS_INC(tunnel->dev, rx_errors);
 722				goto out;
 723			}
 724		}
 725
 726		dev_sw_netstats_rx_add(tunnel->dev, skb->len);
 727
 728		netif_rx(skb);
 729
 730		return 0;
 731	}
 732
 733	/* no tunnel matched,  let upstream know, ipsec may handle it */
 734	return 1;
 735out:
 736	kfree_skb(skb);
 737	return 0;
 738}
 739
 740static const struct tnl_ptk_info ipip_tpi = {
 741	/* no tunnel info required for ipip. */
 742	.proto = htons(ETH_P_IP),
 743};
 744
 745#if IS_ENABLED(CONFIG_MPLS)
 746static const struct tnl_ptk_info mplsip_tpi = {
 747	/* no tunnel info required for mplsip. */
 748	.proto = htons(ETH_P_MPLS_UC),
 749};
 750#endif
 751
 752static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
 753{
 754	const struct iphdr *iph;
 755	struct ip_tunnel *tunnel;
 756	int sifindex;
 757
 758	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 759
 760	iph = ip_hdr(skb);
 761	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 762				     iph->saddr, iph->daddr, sifindex);
 763	if (tunnel) {
 764		const struct tnl_ptk_info *tpi;
 765
 766		if (tunnel->parms.iph.protocol != ipproto &&
 767		    tunnel->parms.iph.protocol != 0)
 768			goto drop;
 769
 770		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 771			goto drop;
 772#if IS_ENABLED(CONFIG_MPLS)
 773		if (ipproto == IPPROTO_MPLS)
 774			tpi = &mplsip_tpi;
 775		else
 776#endif
 777			tpi = &ipip_tpi;
 778		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 779			goto drop;
 780		skb_reset_mac_header(skb);
 781
 782		return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 783	}
 784
 785	return 1;
 786
 787drop:
 788	kfree_skb(skb);
 789	return 0;
 790}
 791
 792static int ipip_rcv(struct sk_buff *skb)
 793{
 794	return sit_tunnel_rcv(skb, IPPROTO_IPIP);
 795}
 796
 797#if IS_ENABLED(CONFIG_MPLS)
 798static int mplsip_rcv(struct sk_buff *skb)
 799{
 800	return sit_tunnel_rcv(skb, IPPROTO_MPLS);
 801}
 802#endif
 803
 804/*
 805 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 806 * stores the embedded IPv4 address in v4dst and returns true.
 807 */
 808static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 809		      __be32 *v4dst)
 810{
 811#ifdef CONFIG_IPV6_SIT_6RD
 812	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 813			      tunnel->ip6rd.prefixlen)) {
 814		unsigned int pbw0, pbi0;
 815		int pbi1;
 816		u32 d;
 817
 818		pbw0 = tunnel->ip6rd.prefixlen >> 5;
 819		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 820
 821		d = tunnel->ip6rd.relay_prefixlen < 32 ?
 822			(ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 823		    tunnel->ip6rd.relay_prefixlen : 0;
 824
 825		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 826		if (pbi1 > 0)
 827			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 828			     (32 - pbi1);
 829
 830		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 831		return true;
 832	}
 833#else
 834	if (v6dst->s6_addr16[0] == htons(0x2002)) {
 835		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 836		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 837		return true;
 838	}
 839#endif
 840	return false;
 841}
 842
 843static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 844			     const struct in6_addr *v6dst)
 845{
 846	__be32 dst = 0;
 847	check_6rd(tunnel, v6dst, &dst);
 848	return dst;
 849}
 850
 851static bool ipip6_tunnel_dst_find(struct sk_buff *skb, __be32 *dst,
 852				  bool is_isatap)
 853{
 854	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 855	struct neighbour *neigh = NULL;
 856	const struct in6_addr *addr6;
 857	bool found = false;
 858	int addr_type;
 859
 860	if (skb_dst(skb))
 861		neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 862
 863	if (!neigh) {
 864		net_dbg_ratelimited("nexthop == NULL\n");
 865		return false;
 866	}
 867
 868	addr6 = (const struct in6_addr *)&neigh->primary_key;
 869	addr_type = ipv6_addr_type(addr6);
 870
 871	if (is_isatap) {
 872		if ((addr_type & IPV6_ADDR_UNICAST) &&
 873		    ipv6_addr_is_isatap(addr6)) {
 874			*dst = addr6->s6_addr32[3];
 875			found = true;
 876		}
 877	} else {
 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			found = true;
 886		}
 887	}
 888
 889	neigh_release(neigh);
 890
 891	return found;
 892}
 893
 894/*
 895 *	This function assumes it is being called from dev_queue_xmit()
 896 *	and that skb is filled properly by that function.
 897 */
 898
 899static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 900				     struct net_device *dev)
 901{
 902	struct ip_tunnel *tunnel = netdev_priv(dev);
 903	const struct iphdr  *tiph = &tunnel->parms.iph;
 904	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 905	u8     tos = tunnel->parms.iph.tos;
 906	__be16 df = tiph->frag_off;
 907	struct rtable *rt;		/* Route to the other host */
 908	struct net_device *tdev;	/* Device to other host */
 909	unsigned int max_headroom;	/* The extra header space needed */
 910	__be32 dst = tiph->daddr;
 911	struct flowi4 fl4;
 912	int    mtu;
 913	u8 ttl;
 914	u8 protocol = IPPROTO_IPV6;
 915	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 916
 917	if (tos == 1)
 918		tos = ipv6_get_dsfield(iph6);
 919
 920	/* ISATAP (RFC4214) - must come before 6to4 */
 921	if ((dev->priv_flags & IFF_ISATAP) &&
 922	    !ipip6_tunnel_dst_find(skb, &dst, true))
 923		goto tx_error;
 924
 925	if (!dst)
 926		dst = try_6rd(tunnel, &iph6->daddr);
 927
 928	if (!dst && !ipip6_tunnel_dst_find(skb, &dst, false))
 929		goto tx_error;
 930
 931	flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
 932			   tos & INET_DSCP_MASK, RT_SCOPE_UNIVERSE,
 933			   IPPROTO_IPV6, 0, dst, tiph->saddr, 0, 0,
 934			   sock_net_uid(tunnel->net, NULL));
 935
 936	rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
 937	if (!rt) {
 938		rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
 939		if (IS_ERR(rt)) {
 940			DEV_STATS_INC(dev, tx_carrier_errors);
 941			goto tx_error_icmp;
 942		}
 943		dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
 944	}
 945
 946	if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
 947		ip_rt_put(rt);
 948		DEV_STATS_INC(dev, tx_carrier_errors);
 949		goto tx_error_icmp;
 950	}
 951	tdev = rt->dst.dev;
 952
 953	if (tdev == dev) {
 954		ip_rt_put(rt);
 955		DEV_STATS_INC(dev, collisions);
 956		goto tx_error;
 957	}
 958
 959	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
 960		ip_rt_put(rt);
 961		goto tx_error;
 962	}
 963
 964	if (df) {
 965		mtu = dst_mtu(&rt->dst) - t_hlen;
 966
 967		if (mtu < IPV4_MIN_MTU) {
 968			DEV_STATS_INC(dev, collisions);
 969			ip_rt_put(rt);
 970			goto tx_error;
 971		}
 972
 973		if (mtu < IPV6_MIN_MTU) {
 974			mtu = IPV6_MIN_MTU;
 975			df = 0;
 976		}
 977
 978		if (tunnel->parms.iph.daddr)
 979			skb_dst_update_pmtu_no_confirm(skb, mtu);
 980
 981		if (skb->len > mtu && !skb_is_gso(skb)) {
 982			icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 983			ip_rt_put(rt);
 984			goto tx_error;
 985		}
 986	}
 987
 988	if (tunnel->err_count > 0) {
 989		if (time_before(jiffies,
 990				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 991			tunnel->err_count--;
 992			dst_link_failure(skb);
 993		} else
 994			tunnel->err_count = 0;
 995	}
 996
 997	/*
 998	 * Okay, now see if we can stuff it in the buffer as-is.
 999	 */
1000	max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
1001
1002	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1003	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1004		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
1005		if (!new_skb) {
1006			ip_rt_put(rt);
1007			DEV_STATS_INC(dev, tx_dropped);
1008			kfree_skb(skb);
1009			return NETDEV_TX_OK;
1010		}
1011		if (skb->sk)
1012			skb_set_owner_w(new_skb, skb->sk);
1013		dev_kfree_skb(skb);
1014		skb = new_skb;
1015		iph6 = ipv6_hdr(skb);
1016	}
1017	ttl = tiph->ttl;
1018	if (ttl == 0)
1019		ttl = iph6->hop_limit;
1020	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
1021
1022	if (ip_tunnel_encap(skb, &tunnel->encap, &protocol, &fl4) < 0) {
1023		ip_rt_put(rt);
1024		goto tx_error;
1025	}
1026
1027	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1028
1029	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
1030		      df, !net_eq(tunnel->net, dev_net(dev)), 0);
1031	return NETDEV_TX_OK;
1032
1033tx_error_icmp:
1034	dst_link_failure(skb);
1035tx_error:
1036	kfree_skb(skb);
1037	DEV_STATS_INC(dev, tx_errors);
1038	return NETDEV_TX_OK;
1039}
1040
1041static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1042				     struct net_device *dev, u8 ipproto)
1043{
1044	struct ip_tunnel *tunnel = netdev_priv(dev);
1045	const struct iphdr  *tiph = &tunnel->parms.iph;
1046
1047	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1048		goto tx_error;
1049
1050	skb_set_inner_ipproto(skb, ipproto);
1051
1052	ip_tunnel_xmit(skb, dev, tiph, ipproto);
1053	return NETDEV_TX_OK;
1054tx_error:
1055	kfree_skb(skb);
1056	DEV_STATS_INC(dev, tx_errors);
1057	return NETDEV_TX_OK;
1058}
1059
1060static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1061				   struct net_device *dev)
1062{
1063	if (!pskb_inet_may_pull(skb))
1064		goto tx_err;
1065
1066	switch (skb->protocol) {
1067	case htons(ETH_P_IP):
1068		sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1069		break;
1070	case htons(ETH_P_IPV6):
1071		ipip6_tunnel_xmit(skb, dev);
1072		break;
1073#if IS_ENABLED(CONFIG_MPLS)
1074	case htons(ETH_P_MPLS_UC):
1075		sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1076		break;
1077#endif
1078	default:
1079		goto tx_err;
1080	}
1081
1082	return NETDEV_TX_OK;
1083
1084tx_err:
1085	DEV_STATS_INC(dev, tx_errors);
1086	kfree_skb(skb);
1087	return NETDEV_TX_OK;
1088
1089}
1090
1091static void ipip6_tunnel_bind_dev(struct net_device *dev)
1092{
1093	struct ip_tunnel *tunnel = netdev_priv(dev);
1094	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1095	struct net_device *tdev = NULL;
1096	int hlen = LL_MAX_HEADER;
1097	const struct iphdr *iph;
1098	struct flowi4 fl4;
1099
1100	iph = &tunnel->parms.iph;
1101
1102	if (iph->daddr) {
1103		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1104							  NULL,
1105							  iph->daddr, iph->saddr,
1106							  0, 0,
1107							  IPPROTO_IPV6,
1108							  iph->tos & INET_DSCP_MASK,
1109							  tunnel->parms.link);
1110
1111		if (!IS_ERR(rt)) {
1112			tdev = rt->dst.dev;
1113			ip_rt_put(rt);
1114		}
1115		dev->flags |= IFF_POINTOPOINT;
1116	}
1117
1118	if (!tdev && tunnel->parms.link)
1119		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1120
1121	if (tdev && !netif_is_l3_master(tdev)) {
1122		int mtu;
1123
1124		mtu = tdev->mtu - t_hlen;
1125		if (mtu < IPV6_MIN_MTU)
1126			mtu = IPV6_MIN_MTU;
1127		WRITE_ONCE(dev->mtu, mtu);
1128		hlen = tdev->hard_header_len + tdev->needed_headroom;
1129	}
1130	dev->needed_headroom = t_hlen + hlen;
1131}
1132
1133static void ipip6_tunnel_update(struct ip_tunnel *t,
1134				struct ip_tunnel_parm_kern *p,
1135				__u32 fwmark)
1136{
1137	struct net *net = t->net;
1138	struct sit_net *sitn = net_generic(net, sit_net_id);
1139
1140	ipip6_tunnel_unlink(sitn, t);
1141	synchronize_net();
1142	t->parms.iph.saddr = p->iph.saddr;
1143	t->parms.iph.daddr = p->iph.daddr;
1144	__dev_addr_set(t->dev, &p->iph.saddr, 4);
1145	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1146	ipip6_tunnel_link(sitn, t);
1147	t->parms.iph.ttl = p->iph.ttl;
1148	t->parms.iph.tos = p->iph.tos;
1149	t->parms.iph.frag_off = p->iph.frag_off;
1150	if (t->parms.link != p->link || t->fwmark != fwmark) {
1151		t->parms.link = p->link;
1152		t->fwmark = fwmark;
1153		ipip6_tunnel_bind_dev(t->dev);
1154	}
1155	dst_cache_reset(&t->dst_cache);
1156	netdev_state_change(t->dev);
1157}
1158
1159#ifdef CONFIG_IPV6_SIT_6RD
1160static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1161				   struct ip_tunnel_6rd *ip6rd)
1162{
1163	struct in6_addr prefix;
1164	__be32 relay_prefix;
1165
1166	if (ip6rd->relay_prefixlen > 32 ||
1167	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1168		return -EINVAL;
1169
1170	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1171	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1172		return -EINVAL;
1173	if (ip6rd->relay_prefixlen)
1174		relay_prefix = ip6rd->relay_prefix &
1175			       htonl(0xffffffffUL <<
1176				     (32 - ip6rd->relay_prefixlen));
1177	else
1178		relay_prefix = 0;
1179	if (relay_prefix != ip6rd->relay_prefix)
1180		return -EINVAL;
1181
1182	t->ip6rd.prefix = prefix;
1183	t->ip6rd.relay_prefix = relay_prefix;
1184	t->ip6rd.prefixlen = ip6rd->prefixlen;
1185	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1186	dst_cache_reset(&t->dst_cache);
1187	netdev_state_change(t->dev);
1188	return 0;
1189}
1190
1191static int
1192ipip6_tunnel_get6rd(struct net_device *dev, struct ip_tunnel_parm __user *data)
1193{
1194	struct ip_tunnel *t = netdev_priv(dev);
1195	struct ip_tunnel_parm_kern p;
1196	struct ip_tunnel_6rd ip6rd;
1197
1198	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1199		if (!ip_tunnel_parm_from_user(&p, data))
1200			return -EFAULT;
1201		t = ipip6_tunnel_locate(t->net, &p, 0);
1202	}
1203	if (!t)
1204		t = netdev_priv(dev);
1205
1206	ip6rd.prefix = t->ip6rd.prefix;
1207	ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1208	ip6rd.prefixlen = t->ip6rd.prefixlen;
1209	ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1210	if (copy_to_user(data, &ip6rd, sizeof(ip6rd)))
1211		return -EFAULT;
1212	return 0;
1213}
1214
1215static int
1216ipip6_tunnel_6rdctl(struct net_device *dev, struct ip_tunnel_6rd __user *data,
1217		    int cmd)
1218{
1219	struct ip_tunnel *t = netdev_priv(dev);
1220	struct ip_tunnel_6rd ip6rd;
1221	int err;
1222
1223	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1224		return -EPERM;
1225	if (copy_from_user(&ip6rd, data, sizeof(ip6rd)))
1226		return -EFAULT;
1227
1228	if (cmd != SIOCDEL6RD) {
1229		err = ipip6_tunnel_update_6rd(t, &ip6rd);
1230		if (err < 0)
1231			return err;
1232	} else
1233		ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev));
1234	return 0;
1235}
1236
1237#endif /* CONFIG_IPV6_SIT_6RD */
1238
1239static bool ipip6_valid_ip_proto(u8 ipproto)
1240{
1241	return ipproto == IPPROTO_IPV6 ||
1242		ipproto == IPPROTO_IPIP ||
1243#if IS_ENABLED(CONFIG_MPLS)
1244		ipproto == IPPROTO_MPLS ||
1245#endif
1246		ipproto == 0;
1247}
1248
1249static int
1250__ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm_kern *p)
1251{
1252	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1253		return -EPERM;
1254
1255	if (!ipip6_valid_ip_proto(p->iph.protocol))
1256		return -EINVAL;
1257	if (p->iph.version != 4 ||
1258	    p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
1259		return -EINVAL;
1260
1261	if (p->iph.ttl)
1262		p->iph.frag_off |= htons(IP_DF);
1263	return 0;
1264}
1265
1266static int
1267ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1268{
1269	struct ip_tunnel *t = netdev_priv(dev);
1270
1271	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
1272		t = ipip6_tunnel_locate(t->net, p, 0);
1273	if (!t)
1274		t = netdev_priv(dev);
1275	memcpy(p, &t->parms, sizeof(*p));
1276	return 0;
1277}
1278
1279static int
1280ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1281{
1282	struct ip_tunnel *t = netdev_priv(dev);
1283	int err;
1284
1285	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1286	if (err)
1287		return err;
1288
1289	t = ipip6_tunnel_locate(t->net, p, 1);
1290	if (!t)
1291		return -ENOBUFS;
1292	return 0;
1293}
1294
1295static int
1296ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1297{
1298	struct ip_tunnel *t = netdev_priv(dev);
1299	int err;
1300
1301	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1302	if (err)
1303		return err;
1304
1305	t = ipip6_tunnel_locate(t->net, p, 0);
1306	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1307		if (!t)
1308			return -ENOENT;
1309	} else {
1310		if (t) {
1311			if (t->dev != dev)
1312				return -EEXIST;
1313		} else {
1314			if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) ||
1315			    (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr))
1316				return -EINVAL;
1317			t = netdev_priv(dev);
1318		}
1319
1320		ipip6_tunnel_update(t, p, t->fwmark);
1321	}
1322
1323	return 0;
1324}
1325
1326static int
1327ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1328{
1329	struct ip_tunnel *t = netdev_priv(dev);
1330
1331	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1332		return -EPERM;
1333
1334	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1335		t = ipip6_tunnel_locate(t->net, p, 0);
1336		if (!t)
1337			return -ENOENT;
1338		if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev))
1339			return -EPERM;
1340		dev = t->dev;
1341	}
1342	unregister_netdevice(dev);
1343	return 0;
1344}
1345
1346static int
1347ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm_kern *p,
1348		 int cmd)
1349{
1350	switch (cmd) {
1351	case SIOCGETTUNNEL:
1352		return ipip6_tunnel_get(dev, p);
1353	case SIOCADDTUNNEL:
1354		return ipip6_tunnel_add(dev, p);
1355	case SIOCCHGTUNNEL:
1356		return ipip6_tunnel_change(dev, p);
1357	case SIOCDELTUNNEL:
1358		return ipip6_tunnel_del(dev, p);
1359	default:
1360		return -EINVAL;
1361	}
1362}
1363
1364static int
1365ipip6_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1366			    void __user *data, int cmd)
1367{
1368	switch (cmd) {
1369	case SIOCGETTUNNEL:
1370	case SIOCADDTUNNEL:
1371	case SIOCCHGTUNNEL:
1372	case SIOCDELTUNNEL:
1373		return ip_tunnel_siocdevprivate(dev, ifr, data, cmd);
1374	case SIOCGETPRL:
1375		return ipip6_tunnel_get_prl(dev, data);
1376	case SIOCADDPRL:
1377	case SIOCDELPRL:
1378	case SIOCCHGPRL:
1379		return ipip6_tunnel_prl_ctl(dev, data, cmd);
1380#ifdef CONFIG_IPV6_SIT_6RD
1381	case SIOCGET6RD:
1382		return ipip6_tunnel_get6rd(dev, data);
1383	case SIOCADD6RD:
1384	case SIOCCHG6RD:
1385	case SIOCDEL6RD:
1386		return ipip6_tunnel_6rdctl(dev, data, cmd);
1387#endif
1388	default:
1389		return -EINVAL;
1390	}
1391}
1392
1393static const struct net_device_ops ipip6_netdev_ops = {
1394	.ndo_init	= ipip6_tunnel_init,
1395	.ndo_uninit	= ipip6_tunnel_uninit,
1396	.ndo_start_xmit	= sit_tunnel_xmit,
1397	.ndo_siocdevprivate = ipip6_tunnel_siocdevprivate,
1398	.ndo_get_iflink = ip_tunnel_get_iflink,
1399	.ndo_tunnel_ctl = ipip6_tunnel_ctl,
1400};
1401
1402static void ipip6_dev_free(struct net_device *dev)
1403{
1404	struct ip_tunnel *tunnel = netdev_priv(dev);
1405
1406	dst_cache_destroy(&tunnel->dst_cache);
1407}
1408
1409#define SIT_FEATURES (NETIF_F_SG	   | \
1410		      NETIF_F_FRAGLIST	   | \
1411		      NETIF_F_HIGHDMA	   | \
1412		      NETIF_F_GSO_SOFTWARE | \
1413		      NETIF_F_HW_CSUM)
1414
1415static void ipip6_tunnel_setup(struct net_device *dev)
1416{
1417	struct ip_tunnel *tunnel = netdev_priv(dev);
1418	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1419
1420	dev->netdev_ops		= &ipip6_netdev_ops;
1421	dev->header_ops		= &ip_tunnel_header_ops;
1422	dev->needs_free_netdev	= true;
1423	dev->priv_destructor	= ipip6_dev_free;
1424
1425	dev->type		= ARPHRD_SIT;
1426	dev->mtu		= ETH_DATA_LEN - t_hlen;
1427	dev->min_mtu		= IPV6_MIN_MTU;
1428	dev->max_mtu		= IP6_MAX_MTU - t_hlen;
1429	dev->flags		= IFF_NOARP;
1430	netif_keep_dst(dev);
1431	dev->addr_len		= 4;
1432	dev->lltx		= true;
1433	dev->features		|= SIT_FEATURES;
1434	dev->hw_features	|= SIT_FEATURES;
1435	dev->pcpu_stat_type	= NETDEV_PCPU_STAT_TSTATS;
1436
1437}
1438
1439static int ipip6_tunnel_init(struct net_device *dev)
1440{
1441	struct ip_tunnel *tunnel = netdev_priv(dev);
1442	int err;
1443
1444	tunnel->dev = dev;
1445	strcpy(tunnel->parms.name, dev->name);
1446
1447	ipip6_tunnel_bind_dev(dev);
1448
1449	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1450	if (err)
1451		return err;
1452
1453	netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1454	netdev_lockdep_set_classes(dev);
1455	return 0;
1456}
1457
1458static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1459{
1460	struct ip_tunnel *tunnel = netdev_priv(dev);
1461	struct iphdr *iph = &tunnel->parms.iph;
1462	struct net *net = dev_net(dev);
1463	struct sit_net *sitn = net_generic(net, sit_net_id);
1464
1465	iph->version		= 4;
1466	iph->protocol		= IPPROTO_IPV6;
1467	iph->ihl		= 5;
1468	iph->ttl		= 64;
1469
1470	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1471}
1472
1473static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1474			  struct netlink_ext_ack *extack)
1475{
1476	u8 proto;
1477
1478	if (!data || !data[IFLA_IPTUN_PROTO])
1479		return 0;
1480
1481	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1482	if (!ipip6_valid_ip_proto(proto))
1483		return -EINVAL;
1484
1485	return 0;
1486}
1487
1488static void ipip6_netlink_parms(struct nlattr *data[],
1489				struct ip_tunnel_parm_kern *parms,
1490				__u32 *fwmark)
1491{
1492	memset(parms, 0, sizeof(*parms));
1493
1494	parms->iph.version = 4;
1495	parms->iph.protocol = IPPROTO_IPV6;
1496	parms->iph.ihl = 5;
1497	parms->iph.ttl = 64;
1498
1499	if (!data)
1500		return;
1501
1502	ip_tunnel_netlink_parms(data, parms);
1503
1504	if (data[IFLA_IPTUN_FWMARK])
1505		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1506}
1507
1508#ifdef CONFIG_IPV6_SIT_6RD
1509/* This function returns true when 6RD attributes are present in the nl msg */
1510static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1511				    struct ip_tunnel_6rd *ip6rd)
1512{
1513	bool ret = false;
1514	memset(ip6rd, 0, sizeof(*ip6rd));
1515
1516	if (!data)
1517		return ret;
1518
1519	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1520		ret = true;
1521		ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1522	}
1523
1524	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1525		ret = true;
1526		ip6rd->relay_prefix =
1527			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1528	}
1529
1530	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1531		ret = true;
1532		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1533	}
1534
1535	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1536		ret = true;
1537		ip6rd->relay_prefixlen =
1538			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1539	}
1540
1541	return ret;
1542}
1543#endif
1544
1545static int ipip6_newlink(struct net_device *dev,
1546			 struct rtnl_newlink_params *params,
1547			 struct netlink_ext_ack *extack)
1548{
1549	struct nlattr **data = params->data;
1550	struct nlattr **tb = params->tb;
1551	struct ip_tunnel *nt;
1552	struct ip_tunnel_encap ipencap;
1553#ifdef CONFIG_IPV6_SIT_6RD
1554	struct ip_tunnel_6rd ip6rd;
1555#endif
1556	struct net *net;
1557	int err;
1558
1559	net = params->link_net ? : dev_net(dev);
1560	nt = netdev_priv(dev);
1561	nt->net = net;
1562
1563	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1564		err = ip_tunnel_encap_setup(nt, &ipencap);
1565		if (err < 0)
1566			return err;
1567	}
1568
1569	ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1570
1571	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1572		return -EEXIST;
1573
1574	err = ipip6_tunnel_create(dev);
1575	if (err < 0)
1576		return err;
1577
1578	if (tb[IFLA_MTU]) {
1579		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1580
1581		if (mtu >= IPV6_MIN_MTU &&
1582		    mtu <= IP6_MAX_MTU - dev->hard_header_len)
1583			dev->mtu = mtu;
1584	}
1585
1586#ifdef CONFIG_IPV6_SIT_6RD
1587	if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1588		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1589		if (err < 0)
1590			unregister_netdevice_queue(dev, NULL);
1591	}
1592#endif
1593
1594	return err;
1595}
1596
1597static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1598			    struct nlattr *data[],
1599			    struct netlink_ext_ack *extack)
1600{
1601	struct ip_tunnel *t = netdev_priv(dev);
1602	struct ip_tunnel_encap ipencap;
1603	struct ip_tunnel_parm_kern p;
1604	struct net *net = t->net;
1605	struct sit_net *sitn = net_generic(net, sit_net_id);
1606#ifdef CONFIG_IPV6_SIT_6RD
1607	struct ip_tunnel_6rd ip6rd;
1608#endif
1609	__u32 fwmark = t->fwmark;
1610	int err;
1611
1612	if (dev == sitn->fb_tunnel_dev)
1613		return -EINVAL;
1614
1615	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1616		err = ip_tunnel_encap_setup(t, &ipencap);
1617		if (err < 0)
1618			return err;
1619	}
1620
1621	ipip6_netlink_parms(data, &p, &fwmark);
1622
1623	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1624	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1625		return -EINVAL;
1626
1627	t = ipip6_tunnel_locate(net, &p, 0);
1628
1629	if (t) {
1630		if (t->dev != dev)
1631			return -EEXIST;
1632	} else
1633		t = netdev_priv(dev);
1634
1635	ipip6_tunnel_update(t, &p, fwmark);
1636
1637#ifdef CONFIG_IPV6_SIT_6RD
1638	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1639		return ipip6_tunnel_update_6rd(t, &ip6rd);
1640#endif
1641
1642	return 0;
1643}
1644
1645static size_t ipip6_get_size(const struct net_device *dev)
1646{
1647	return
1648		/* IFLA_IPTUN_LINK */
1649		nla_total_size(4) +
1650		/* IFLA_IPTUN_LOCAL */
1651		nla_total_size(4) +
1652		/* IFLA_IPTUN_REMOTE */
1653		nla_total_size(4) +
1654		/* IFLA_IPTUN_TTL */
1655		nla_total_size(1) +
1656		/* IFLA_IPTUN_TOS */
1657		nla_total_size(1) +
1658		/* IFLA_IPTUN_PMTUDISC */
1659		nla_total_size(1) +
1660		/* IFLA_IPTUN_FLAGS */
1661		nla_total_size(2) +
1662		/* IFLA_IPTUN_PROTO */
1663		nla_total_size(1) +
1664#ifdef CONFIG_IPV6_SIT_6RD
1665		/* IFLA_IPTUN_6RD_PREFIX */
1666		nla_total_size(sizeof(struct in6_addr)) +
1667		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1668		nla_total_size(4) +
1669		/* IFLA_IPTUN_6RD_PREFIXLEN */
1670		nla_total_size(2) +
1671		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1672		nla_total_size(2) +
1673#endif
1674		/* IFLA_IPTUN_ENCAP_TYPE */
1675		nla_total_size(2) +
1676		/* IFLA_IPTUN_ENCAP_FLAGS */
1677		nla_total_size(2) +
1678		/* IFLA_IPTUN_ENCAP_SPORT */
1679		nla_total_size(2) +
1680		/* IFLA_IPTUN_ENCAP_DPORT */
1681		nla_total_size(2) +
1682		/* IFLA_IPTUN_FWMARK */
1683		nla_total_size(4) +
1684		0;
1685}
1686
1687static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1688{
1689	struct ip_tunnel *tunnel = netdev_priv(dev);
1690	struct ip_tunnel_parm_kern *parm = &tunnel->parms;
1691
1692	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1693	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1694	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1695	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1696	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1697	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1698		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1699	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1700	    nla_put_be16(skb, IFLA_IPTUN_FLAGS,
1701			 ip_tunnel_flags_to_be16(parm->i_flags)) ||
1702	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1703		goto nla_put_failure;
1704
1705#ifdef CONFIG_IPV6_SIT_6RD
1706	if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1707			     &tunnel->ip6rd.prefix) ||
1708	    nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1709			    tunnel->ip6rd.relay_prefix) ||
1710	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1711			tunnel->ip6rd.prefixlen) ||
1712	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1713			tunnel->ip6rd.relay_prefixlen))
1714		goto nla_put_failure;
1715#endif
1716
1717	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1718			tunnel->encap.type) ||
1719	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1720			tunnel->encap.sport) ||
1721	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1722			tunnel->encap.dport) ||
1723	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1724			tunnel->encap.flags))
1725		goto nla_put_failure;
1726
1727	return 0;
1728
1729nla_put_failure:
1730	return -EMSGSIZE;
1731}
1732
1733static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1734	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1735	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1736	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1737	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1738	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1739	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1740	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1741	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1742#ifdef CONFIG_IPV6_SIT_6RD
1743	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1744	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1745	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1746	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1747#endif
1748	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1749	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1750	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1751	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1752	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
1753};
1754
1755static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1756{
1757	struct net *net = dev_net(dev);
1758	struct sit_net *sitn = net_generic(net, sit_net_id);
1759
1760	if (dev != sitn->fb_tunnel_dev)
1761		unregister_netdevice_queue(dev, head);
1762}
1763
1764static struct rtnl_link_ops sit_link_ops __read_mostly = {
1765	.kind		= "sit",
1766	.maxtype	= IFLA_IPTUN_MAX,
1767	.policy		= ipip6_policy,
1768	.priv_size	= sizeof(struct ip_tunnel),
1769	.setup		= ipip6_tunnel_setup,
1770	.validate	= ipip6_validate,
1771	.newlink	= ipip6_newlink,
1772	.changelink	= ipip6_changelink,
1773	.get_size	= ipip6_get_size,
1774	.fill_info	= ipip6_fill_info,
1775	.dellink	= ipip6_dellink,
1776	.get_link_net	= ip_tunnel_get_link_net,
1777};
1778
1779static struct xfrm_tunnel sit_handler __read_mostly = {
1780	.handler	=	ipip6_rcv,
1781	.err_handler	=	ipip6_err,
1782	.priority	=	1,
1783};
1784
1785static struct xfrm_tunnel ipip_handler __read_mostly = {
1786	.handler	=	ipip_rcv,
1787	.err_handler	=	ipip6_err,
1788	.priority	=	2,
1789};
1790
1791#if IS_ENABLED(CONFIG_MPLS)
1792static struct xfrm_tunnel mplsip_handler __read_mostly = {
1793	.handler	=	mplsip_rcv,
1794	.err_handler	=	ipip6_err,
1795	.priority	=	2,
1796};
1797#endif
1798
1799static void __net_exit sit_exit_rtnl_net(struct net *net, struct list_head *head)
1800{
1801	struct sit_net *sitn = net_generic(net, sit_net_id);
1802	struct net_device *dev, *aux;
1803	int prio;
1804
1805	for_each_netdev_safe(net, dev, aux)
1806		if (dev->rtnl_link_ops == &sit_link_ops)
1807			unregister_netdevice_queue(dev, head);
1808
1809	for (prio = 0; prio < 4; prio++) {
1810		int h;
1811		for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1812			struct ip_tunnel *t;
1813
1814			t = rtnl_net_dereference(net, sitn->tunnels[prio][h]);
1815			while (t) {
1816				/* If dev is in the same netns, it has already
1817				 * been added to the list by the previous loop.
1818				 */
1819				if (!net_eq(dev_net(t->dev), net))
1820					unregister_netdevice_queue(t->dev, head);
1821
1822				t = rtnl_net_dereference(net, t->next);
1823			}
1824		}
1825	}
1826}
1827
1828static int __net_init sit_init_net(struct net *net)
1829{
1830	struct sit_net *sitn = net_generic(net, sit_net_id);
1831	struct ip_tunnel *t;
1832	int err;
1833
1834	sitn->tunnels[0] = sitn->tunnels_wc;
1835	sitn->tunnels[1] = sitn->tunnels_l;
1836	sitn->tunnels[2] = sitn->tunnels_r;
1837	sitn->tunnels[3] = sitn->tunnels_r_l;
1838
1839	if (!net_has_fallback_tunnels(net))
1840		return 0;
1841
1842	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1843					   NET_NAME_UNKNOWN,
1844					   ipip6_tunnel_setup);
1845	if (!sitn->fb_tunnel_dev) {
1846		err = -ENOMEM;
1847		goto err_alloc_dev;
1848	}
1849	dev_net_set(sitn->fb_tunnel_dev, net);
1850	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1851	/* FB netdevice is special: we have one, and only one per netns.
1852	 * Allowing to move it to another netns is clearly unsafe.
1853	 */
1854	sitn->fb_tunnel_dev->netns_immutable = true;
1855
1856	t = netdev_priv(sitn->fb_tunnel_dev);
1857	t->net = net;
1858
1859	err = register_netdev(sitn->fb_tunnel_dev);
1860	if (err)
1861		goto err_reg_dev;
1862
1863	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1864	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1865
1866	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1867	return 0;
1868
1869err_reg_dev:
1870	free_netdev(sitn->fb_tunnel_dev);
1871err_alloc_dev:
1872	return err;
1873}
1874
1875static struct pernet_operations sit_net_ops = {
1876	.init = sit_init_net,
1877	.exit_rtnl = sit_exit_rtnl_net,
1878	.id   = &sit_net_id,
1879	.size = sizeof(struct sit_net),
1880};
1881
1882static void __exit sit_cleanup(void)
1883{
1884	rtnl_link_unregister(&sit_link_ops);
1885	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1886	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1887#if IS_ENABLED(CONFIG_MPLS)
1888	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1889#endif
1890
1891	unregister_pernet_device(&sit_net_ops);
1892	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1893}
1894
1895static int __init sit_init(void)
1896{
1897	int err;
1898
1899	pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1900
1901	err = register_pernet_device(&sit_net_ops);
1902	if (err < 0)
1903		return err;
1904	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1905	if (err < 0) {
1906		pr_info("%s: can't register ip6ip4\n", __func__);
1907		goto xfrm_tunnel_failed;
1908	}
1909	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1910	if (err < 0) {
1911		pr_info("%s: can't register ip4ip4\n", __func__);
1912		goto xfrm_tunnel4_failed;
1913	}
1914#if IS_ENABLED(CONFIG_MPLS)
1915	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1916	if (err < 0) {
1917		pr_info("%s: can't register mplsip\n", __func__);
1918		goto xfrm_tunnel_mpls_failed;
1919	}
1920#endif
1921	err = rtnl_link_register(&sit_link_ops);
1922	if (err < 0)
1923		goto rtnl_link_failed;
1924
1925out:
1926	return err;
1927
1928rtnl_link_failed:
1929#if IS_ENABLED(CONFIG_MPLS)
1930	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1931xfrm_tunnel_mpls_failed:
1932#endif
1933	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1934xfrm_tunnel4_failed:
1935	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1936xfrm_tunnel_failed:
1937	unregister_pernet_device(&sit_net_ops);
1938	goto out;
1939}
1940
1941module_init(sit_init);
1942module_exit(sit_cleanup);
1943MODULE_DESCRIPTION("IPv6-in-IPv4 tunnel SIT driver");
1944MODULE_LICENSE("GPL");
1945MODULE_ALIAS_RTNL_LINK("sit");
1946MODULE_ALIAS_NETDEV("sit0");