Linux Audio

Check our new training course

Loading...
   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
  55/*
  56   This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  57
  58   For comments look at net/ipv4/ip_gre.c --ANK
  59 */
  60
  61#define IP6_SIT_HASH_SIZE  16
  62#define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  63
  64static bool log_ecn_error = true;
  65module_param(log_ecn_error, bool, 0644);
  66MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  67
  68static int ipip6_tunnel_init(struct net_device *dev);
  69static void ipip6_tunnel_setup(struct net_device *dev);
  70static void ipip6_dev_free(struct net_device *dev);
  71static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
  72		      __be32 *v4dst);
  73static struct rtnl_link_ops sit_link_ops __read_mostly;
  74
  75static unsigned int sit_net_id __read_mostly;
  76struct sit_net {
  77	struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
  78	struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
  79	struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE];
  80	struct ip_tunnel __rcu *tunnels_wc[1];
  81	struct ip_tunnel __rcu **tunnels[4];
  82
  83	struct net_device *fb_tunnel_dev;
  84};
  85
  86/*
  87 * Must be invoked with rcu_read_lock
  88 */
  89static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
  90					     struct net_device *dev,
  91					     __be32 remote, __be32 local,
  92					     int sifindex)
  93{
  94	unsigned int h0 = HASH(remote);
  95	unsigned int h1 = HASH(local);
  96	struct ip_tunnel *t;
  97	struct sit_net *sitn = net_generic(net, sit_net_id);
  98	int ifindex = dev ? dev->ifindex : 0;
  99
 100	for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 101		if (local == t->parms.iph.saddr &&
 102		    remote == t->parms.iph.daddr &&
 103		    (!dev || !t->parms.link || ifindex == t->parms.link ||
 104		     sifindex == 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 || ifindex == t->parms.link ||
 111		     sifindex == t->parms.link) &&
 112		    (t->dev->flags & IFF_UP))
 113			return t;
 114	}
 115	for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 116		if (local == t->parms.iph.saddr &&
 117		    (!dev || !t->parms.link || ifindex == t->parms.link ||
 118		     sifindex == t->parms.link) &&
 119		    (t->dev->flags & IFF_UP))
 120			return t;
 121	}
 122	t = rcu_dereference(sitn->tunnels_wc[0]);
 123	if (t && (t->dev->flags & IFF_UP))
 124		return t;
 125	return NULL;
 126}
 127
 128static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 129		struct ip_tunnel_parm *parms)
 130{
 131	__be32 remote = parms->iph.daddr;
 132	__be32 local = parms->iph.saddr;
 133	unsigned int h = 0;
 134	int prio = 0;
 135
 136	if (remote) {
 137		prio |= 2;
 138		h ^= HASH(remote);
 139	}
 140	if (local) {
 141		prio |= 1;
 142		h ^= HASH(local);
 143	}
 144	return &sitn->tunnels[prio][h];
 145}
 146
 147static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 148		struct ip_tunnel *t)
 149{
 150	return __ipip6_bucket(sitn, &t->parms);
 151}
 152
 153static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 154{
 155	struct ip_tunnel __rcu **tp;
 156	struct ip_tunnel *iter;
 157
 158	for (tp = ipip6_bucket(sitn, t);
 159	     (iter = rtnl_dereference(*tp)) != NULL;
 160	     tp = &iter->next) {
 161		if (t == iter) {
 162			rcu_assign_pointer(*tp, t->next);
 163			break;
 164		}
 165	}
 166}
 167
 168static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 169{
 170	struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 171
 172	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 173	rcu_assign_pointer(*tp, t);
 174}
 175
 176static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 177{
 178#ifdef CONFIG_IPV6_SIT_6RD
 179	struct ip_tunnel *t = netdev_priv(dev);
 180
 181	if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
 182		ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 183		t->ip6rd.relay_prefix = 0;
 184		t->ip6rd.prefixlen = 16;
 185		t->ip6rd.relay_prefixlen = 0;
 186	} else {
 187		struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 188		memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 189	}
 190#endif
 191}
 192
 193static int ipip6_tunnel_create(struct net_device *dev)
 194{
 195	struct ip_tunnel *t = netdev_priv(dev);
 196	struct net *net = dev_net(dev);
 197	struct sit_net *sitn = net_generic(net, sit_net_id);
 198	int err;
 199
 200	memcpy(dev->dev_addr, &t->parms.iph.saddr, 4);
 201	memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
 202
 203	if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 204		dev->priv_flags |= IFF_ISATAP;
 205
 206	dev->rtnl_link_ops = &sit_link_ops;
 207
 208	err = register_netdevice(dev);
 209	if (err < 0)
 210		goto out;
 211
 212	ipip6_tunnel_clone_6rd(dev, sitn);
 213
 214	dev_hold(dev);
 215
 216	ipip6_tunnel_link(sitn, t);
 217	return 0;
 218
 219out:
 220	return err;
 221}
 222
 223static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 224		struct ip_tunnel_parm *parms, int create)
 225{
 226	__be32 remote = parms->iph.daddr;
 227	__be32 local = parms->iph.saddr;
 228	struct ip_tunnel *t, *nt;
 229	struct ip_tunnel __rcu **tp;
 230	struct net_device *dev;
 231	char name[IFNAMSIZ];
 232	struct sit_net *sitn = net_generic(net, sit_net_id);
 233
 234	for (tp = __ipip6_bucket(sitn, parms);
 235	    (t = rtnl_dereference(*tp)) != NULL;
 236	     tp = &t->next) {
 237		if (local == t->parms.iph.saddr &&
 238		    remote == t->parms.iph.daddr &&
 239		    parms->link == t->parms.link) {
 240			if (create)
 241				return NULL;
 242			else
 243				return t;
 244		}
 245	}
 246	if (!create)
 247		goto failed;
 248
 249	if (parms->name[0]) {
 250		if (!dev_valid_name(parms->name))
 251			goto failed;
 252		strlcpy(name, parms->name, IFNAMSIZ);
 253	} else {
 254		strcpy(name, "sit%d");
 255	}
 256	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 257			   ipip6_tunnel_setup);
 258	if (!dev)
 259		return NULL;
 260
 261	dev_net_set(dev, net);
 262
 263	nt = netdev_priv(dev);
 264
 265	nt->parms = *parms;
 266	if (ipip6_tunnel_create(dev) < 0)
 267		goto failed_free;
 268
 269	return nt;
 270
 271failed_free:
 272	free_netdev(dev);
 273failed:
 274	return NULL;
 275}
 276
 277#define for_each_prl_rcu(start)			\
 278	for (prl = rcu_dereference(start);	\
 279	     prl;				\
 280	     prl = rcu_dereference(prl->next))
 281
 282static struct ip_tunnel_prl_entry *
 283__ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 284{
 285	struct ip_tunnel_prl_entry *prl;
 286
 287	for_each_prl_rcu(t->prl)
 288		if (prl->addr == addr)
 289			break;
 290	return prl;
 291
 292}
 293
 294static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
 295				struct ip_tunnel_prl __user *a)
 296{
 297	struct ip_tunnel_prl kprl, *kp;
 298	struct ip_tunnel_prl_entry *prl;
 299	unsigned int cmax, c = 0, ca, len;
 300	int ret = 0;
 301
 302	if (copy_from_user(&kprl, a, sizeof(kprl)))
 303		return -EFAULT;
 304	cmax = kprl.datalen / sizeof(kprl);
 305	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 306		cmax = 1;
 307
 308	/* For simple GET or for root users,
 309	 * we try harder to allocate.
 310	 */
 311	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 312		kcalloc(cmax, sizeof(*kp), GFP_KERNEL | __GFP_NOWARN) :
 313		NULL;
 314
 315	rcu_read_lock();
 316
 317	ca = t->prl_count < cmax ? t->prl_count : cmax;
 318
 319	if (!kp) {
 320		/* We don't try hard to allocate much memory for
 321		 * non-root users.
 322		 * For root users, retry allocating enough memory for
 323		 * the answer.
 324		 */
 325		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 326		if (!kp) {
 327			ret = -ENOMEM;
 328			goto out;
 329		}
 330	}
 331
 332	c = 0;
 333	for_each_prl_rcu(t->prl) {
 334		if (c >= cmax)
 335			break;
 336		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 337			continue;
 338		kp[c].addr = prl->addr;
 339		kp[c].flags = prl->flags;
 340		c++;
 341		if (kprl.addr != htonl(INADDR_ANY))
 342			break;
 343	}
 344out:
 345	rcu_read_unlock();
 346
 347	len = sizeof(*kp) * c;
 348	ret = 0;
 349	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 350		ret = -EFAULT;
 351
 352	kfree(kp);
 353
 354	return ret;
 355}
 356
 357static int
 358ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 359{
 360	struct ip_tunnel_prl_entry *p;
 361	int err = 0;
 362
 363	if (a->addr == htonl(INADDR_ANY))
 364		return -EINVAL;
 365
 366	ASSERT_RTNL();
 367
 368	for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 369		if (p->addr == a->addr) {
 370			if (chg) {
 371				p->flags = a->flags;
 372				goto out;
 373			}
 374			err = -EEXIST;
 375			goto out;
 376		}
 377	}
 378
 379	if (chg) {
 380		err = -ENXIO;
 381		goto out;
 382	}
 383
 384	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 385	if (!p) {
 386		err = -ENOBUFS;
 387		goto out;
 388	}
 389
 390	p->next = t->prl;
 391	p->addr = a->addr;
 392	p->flags = a->flags;
 393	t->prl_count++;
 394	rcu_assign_pointer(t->prl, p);
 395out:
 396	return err;
 397}
 398
 399static void prl_list_destroy_rcu(struct rcu_head *head)
 400{
 401	struct ip_tunnel_prl_entry *p, *n;
 402
 403	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 404	do {
 405		n = rcu_dereference_protected(p->next, 1);
 406		kfree(p);
 407		p = n;
 408	} while (p);
 409}
 410
 411static int
 412ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 413{
 414	struct ip_tunnel_prl_entry *x;
 415	struct ip_tunnel_prl_entry __rcu **p;
 416	int err = 0;
 417
 418	ASSERT_RTNL();
 419
 420	if (a && a->addr != htonl(INADDR_ANY)) {
 421		for (p = &t->prl;
 422		     (x = rtnl_dereference(*p)) != NULL;
 423		     p = &x->next) {
 424			if (x->addr == a->addr) {
 425				*p = x->next;
 426				kfree_rcu(x, rcu_head);
 427				t->prl_count--;
 428				goto out;
 429			}
 430		}
 431		err = -ENXIO;
 432	} else {
 433		x = rtnl_dereference(t->prl);
 434		if (x) {
 435			t->prl_count = 0;
 436			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 437			t->prl = NULL;
 438		}
 439	}
 440out:
 441	return err;
 442}
 443
 444static int
 445isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 446{
 447	struct ip_tunnel_prl_entry *p;
 448	int ok = 1;
 449
 450	rcu_read_lock();
 451	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 452	if (p) {
 453		if (p->flags & PRL_DEFAULT)
 454			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 455		else
 456			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 457	} else {
 458		const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 459
 460		if (ipv6_addr_is_isatap(addr6) &&
 461		    (addr6->s6_addr32[3] == iph->saddr) &&
 462		    ipv6_chk_prefix(addr6, t->dev))
 463			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 464		else
 465			ok = 0;
 466	}
 467	rcu_read_unlock();
 468	return ok;
 469}
 470
 471static void ipip6_tunnel_uninit(struct net_device *dev)
 472{
 473	struct ip_tunnel *tunnel = netdev_priv(dev);
 474	struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
 475
 476	if (dev == sitn->fb_tunnel_dev) {
 477		RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 478	} else {
 479		ipip6_tunnel_unlink(sitn, tunnel);
 480		ipip6_tunnel_del_prl(tunnel, NULL);
 481	}
 482	dst_cache_reset(&tunnel->dst_cache);
 483	dev_put(dev);
 484}
 485
 486static int ipip6_err(struct sk_buff *skb, u32 info)
 487{
 488	const struct iphdr *iph = (const struct iphdr *)skb->data;
 489	const int type = icmp_hdr(skb)->type;
 490	const int code = icmp_hdr(skb)->code;
 491	unsigned int data_len = 0;
 492	struct ip_tunnel *t;
 493	int sifindex;
 494	int err;
 495
 496	switch (type) {
 497	default:
 498	case ICMP_PARAMETERPROB:
 499		return 0;
 500
 501	case ICMP_DEST_UNREACH:
 502		switch (code) {
 503		case ICMP_SR_FAILED:
 504			/* Impossible event. */
 505			return 0;
 506		default:
 507			/* All others are translated to HOST_UNREACH.
 508			   rfc2003 contains "deep thoughts" about NET_UNREACH,
 509			   I believe they are just ether pollution. --ANK
 510			 */
 511			break;
 512		}
 513		break;
 514	case ICMP_TIME_EXCEEDED:
 515		if (code != ICMP_EXC_TTL)
 516			return 0;
 517		data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
 518		break;
 519	case ICMP_REDIRECT:
 520		break;
 521	}
 522
 523	err = -ENOENT;
 524
 525	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 526	t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 527				iph->daddr, iph->saddr, sifindex);
 528	if (!t)
 529		goto out;
 530
 531	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 532		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 533				 t->parms.link, iph->protocol);
 534		err = 0;
 535		goto out;
 536	}
 537	if (type == ICMP_REDIRECT) {
 538		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
 539			      iph->protocol);
 540		err = 0;
 541		goto out;
 542	}
 543
 544	err = 0;
 545	if (__in6_dev_get(skb->dev) &&
 546	    !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
 547		goto out;
 548
 549	if (t->parms.iph.daddr == 0)
 550		goto out;
 551
 552	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 553		goto out;
 554
 555	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 556		t->err_count++;
 557	else
 558		t->err_count = 1;
 559	t->err_time = jiffies;
 560out:
 561	return err;
 562}
 563
 564static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 565				  const struct in6_addr *v6addr)
 566{
 567	__be32 v4embed = 0;
 568	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 569		return true;
 570	return false;
 571}
 572
 573/* Checks if an address matches an address on the tunnel interface.
 574 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 575 * Long story:
 576 * This function is called after we considered the packet as spoofed
 577 * in is_spoofed_6rd.
 578 * We may have a router that is doing NAT for proto 41 packets
 579 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 580 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 581 * function will return true, dropping the packet.
 582 * But, we can still check if is spoofed against the IP
 583 * addresses associated with the interface.
 584 */
 585static bool only_dnatted(const struct ip_tunnel *tunnel,
 586	const struct in6_addr *v6dst)
 587{
 588	int prefix_len;
 589
 590#ifdef CONFIG_IPV6_SIT_6RD
 591	prefix_len = tunnel->ip6rd.prefixlen + 32
 592		- tunnel->ip6rd.relay_prefixlen;
 593#else
 594	prefix_len = 48;
 595#endif
 596	return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 597}
 598
 599/* Returns true if a packet is spoofed */
 600static bool packet_is_spoofed(struct sk_buff *skb,
 601			      const struct iphdr *iph,
 602			      struct ip_tunnel *tunnel)
 603{
 604	const struct ipv6hdr *ipv6h;
 605
 606	if (tunnel->dev->priv_flags & IFF_ISATAP) {
 607		if (!isatap_chksrc(skb, iph, tunnel))
 608			return true;
 609
 610		return false;
 611	}
 612
 613	if (tunnel->dev->flags & IFF_POINTOPOINT)
 614		return false;
 615
 616	ipv6h = ipv6_hdr(skb);
 617
 618	if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 619		net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 620				     &iph->saddr, &ipv6h->saddr,
 621				     &iph->daddr, &ipv6h->daddr);
 622		return true;
 623	}
 624
 625	if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 626		return false;
 627
 628	if (only_dnatted(tunnel, &ipv6h->daddr))
 629		return false;
 630
 631	net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 632			     &iph->saddr, &ipv6h->saddr,
 633			     &iph->daddr, &ipv6h->daddr);
 634	return true;
 635}
 636
 637static int ipip6_rcv(struct sk_buff *skb)
 638{
 639	const struct iphdr *iph = ip_hdr(skb);
 640	struct ip_tunnel *tunnel;
 641	int sifindex;
 642	int err;
 643
 644	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 645	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 646				     iph->saddr, iph->daddr, sifindex);
 647	if (tunnel) {
 648		struct pcpu_sw_netstats *tstats;
 649
 650		if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 651		    tunnel->parms.iph.protocol != 0)
 652			goto out;
 653
 654		skb->mac_header = skb->network_header;
 655		skb_reset_network_header(skb);
 656		IPCB(skb)->flags = 0;
 657		skb->dev = tunnel->dev;
 658
 659		if (packet_is_spoofed(skb, iph, tunnel)) {
 660			tunnel->dev->stats.rx_errors++;
 661			goto out;
 662		}
 663
 664		if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
 665		    !net_eq(tunnel->net, dev_net(tunnel->dev))))
 666			goto out;
 667
 668		/* skb can be uncloned in iptunnel_pull_header, so
 669		 * old iph is no longer valid
 670		 */
 671		iph = (const struct iphdr *)skb_mac_header(skb);
 672		err = IP_ECN_decapsulate(iph, skb);
 673		if (unlikely(err)) {
 674			if (log_ecn_error)
 675				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 676						     &iph->saddr, iph->tos);
 677			if (err > 1) {
 678				++tunnel->dev->stats.rx_frame_errors;
 679				++tunnel->dev->stats.rx_errors;
 680				goto out;
 681			}
 682		}
 683
 684		tstats = this_cpu_ptr(tunnel->dev->tstats);
 685		u64_stats_update_begin(&tstats->syncp);
 686		tstats->rx_packets++;
 687		tstats->rx_bytes += skb->len;
 688		u64_stats_update_end(&tstats->syncp);
 689
 690		netif_rx(skb);
 691
 692		return 0;
 693	}
 694
 695	/* no tunnel matched,  let upstream know, ipsec may handle it */
 696	return 1;
 697out:
 698	kfree_skb(skb);
 699	return 0;
 700}
 701
 702static const struct tnl_ptk_info ipip_tpi = {
 703	/* no tunnel info required for ipip. */
 704	.proto = htons(ETH_P_IP),
 705};
 706
 707#if IS_ENABLED(CONFIG_MPLS)
 708static const struct tnl_ptk_info mplsip_tpi = {
 709	/* no tunnel info required for mplsip. */
 710	.proto = htons(ETH_P_MPLS_UC),
 711};
 712#endif
 713
 714static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
 715{
 716	const struct iphdr *iph;
 717	struct ip_tunnel *tunnel;
 718	int sifindex;
 719
 720	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 721
 722	iph = ip_hdr(skb);
 723	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 724				     iph->saddr, iph->daddr, sifindex);
 725	if (tunnel) {
 726		const struct tnl_ptk_info *tpi;
 727
 728		if (tunnel->parms.iph.protocol != ipproto &&
 729		    tunnel->parms.iph.protocol != 0)
 730			goto drop;
 731
 732		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 733			goto drop;
 734#if IS_ENABLED(CONFIG_MPLS)
 735		if (ipproto == IPPROTO_MPLS)
 736			tpi = &mplsip_tpi;
 737		else
 738#endif
 739			tpi = &ipip_tpi;
 740		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 741			goto drop;
 742		return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 743	}
 744
 745	return 1;
 746
 747drop:
 748	kfree_skb(skb);
 749	return 0;
 750}
 751
 752static int ipip_rcv(struct sk_buff *skb)
 753{
 754	return sit_tunnel_rcv(skb, IPPROTO_IPIP);
 755}
 756
 757#if IS_ENABLED(CONFIG_MPLS)
 758static int mplsip_rcv(struct sk_buff *skb)
 759{
 760	return sit_tunnel_rcv(skb, IPPROTO_MPLS);
 761}
 762#endif
 763
 764/*
 765 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 766 * stores the embedded IPv4 address in v4dst and returns true.
 767 */
 768static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 769		      __be32 *v4dst)
 770{
 771#ifdef CONFIG_IPV6_SIT_6RD
 772	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 773			      tunnel->ip6rd.prefixlen)) {
 774		unsigned int pbw0, pbi0;
 775		int pbi1;
 776		u32 d;
 777
 778		pbw0 = tunnel->ip6rd.prefixlen >> 5;
 779		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 780
 781		d = tunnel->ip6rd.relay_prefixlen < 32 ?
 782			(ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 783		    tunnel->ip6rd.relay_prefixlen : 0;
 784
 785		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 786		if (pbi1 > 0)
 787			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 788			     (32 - pbi1);
 789
 790		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 791		return true;
 792	}
 793#else
 794	if (v6dst->s6_addr16[0] == htons(0x2002)) {
 795		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 796		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 797		return true;
 798	}
 799#endif
 800	return false;
 801}
 802
 803static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 804			     const struct in6_addr *v6dst)
 805{
 806	__be32 dst = 0;
 807	check_6rd(tunnel, v6dst, &dst);
 808	return dst;
 809}
 810
 811/*
 812 *	This function assumes it is being called from dev_queue_xmit()
 813 *	and that skb is filled properly by that function.
 814 */
 815
 816static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 817				     struct net_device *dev)
 818{
 819	struct ip_tunnel *tunnel = netdev_priv(dev);
 820	const struct iphdr  *tiph = &tunnel->parms.iph;
 821	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 822	u8     tos = tunnel->parms.iph.tos;
 823	__be16 df = tiph->frag_off;
 824	struct rtable *rt;		/* Route to the other host */
 825	struct net_device *tdev;	/* Device to other host */
 826	unsigned int max_headroom;	/* The extra header space needed */
 827	__be32 dst = tiph->daddr;
 828	struct flowi4 fl4;
 829	int    mtu;
 830	const struct in6_addr *addr6;
 831	int addr_type;
 832	u8 ttl;
 833	u8 protocol = IPPROTO_IPV6;
 834	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 835
 836	if (tos == 1)
 837		tos = ipv6_get_dsfield(iph6);
 838
 839	/* ISATAP (RFC4214) - must come before 6to4 */
 840	if (dev->priv_flags & IFF_ISATAP) {
 841		struct neighbour *neigh = NULL;
 842		bool do_tx_error = false;
 843
 844		if (skb_dst(skb))
 845			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 846
 847		if (!neigh) {
 848			net_dbg_ratelimited("nexthop == NULL\n");
 849			goto tx_error;
 850		}
 851
 852		addr6 = (const struct in6_addr *)&neigh->primary_key;
 853		addr_type = ipv6_addr_type(addr6);
 854
 855		if ((addr_type & IPV6_ADDR_UNICAST) &&
 856		     ipv6_addr_is_isatap(addr6))
 857			dst = addr6->s6_addr32[3];
 858		else
 859			do_tx_error = true;
 860
 861		neigh_release(neigh);
 862		if (do_tx_error)
 863			goto tx_error;
 864	}
 865
 866	if (!dst)
 867		dst = try_6rd(tunnel, &iph6->daddr);
 868
 869	if (!dst) {
 870		struct neighbour *neigh = NULL;
 871		bool do_tx_error = false;
 872
 873		if (skb_dst(skb))
 874			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 875
 876		if (!neigh) {
 877			net_dbg_ratelimited("nexthop == NULL\n");
 878			goto tx_error;
 879		}
 880
 881		addr6 = (const struct in6_addr *)&neigh->primary_key;
 882		addr_type = ipv6_addr_type(addr6);
 883
 884		if (addr_type == IPV6_ADDR_ANY) {
 885			addr6 = &ipv6_hdr(skb)->daddr;
 886			addr_type = ipv6_addr_type(addr6);
 887		}
 888
 889		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 890			dst = addr6->s6_addr32[3];
 891		else
 892			do_tx_error = true;
 893
 894		neigh_release(neigh);
 895		if (do_tx_error)
 896			goto tx_error;
 897	}
 898
 899	flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
 900			   RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
 901			   0, dst, tiph->saddr, 0, 0,
 902			   sock_net_uid(tunnel->net, NULL));
 903
 904	rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
 905	if (!rt) {
 906		rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
 907		if (IS_ERR(rt)) {
 908			dev->stats.tx_carrier_errors++;
 909			goto tx_error_icmp;
 910		}
 911		dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
 912	}
 913
 914	if (rt->rt_type != RTN_UNICAST) {
 915		ip_rt_put(rt);
 916		dev->stats.tx_carrier_errors++;
 917		goto tx_error_icmp;
 918	}
 919	tdev = rt->dst.dev;
 920
 921	if (tdev == dev) {
 922		ip_rt_put(rt);
 923		dev->stats.collisions++;
 924		goto tx_error;
 925	}
 926
 927	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
 928		ip_rt_put(rt);
 929		goto tx_error;
 930	}
 931
 932	if (df) {
 933		mtu = dst_mtu(&rt->dst) - t_hlen;
 934
 935		if (mtu < 68) {
 936			dev->stats.collisions++;
 937			ip_rt_put(rt);
 938			goto tx_error;
 939		}
 940
 941		if (mtu < IPV6_MIN_MTU) {
 942			mtu = IPV6_MIN_MTU;
 943			df = 0;
 944		}
 945
 946		if (tunnel->parms.iph.daddr)
 947			skb_dst_update_pmtu(skb, mtu);
 948
 949		if (skb->len > mtu && !skb_is_gso(skb)) {
 950			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 951			ip_rt_put(rt);
 952			goto tx_error;
 953		}
 954	}
 955
 956	if (tunnel->err_count > 0) {
 957		if (time_before(jiffies,
 958				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 959			tunnel->err_count--;
 960			dst_link_failure(skb);
 961		} else
 962			tunnel->err_count = 0;
 963	}
 964
 965	/*
 966	 * Okay, now see if we can stuff it in the buffer as-is.
 967	 */
 968	max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
 969
 970	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 971	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 972		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 973		if (!new_skb) {
 974			ip_rt_put(rt);
 975			dev->stats.tx_dropped++;
 976			kfree_skb(skb);
 977			return NETDEV_TX_OK;
 978		}
 979		if (skb->sk)
 980			skb_set_owner_w(new_skb, skb->sk);
 981		dev_kfree_skb(skb);
 982		skb = new_skb;
 983		iph6 = ipv6_hdr(skb);
 984	}
 985	ttl = tiph->ttl;
 986	if (ttl == 0)
 987		ttl = iph6->hop_limit;
 988	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 989
 990	if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0) {
 991		ip_rt_put(rt);
 992		goto tx_error;
 993	}
 994
 995	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
 996
 997	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
 998		      df, !net_eq(tunnel->net, dev_net(dev)));
 999	return NETDEV_TX_OK;
1000
1001tx_error_icmp:
1002	dst_link_failure(skb);
1003tx_error:
1004	kfree_skb(skb);
1005	dev->stats.tx_errors++;
1006	return NETDEV_TX_OK;
1007}
1008
1009static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1010				     struct net_device *dev, u8 ipproto)
1011{
1012	struct ip_tunnel *tunnel = netdev_priv(dev);
1013	const struct iphdr  *tiph = &tunnel->parms.iph;
1014
1015	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1016		goto tx_error;
1017
1018	skb_set_inner_ipproto(skb, ipproto);
1019
1020	ip_tunnel_xmit(skb, dev, tiph, ipproto);
1021	return NETDEV_TX_OK;
1022tx_error:
1023	kfree_skb(skb);
1024	dev->stats.tx_errors++;
1025	return NETDEV_TX_OK;
1026}
1027
1028static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1029				   struct net_device *dev)
1030{
1031	if (!pskb_inet_may_pull(skb))
1032		goto tx_err;
1033
1034	switch (skb->protocol) {
1035	case htons(ETH_P_IP):
1036		sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1037		break;
1038	case htons(ETH_P_IPV6):
1039		ipip6_tunnel_xmit(skb, dev);
1040		break;
1041#if IS_ENABLED(CONFIG_MPLS)
1042	case htons(ETH_P_MPLS_UC):
1043		sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1044		break;
1045#endif
1046	default:
1047		goto tx_err;
1048	}
1049
1050	return NETDEV_TX_OK;
1051
1052tx_err:
1053	dev->stats.tx_errors++;
1054	kfree_skb(skb);
1055	return NETDEV_TX_OK;
1056
1057}
1058
1059static void ipip6_tunnel_bind_dev(struct net_device *dev)
1060{
1061	struct net_device *tdev = NULL;
1062	struct ip_tunnel *tunnel;
1063	const struct iphdr *iph;
1064	struct flowi4 fl4;
1065
1066	tunnel = netdev_priv(dev);
1067	iph = &tunnel->parms.iph;
1068
1069	if (iph->daddr) {
1070		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1071							  NULL,
1072							  iph->daddr, iph->saddr,
1073							  0, 0,
1074							  IPPROTO_IPV6,
1075							  RT_TOS(iph->tos),
1076							  tunnel->parms.link);
1077
1078		if (!IS_ERR(rt)) {
1079			tdev = rt->dst.dev;
1080			ip_rt_put(rt);
1081		}
1082		dev->flags |= IFF_POINTOPOINT;
1083	}
1084
1085	if (!tdev && tunnel->parms.link)
1086		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1087
1088	if (tdev && !netif_is_l3_master(tdev)) {
1089		int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1090
1091		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1092		dev->mtu = tdev->mtu - t_hlen;
1093		if (dev->mtu < IPV6_MIN_MTU)
1094			dev->mtu = IPV6_MIN_MTU;
1095	}
1096}
1097
1098static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
1099				__u32 fwmark)
1100{
1101	struct net *net = t->net;
1102	struct sit_net *sitn = net_generic(net, sit_net_id);
1103
1104	ipip6_tunnel_unlink(sitn, t);
1105	synchronize_net();
1106	t->parms.iph.saddr = p->iph.saddr;
1107	t->parms.iph.daddr = p->iph.daddr;
1108	memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1109	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1110	ipip6_tunnel_link(sitn, t);
1111	t->parms.iph.ttl = p->iph.ttl;
1112	t->parms.iph.tos = p->iph.tos;
1113	t->parms.iph.frag_off = p->iph.frag_off;
1114	if (t->parms.link != p->link || t->fwmark != fwmark) {
1115		t->parms.link = p->link;
1116		t->fwmark = fwmark;
1117		ipip6_tunnel_bind_dev(t->dev);
1118	}
1119	dst_cache_reset(&t->dst_cache);
1120	netdev_state_change(t->dev);
1121}
1122
1123#ifdef CONFIG_IPV6_SIT_6RD
1124static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1125				   struct ip_tunnel_6rd *ip6rd)
1126{
1127	struct in6_addr prefix;
1128	__be32 relay_prefix;
1129
1130	if (ip6rd->relay_prefixlen > 32 ||
1131	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1132		return -EINVAL;
1133
1134	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1135	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1136		return -EINVAL;
1137	if (ip6rd->relay_prefixlen)
1138		relay_prefix = ip6rd->relay_prefix &
1139			       htonl(0xffffffffUL <<
1140				     (32 - ip6rd->relay_prefixlen));
1141	else
1142		relay_prefix = 0;
1143	if (relay_prefix != ip6rd->relay_prefix)
1144		return -EINVAL;
1145
1146	t->ip6rd.prefix = prefix;
1147	t->ip6rd.relay_prefix = relay_prefix;
1148	t->ip6rd.prefixlen = ip6rd->prefixlen;
1149	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1150	dst_cache_reset(&t->dst_cache);
1151	netdev_state_change(t->dev);
1152	return 0;
1153}
1154#endif
1155
1156static bool ipip6_valid_ip_proto(u8 ipproto)
1157{
1158	return ipproto == IPPROTO_IPV6 ||
1159		ipproto == IPPROTO_IPIP ||
1160#if IS_ENABLED(CONFIG_MPLS)
1161		ipproto == IPPROTO_MPLS ||
1162#endif
1163		ipproto == 0;
1164}
1165
1166static int
1167ipip6_tunnel_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1168{
1169	int err = 0;
1170	struct ip_tunnel_parm p;
1171	struct ip_tunnel_prl prl;
1172	struct ip_tunnel *t = netdev_priv(dev);
1173	struct net *net = t->net;
1174	struct sit_net *sitn = net_generic(net, sit_net_id);
1175#ifdef CONFIG_IPV6_SIT_6RD
1176	struct ip_tunnel_6rd ip6rd;
1177#endif
1178
1179	switch (cmd) {
1180	case SIOCGETTUNNEL:
1181#ifdef CONFIG_IPV6_SIT_6RD
1182	case SIOCGET6RD:
1183#endif
1184		if (dev == sitn->fb_tunnel_dev) {
1185			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1186				err = -EFAULT;
1187				break;
1188			}
1189			t = ipip6_tunnel_locate(net, &p, 0);
1190			if (!t)
1191				t = netdev_priv(dev);
1192		}
1193
1194		err = -EFAULT;
1195		if (cmd == SIOCGETTUNNEL) {
1196			memcpy(&p, &t->parms, sizeof(p));
1197			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1198					 sizeof(p)))
1199				goto done;
1200#ifdef CONFIG_IPV6_SIT_6RD
1201		} else {
1202			ip6rd.prefix = t->ip6rd.prefix;
1203			ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1204			ip6rd.prefixlen = t->ip6rd.prefixlen;
1205			ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1206			if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1207					 sizeof(ip6rd)))
1208				goto done;
1209#endif
1210		}
1211		err = 0;
1212		break;
1213
1214	case SIOCADDTUNNEL:
1215	case SIOCCHGTUNNEL:
1216		err = -EPERM;
1217		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1218			goto done;
1219
1220		err = -EFAULT;
1221		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1222			goto done;
1223
1224		err = -EINVAL;
1225		if (!ipip6_valid_ip_proto(p.iph.protocol))
1226			goto done;
1227		if (p.iph.version != 4 ||
1228		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1229			goto done;
1230		if (p.iph.ttl)
1231			p.iph.frag_off |= htons(IP_DF);
1232
1233		t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1234
1235		if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1236			if (t) {
1237				if (t->dev != dev) {
1238					err = -EEXIST;
1239					break;
1240				}
1241			} else {
1242				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1243				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1244					err = -EINVAL;
1245					break;
1246				}
1247				t = netdev_priv(dev);
1248			}
1249
1250			ipip6_tunnel_update(t, &p, t->fwmark);
1251		}
1252
1253		if (t) {
1254			err = 0;
1255			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1256				err = -EFAULT;
1257		} else
1258			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1259		break;
1260
1261	case SIOCDELTUNNEL:
1262		err = -EPERM;
1263		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1264			goto done;
1265
1266		if (dev == sitn->fb_tunnel_dev) {
1267			err = -EFAULT;
1268			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1269				goto done;
1270			err = -ENOENT;
1271			t = ipip6_tunnel_locate(net, &p, 0);
1272			if (!t)
1273				goto done;
1274			err = -EPERM;
1275			if (t == netdev_priv(sitn->fb_tunnel_dev))
1276				goto done;
1277			dev = t->dev;
1278		}
1279		unregister_netdevice(dev);
1280		err = 0;
1281		break;
1282
1283	case SIOCGETPRL:
1284		err = -EINVAL;
1285		if (dev == sitn->fb_tunnel_dev)
1286			goto done;
1287		err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1288		break;
1289
1290	case SIOCADDPRL:
1291	case SIOCDELPRL:
1292	case SIOCCHGPRL:
1293		err = -EPERM;
1294		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1295			goto done;
1296		err = -EINVAL;
1297		if (dev == sitn->fb_tunnel_dev)
1298			goto done;
1299		err = -EFAULT;
1300		if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1301			goto done;
1302
1303		switch (cmd) {
1304		case SIOCDELPRL:
1305			err = ipip6_tunnel_del_prl(t, &prl);
1306			break;
1307		case SIOCADDPRL:
1308		case SIOCCHGPRL:
1309			err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1310			break;
1311		}
1312		dst_cache_reset(&t->dst_cache);
1313		netdev_state_change(dev);
1314		break;
1315
1316#ifdef CONFIG_IPV6_SIT_6RD
1317	case SIOCADD6RD:
1318	case SIOCCHG6RD:
1319	case SIOCDEL6RD:
1320		err = -EPERM;
1321		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1322			goto done;
1323
1324		err = -EFAULT;
1325		if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1326				   sizeof(ip6rd)))
1327			goto done;
1328
1329		if (cmd != SIOCDEL6RD) {
1330			err = ipip6_tunnel_update_6rd(t, &ip6rd);
1331			if (err < 0)
1332				goto done;
1333		} else
1334			ipip6_tunnel_clone_6rd(dev, sitn);
1335
1336		err = 0;
1337		break;
1338#endif
1339
1340	default:
1341		err = -EINVAL;
1342	}
1343
1344done:
1345	return err;
1346}
1347
1348static const struct net_device_ops ipip6_netdev_ops = {
1349	.ndo_init	= ipip6_tunnel_init,
1350	.ndo_uninit	= ipip6_tunnel_uninit,
1351	.ndo_start_xmit	= sit_tunnel_xmit,
1352	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1353	.ndo_get_stats64 = ip_tunnel_get_stats64,
1354	.ndo_get_iflink = ip_tunnel_get_iflink,
1355};
1356
1357static void ipip6_dev_free(struct net_device *dev)
1358{
1359	struct ip_tunnel *tunnel = netdev_priv(dev);
1360
1361	dst_cache_destroy(&tunnel->dst_cache);
1362	free_percpu(dev->tstats);
1363}
1364
1365#define SIT_FEATURES (NETIF_F_SG	   | \
1366		      NETIF_F_FRAGLIST	   | \
1367		      NETIF_F_HIGHDMA	   | \
1368		      NETIF_F_GSO_SOFTWARE | \
1369		      NETIF_F_HW_CSUM)
1370
1371static void ipip6_tunnel_setup(struct net_device *dev)
1372{
1373	struct ip_tunnel *tunnel = netdev_priv(dev);
1374	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1375
1376	dev->netdev_ops		= &ipip6_netdev_ops;
1377	dev->needs_free_netdev	= true;
1378	dev->priv_destructor	= ipip6_dev_free;
1379
1380	dev->type		= ARPHRD_SIT;
1381	dev->hard_header_len	= LL_MAX_HEADER + t_hlen;
1382	dev->mtu		= ETH_DATA_LEN - t_hlen;
1383	dev->min_mtu		= IPV6_MIN_MTU;
1384	dev->max_mtu		= IP6_MAX_MTU - t_hlen;
1385	dev->flags		= IFF_NOARP;
1386	netif_keep_dst(dev);
1387	dev->addr_len		= 4;
1388	dev->features		|= NETIF_F_LLTX;
1389	dev->features		|= SIT_FEATURES;
1390	dev->hw_features	|= SIT_FEATURES;
1391}
1392
1393static int ipip6_tunnel_init(struct net_device *dev)
1394{
1395	struct ip_tunnel *tunnel = netdev_priv(dev);
1396	int err;
1397
1398	tunnel->dev = dev;
1399	tunnel->net = dev_net(dev);
1400	strcpy(tunnel->parms.name, dev->name);
1401
1402	ipip6_tunnel_bind_dev(dev);
1403	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1404	if (!dev->tstats)
1405		return -ENOMEM;
1406
1407	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1408	if (err) {
1409		free_percpu(dev->tstats);
1410		dev->tstats = NULL;
1411		return err;
1412	}
1413
1414	return 0;
1415}
1416
1417static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1418{
1419	struct ip_tunnel *tunnel = netdev_priv(dev);
1420	struct iphdr *iph = &tunnel->parms.iph;
1421	struct net *net = dev_net(dev);
1422	struct sit_net *sitn = net_generic(net, sit_net_id);
1423
1424	iph->version		= 4;
1425	iph->protocol		= IPPROTO_IPV6;
1426	iph->ihl		= 5;
1427	iph->ttl		= 64;
1428
1429	dev_hold(dev);
1430	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1431}
1432
1433static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1434			  struct netlink_ext_ack *extack)
1435{
1436	u8 proto;
1437
1438	if (!data || !data[IFLA_IPTUN_PROTO])
1439		return 0;
1440
1441	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1442	if (!ipip6_valid_ip_proto(proto))
1443		return -EINVAL;
1444
1445	return 0;
1446}
1447
1448static void ipip6_netlink_parms(struct nlattr *data[],
1449				struct ip_tunnel_parm *parms,
1450				__u32 *fwmark)
1451{
1452	memset(parms, 0, sizeof(*parms));
1453
1454	parms->iph.version = 4;
1455	parms->iph.protocol = IPPROTO_IPV6;
1456	parms->iph.ihl = 5;
1457	parms->iph.ttl = 64;
1458
1459	if (!data)
1460		return;
1461
1462	if (data[IFLA_IPTUN_LINK])
1463		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1464
1465	if (data[IFLA_IPTUN_LOCAL])
1466		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1467
1468	if (data[IFLA_IPTUN_REMOTE])
1469		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1470
1471	if (data[IFLA_IPTUN_TTL]) {
1472		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1473		if (parms->iph.ttl)
1474			parms->iph.frag_off = htons(IP_DF);
1475	}
1476
1477	if (data[IFLA_IPTUN_TOS])
1478		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1479
1480	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1481		parms->iph.frag_off = htons(IP_DF);
1482
1483	if (data[IFLA_IPTUN_FLAGS])
1484		parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1485
1486	if (data[IFLA_IPTUN_PROTO])
1487		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1488
1489	if (data[IFLA_IPTUN_FWMARK])
1490		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1491}
1492
1493/* This function returns true when ENCAP attributes are present in the nl msg */
1494static bool ipip6_netlink_encap_parms(struct nlattr *data[],
1495				      struct ip_tunnel_encap *ipencap)
1496{
1497	bool ret = false;
1498
1499	memset(ipencap, 0, sizeof(*ipencap));
1500
1501	if (!data)
1502		return ret;
1503
1504	if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1505		ret = true;
1506		ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1507	}
1508
1509	if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1510		ret = true;
1511		ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1512	}
1513
1514	if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1515		ret = true;
1516		ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1517	}
1518
1519	if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1520		ret = true;
1521		ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1522	}
1523
1524	return ret;
1525}
1526
1527#ifdef CONFIG_IPV6_SIT_6RD
1528/* This function returns true when 6RD attributes are present in the nl msg */
1529static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1530				    struct ip_tunnel_6rd *ip6rd)
1531{
1532	bool ret = false;
1533	memset(ip6rd, 0, sizeof(*ip6rd));
1534
1535	if (!data)
1536		return ret;
1537
1538	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1539		ret = true;
1540		ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1541	}
1542
1543	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1544		ret = true;
1545		ip6rd->relay_prefix =
1546			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1547	}
1548
1549	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1550		ret = true;
1551		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1552	}
1553
1554	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1555		ret = true;
1556		ip6rd->relay_prefixlen =
1557			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1558	}
1559
1560	return ret;
1561}
1562#endif
1563
1564static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1565			 struct nlattr *tb[], struct nlattr *data[],
1566			 struct netlink_ext_ack *extack)
1567{
1568	struct net *net = dev_net(dev);
1569	struct ip_tunnel *nt;
1570	struct ip_tunnel_encap ipencap;
1571#ifdef CONFIG_IPV6_SIT_6RD
1572	struct ip_tunnel_6rd ip6rd;
1573#endif
1574	int err;
1575
1576	nt = netdev_priv(dev);
1577
1578	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1579		err = ip_tunnel_encap_setup(nt, &ipencap);
1580		if (err < 0)
1581			return err;
1582	}
1583
1584	ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1585
1586	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1587		return -EEXIST;
1588
1589	err = ipip6_tunnel_create(dev);
1590	if (err < 0)
1591		return err;
1592
1593	if (tb[IFLA_MTU]) {
1594		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1595
1596		if (mtu >= IPV6_MIN_MTU &&
1597		    mtu <= IP6_MAX_MTU - dev->hard_header_len)
1598			dev->mtu = mtu;
1599	}
1600
1601#ifdef CONFIG_IPV6_SIT_6RD
1602	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1603		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1604#endif
1605
1606	return err;
1607}
1608
1609static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1610			    struct nlattr *data[],
1611			    struct netlink_ext_ack *extack)
1612{
1613	struct ip_tunnel *t = netdev_priv(dev);
1614	struct ip_tunnel_parm p;
1615	struct ip_tunnel_encap ipencap;
1616	struct net *net = t->net;
1617	struct sit_net *sitn = net_generic(net, sit_net_id);
1618#ifdef CONFIG_IPV6_SIT_6RD
1619	struct ip_tunnel_6rd ip6rd;
1620#endif
1621	__u32 fwmark = t->fwmark;
1622	int err;
1623
1624	if (dev == sitn->fb_tunnel_dev)
1625		return -EINVAL;
1626
1627	if (ipip6_netlink_encap_parms(data, &ipencap)) {
1628		err = ip_tunnel_encap_setup(t, &ipencap);
1629		if (err < 0)
1630			return err;
1631	}
1632
1633	ipip6_netlink_parms(data, &p, &fwmark);
1634
1635	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1636	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1637		return -EINVAL;
1638
1639	t = ipip6_tunnel_locate(net, &p, 0);
1640
1641	if (t) {
1642		if (t->dev != dev)
1643			return -EEXIST;
1644	} else
1645		t = netdev_priv(dev);
1646
1647	ipip6_tunnel_update(t, &p, fwmark);
1648
1649#ifdef CONFIG_IPV6_SIT_6RD
1650	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1651		return ipip6_tunnel_update_6rd(t, &ip6rd);
1652#endif
1653
1654	return 0;
1655}
1656
1657static size_t ipip6_get_size(const struct net_device *dev)
1658{
1659	return
1660		/* IFLA_IPTUN_LINK */
1661		nla_total_size(4) +
1662		/* IFLA_IPTUN_LOCAL */
1663		nla_total_size(4) +
1664		/* IFLA_IPTUN_REMOTE */
1665		nla_total_size(4) +
1666		/* IFLA_IPTUN_TTL */
1667		nla_total_size(1) +
1668		/* IFLA_IPTUN_TOS */
1669		nla_total_size(1) +
1670		/* IFLA_IPTUN_PMTUDISC */
1671		nla_total_size(1) +
1672		/* IFLA_IPTUN_FLAGS */
1673		nla_total_size(2) +
1674		/* IFLA_IPTUN_PROTO */
1675		nla_total_size(1) +
1676#ifdef CONFIG_IPV6_SIT_6RD
1677		/* IFLA_IPTUN_6RD_PREFIX */
1678		nla_total_size(sizeof(struct in6_addr)) +
1679		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1680		nla_total_size(4) +
1681		/* IFLA_IPTUN_6RD_PREFIXLEN */
1682		nla_total_size(2) +
1683		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1684		nla_total_size(2) +
1685#endif
1686		/* IFLA_IPTUN_ENCAP_TYPE */
1687		nla_total_size(2) +
1688		/* IFLA_IPTUN_ENCAP_FLAGS */
1689		nla_total_size(2) +
1690		/* IFLA_IPTUN_ENCAP_SPORT */
1691		nla_total_size(2) +
1692		/* IFLA_IPTUN_ENCAP_DPORT */
1693		nla_total_size(2) +
1694		/* IFLA_IPTUN_FWMARK */
1695		nla_total_size(4) +
1696		0;
1697}
1698
1699static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1700{
1701	struct ip_tunnel *tunnel = netdev_priv(dev);
1702	struct ip_tunnel_parm *parm = &tunnel->parms;
1703
1704	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1705	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1706	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1707	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1708	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1709	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1710		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1711	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1712	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
1713	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1714		goto nla_put_failure;
1715
1716#ifdef CONFIG_IPV6_SIT_6RD
1717	if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1718			     &tunnel->ip6rd.prefix) ||
1719	    nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1720			    tunnel->ip6rd.relay_prefix) ||
1721	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1722			tunnel->ip6rd.prefixlen) ||
1723	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1724			tunnel->ip6rd.relay_prefixlen))
1725		goto nla_put_failure;
1726#endif
1727
1728	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1729			tunnel->encap.type) ||
1730	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1731			tunnel->encap.sport) ||
1732	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1733			tunnel->encap.dport) ||
1734	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1735			tunnel->encap.flags))
1736		goto nla_put_failure;
1737
1738	return 0;
1739
1740nla_put_failure:
1741	return -EMSGSIZE;
1742}
1743
1744static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1745	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1746	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1747	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1748	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1749	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1750	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1751	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1752	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1753#ifdef CONFIG_IPV6_SIT_6RD
1754	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1755	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1756	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1757	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1758#endif
1759	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1760	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1761	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1762	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1763	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
1764};
1765
1766static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1767{
1768	struct net *net = dev_net(dev);
1769	struct sit_net *sitn = net_generic(net, sit_net_id);
1770
1771	if (dev != sitn->fb_tunnel_dev)
1772		unregister_netdevice_queue(dev, head);
1773}
1774
1775static struct rtnl_link_ops sit_link_ops __read_mostly = {
1776	.kind		= "sit",
1777	.maxtype	= IFLA_IPTUN_MAX,
1778	.policy		= ipip6_policy,
1779	.priv_size	= sizeof(struct ip_tunnel),
1780	.setup		= ipip6_tunnel_setup,
1781	.validate	= ipip6_validate,
1782	.newlink	= ipip6_newlink,
1783	.changelink	= ipip6_changelink,
1784	.get_size	= ipip6_get_size,
1785	.fill_info	= ipip6_fill_info,
1786	.dellink	= ipip6_dellink,
1787	.get_link_net	= ip_tunnel_get_link_net,
1788};
1789
1790static struct xfrm_tunnel sit_handler __read_mostly = {
1791	.handler	=	ipip6_rcv,
1792	.err_handler	=	ipip6_err,
1793	.priority	=	1,
1794};
1795
1796static struct xfrm_tunnel ipip_handler __read_mostly = {
1797	.handler	=	ipip_rcv,
1798	.err_handler	=	ipip6_err,
1799	.priority	=	2,
1800};
1801
1802#if IS_ENABLED(CONFIG_MPLS)
1803static struct xfrm_tunnel mplsip_handler __read_mostly = {
1804	.handler	=	mplsip_rcv,
1805	.err_handler	=	ipip6_err,
1806	.priority	=	2,
1807};
1808#endif
1809
1810static void __net_exit sit_destroy_tunnels(struct net *net,
1811					   struct list_head *head)
1812{
1813	struct sit_net *sitn = net_generic(net, sit_net_id);
1814	struct net_device *dev, *aux;
1815	int prio;
1816
1817	for_each_netdev_safe(net, dev, aux)
1818		if (dev->rtnl_link_ops == &sit_link_ops)
1819			unregister_netdevice_queue(dev, head);
1820
1821	for (prio = 1; prio < 4; prio++) {
1822		int h;
1823		for (h = 0; h < IP6_SIT_HASH_SIZE; h++) {
1824			struct ip_tunnel *t;
1825
1826			t = rtnl_dereference(sitn->tunnels[prio][h]);
1827			while (t) {
1828				/* If dev is in the same netns, it has already
1829				 * been added to the list by the previous loop.
1830				 */
1831				if (!net_eq(dev_net(t->dev), net))
1832					unregister_netdevice_queue(t->dev,
1833								   head);
1834				t = rtnl_dereference(t->next);
1835			}
1836		}
1837	}
1838}
1839
1840static int __net_init sit_init_net(struct net *net)
1841{
1842	struct sit_net *sitn = net_generic(net, sit_net_id);
1843	struct ip_tunnel *t;
1844	int err;
1845
1846	sitn->tunnels[0] = sitn->tunnels_wc;
1847	sitn->tunnels[1] = sitn->tunnels_l;
1848	sitn->tunnels[2] = sitn->tunnels_r;
1849	sitn->tunnels[3] = sitn->tunnels_r_l;
1850
1851	if (!net_has_fallback_tunnels(net))
1852		return 0;
1853
1854	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1855					   NET_NAME_UNKNOWN,
1856					   ipip6_tunnel_setup);
1857	if (!sitn->fb_tunnel_dev) {
1858		err = -ENOMEM;
1859		goto err_alloc_dev;
1860	}
1861	dev_net_set(sitn->fb_tunnel_dev, net);
1862	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1863	/* FB netdevice is special: we have one, and only one per netns.
1864	 * Allowing to move it to another netns is clearly unsafe.
1865	 */
1866	sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1867
1868	err = register_netdev(sitn->fb_tunnel_dev);
1869	if (err)
1870		goto err_reg_dev;
1871
1872	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1873	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1874
1875	t = netdev_priv(sitn->fb_tunnel_dev);
1876
1877	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1878	return 0;
1879
1880err_reg_dev:
1881	ipip6_dev_free(sitn->fb_tunnel_dev);
1882	free_netdev(sitn->fb_tunnel_dev);
1883err_alloc_dev:
1884	return err;
1885}
1886
1887static void __net_exit sit_exit_batch_net(struct list_head *net_list)
1888{
1889	LIST_HEAD(list);
1890	struct net *net;
1891
1892	rtnl_lock();
1893	list_for_each_entry(net, net_list, exit_list)
1894		sit_destroy_tunnels(net, &list);
1895
1896	unregister_netdevice_many(&list);
1897	rtnl_unlock();
1898}
1899
1900static struct pernet_operations sit_net_ops = {
1901	.init = sit_init_net,
1902	.exit_batch = sit_exit_batch_net,
1903	.id   = &sit_net_id,
1904	.size = sizeof(struct sit_net),
1905};
1906
1907static void __exit sit_cleanup(void)
1908{
1909	rtnl_link_unregister(&sit_link_ops);
1910	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1911	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1912#if IS_ENABLED(CONFIG_MPLS)
1913	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1914#endif
1915
1916	unregister_pernet_device(&sit_net_ops);
1917	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1918}
1919
1920static int __init sit_init(void)
1921{
1922	int err;
1923
1924	pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1925
1926	err = register_pernet_device(&sit_net_ops);
1927	if (err < 0)
1928		return err;
1929	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1930	if (err < 0) {
1931		pr_info("%s: can't register ip6ip4\n", __func__);
1932		goto xfrm_tunnel_failed;
1933	}
1934	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1935	if (err < 0) {
1936		pr_info("%s: can't register ip4ip4\n", __func__);
1937		goto xfrm_tunnel4_failed;
1938	}
1939#if IS_ENABLED(CONFIG_MPLS)
1940	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1941	if (err < 0) {
1942		pr_info("%s: can't register mplsip\n", __func__);
1943		goto xfrm_tunnel_mpls_failed;
1944	}
1945#endif
1946	err = rtnl_link_register(&sit_link_ops);
1947	if (err < 0)
1948		goto rtnl_link_failed;
1949
1950out:
1951	return err;
1952
1953rtnl_link_failed:
1954#if IS_ENABLED(CONFIG_MPLS)
1955	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1956xfrm_tunnel_mpls_failed:
1957#endif
1958	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1959xfrm_tunnel4_failed:
1960	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1961xfrm_tunnel_failed:
1962	unregister_pernet_device(&sit_net_ops);
1963	goto out;
1964}
1965
1966module_init(sit_init);
1967module_exit(sit_cleanup);
1968MODULE_LICENSE("GPL");
1969MODULE_ALIAS_RTNL_LINK("sit");
1970MODULE_ALIAS_NETDEV("sit0");
   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->iflink == 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->iflink == 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->iflink == t->parms.link) &&
 117		    (t->dev->flags & IFF_UP))
 118			return t;
 119	}
 120	t = rcu_dereference(sitn->tunnels_wc[0]);
 121	if ((t != NULL) && (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	err = ipip6_tunnel_init(dev);
 199	if (err < 0)
 200		goto out;
 201	ipip6_tunnel_clone_6rd(dev, sitn);
 202
 203	if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 204		dev->priv_flags |= IFF_ISATAP;
 205
 206	err = register_netdevice(dev);
 207	if (err < 0)
 208		goto out;
 209
 210	strcpy(t->parms.name, dev->name);
 211	dev->rtnl_link_ops = &sit_link_ops;
 212
 213	dev_hold(dev);
 214
 215	ipip6_tunnel_link(sitn, t);
 216	return 0;
 217
 218out:
 219	return err;
 220}
 221
 222static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 223		struct ip_tunnel_parm *parms, int create)
 224{
 225	__be32 remote = parms->iph.daddr;
 226	__be32 local = parms->iph.saddr;
 227	struct ip_tunnel *t, *nt;
 228	struct ip_tunnel __rcu **tp;
 229	struct net_device *dev;
 230	char name[IFNAMSIZ];
 231	struct sit_net *sitn = net_generic(net, sit_net_id);
 232
 233	for (tp = __ipip6_bucket(sitn, parms);
 234	    (t = rtnl_dereference(*tp)) != NULL;
 235	     tp = &t->next) {
 236		if (local == t->parms.iph.saddr &&
 237		    remote == t->parms.iph.daddr &&
 238		    parms->link == t->parms.link) {
 239			if (create)
 240				return NULL;
 241			else
 242				return t;
 243		}
 244	}
 245	if (!create)
 246		goto failed;
 247
 248	if (parms->name[0])
 249		strlcpy(name, parms->name, IFNAMSIZ);
 250	else
 251		strcpy(name, "sit%d");
 252
 253	dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 254	if (dev == NULL)
 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	const struct iphdr *iph = (const struct iphdr *) skb->data;
 488	struct rt6_info *rt;
 489	struct sk_buff *skb2;
 490
 491	if (!pskb_may_pull(skb, iph->ihl * 4 + 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, iph->ihl * 4);
 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 == NULL)
 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->dev->ifindex, 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->dev->ifindex, 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 != NULL) {
 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 != NULL) {
 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
 825	if (skb->protocol != htons(ETH_P_IPV6))
 826		goto tx_error;
 827
 828	if (tos == 1)
 829		tos = ipv6_get_dsfield(iph6);
 830
 831	/* ISATAP (RFC4214) - must come before 6to4 */
 832	if (dev->priv_flags & IFF_ISATAP) {
 833		struct neighbour *neigh = NULL;
 834		bool do_tx_error = false;
 835
 836		if (skb_dst(skb))
 837			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 838
 839		if (neigh == NULL) {
 840			net_dbg_ratelimited("nexthop == NULL\n");
 841			goto tx_error;
 842		}
 843
 844		addr6 = (const struct in6_addr *)&neigh->primary_key;
 845		addr_type = ipv6_addr_type(addr6);
 846
 847		if ((addr_type & IPV6_ADDR_UNICAST) &&
 848		     ipv6_addr_is_isatap(addr6))
 849			dst = addr6->s6_addr32[3];
 850		else
 851			do_tx_error = true;
 852
 853		neigh_release(neigh);
 854		if (do_tx_error)
 855			goto tx_error;
 856	}
 857
 858	if (!dst)
 859		dst = try_6rd(tunnel, &iph6->daddr);
 860
 861	if (!dst) {
 862		struct neighbour *neigh = NULL;
 863		bool do_tx_error = false;
 864
 865		if (skb_dst(skb))
 866			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 867
 868		if (neigh == NULL) {
 869			net_dbg_ratelimited("nexthop == NULL\n");
 870			goto tx_error;
 871		}
 872
 873		addr6 = (const struct in6_addr *)&neigh->primary_key;
 874		addr_type = ipv6_addr_type(addr6);
 875
 876		if (addr_type == IPV6_ADDR_ANY) {
 877			addr6 = &ipv6_hdr(skb)->daddr;
 878			addr_type = ipv6_addr_type(addr6);
 879		}
 880
 881		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 882			dst = addr6->s6_addr32[3];
 883		else
 884			do_tx_error = true;
 885
 886		neigh_release(neigh);
 887		if (do_tx_error)
 888			goto tx_error;
 889	}
 890
 891	rt = ip_route_output_ports(tunnel->net, &fl4, NULL,
 892				   dst, tiph->saddr,
 893				   0, 0,
 894				   IPPROTO_IPV6, RT_TOS(tos),
 895				   tunnel->parms.link);
 896	if (IS_ERR(rt)) {
 897		dev->stats.tx_carrier_errors++;
 898		goto tx_error_icmp;
 899	}
 900	if (rt->rt_type != RTN_UNICAST) {
 901		ip_rt_put(rt);
 902		dev->stats.tx_carrier_errors++;
 903		goto tx_error_icmp;
 904	}
 905	tdev = rt->dst.dev;
 906
 907	if (tdev == dev) {
 908		ip_rt_put(rt);
 909		dev->stats.collisions++;
 910		goto tx_error;
 911	}
 912
 913	if (df) {
 914		mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
 915
 916		if (mtu < 68) {
 917			dev->stats.collisions++;
 918			ip_rt_put(rt);
 919			goto tx_error;
 920		}
 921
 922		if (mtu < IPV6_MIN_MTU) {
 923			mtu = IPV6_MIN_MTU;
 924			df = 0;
 925		}
 926
 927		if (tunnel->parms.iph.daddr && skb_dst(skb))
 928			skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
 929
 930		if (skb->len > mtu && !skb_is_gso(skb)) {
 931			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 932			ip_rt_put(rt);
 933			goto tx_error;
 934		}
 935	}
 936
 937	if (tunnel->err_count > 0) {
 938		if (time_before(jiffies,
 939				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 940			tunnel->err_count--;
 941			dst_link_failure(skb);
 942		} else
 943			tunnel->err_count = 0;
 944	}
 945
 946	/*
 947	 * Okay, now see if we can stuff it in the buffer as-is.
 948	 */
 949	max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 950
 951	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 952	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 953		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 954		if (!new_skb) {
 955			ip_rt_put(rt);
 956			dev->stats.tx_dropped++;
 957			kfree_skb(skb);
 958			return NETDEV_TX_OK;
 959		}
 960		if (skb->sk)
 961			skb_set_owner_w(new_skb, skb->sk);
 962		dev_kfree_skb(skb);
 963		skb = new_skb;
 964		iph6 = ipv6_hdr(skb);
 965	}
 966	ttl = tiph->ttl;
 967	if (ttl == 0)
 968		ttl = iph6->hop_limit;
 969	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 970
 971	skb = iptunnel_handle_offloads(skb, false, SKB_GSO_SIT);
 972	if (IS_ERR(skb)) {
 973		ip_rt_put(rt);
 974		goto out;
 975	}
 976
 977	err = iptunnel_xmit(skb->sk, rt, skb, fl4.saddr, fl4.daddr,
 978			    IPPROTO_IPV6, tos, ttl, df,
 979			    !net_eq(tunnel->net, dev_net(dev)));
 980	iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
 981	return NETDEV_TX_OK;
 982
 983tx_error_icmp:
 984	dst_link_failure(skb);
 985tx_error:
 986	kfree_skb(skb);
 987out:
 988	dev->stats.tx_errors++;
 989	return NETDEV_TX_OK;
 990}
 991
 992static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 993{
 994	struct ip_tunnel *tunnel = netdev_priv(dev);
 995	const struct iphdr  *tiph = &tunnel->parms.iph;
 996
 997	skb = iptunnel_handle_offloads(skb, false, SKB_GSO_IPIP);
 998	if (IS_ERR(skb))
 999		goto out;
1000
1001	ip_tunnel_xmit(skb, dev, tiph, IPPROTO_IPIP);
1002	return NETDEV_TX_OK;
1003out:
1004	dev->stats.tx_errors++;
1005	return NETDEV_TX_OK;
1006}
1007
1008static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1009				   struct net_device *dev)
1010{
1011	switch (skb->protocol) {
1012	case htons(ETH_P_IP):
1013		ipip_tunnel_xmit(skb, dev);
1014		break;
1015	case htons(ETH_P_IPV6):
1016		ipip6_tunnel_xmit(skb, dev);
1017		break;
1018	default:
1019		goto tx_err;
1020	}
1021
1022	return NETDEV_TX_OK;
1023
1024tx_err:
1025	dev->stats.tx_errors++;
1026	kfree_skb(skb);
1027	return NETDEV_TX_OK;
1028
1029}
1030
1031static void ipip6_tunnel_bind_dev(struct net_device *dev)
1032{
1033	struct net_device *tdev = NULL;
1034	struct ip_tunnel *tunnel;
1035	const struct iphdr *iph;
1036	struct flowi4 fl4;
1037
1038	tunnel = netdev_priv(dev);
1039	iph = &tunnel->parms.iph;
1040
1041	if (iph->daddr) {
1042		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1043							  NULL,
1044							  iph->daddr, iph->saddr,
1045							  0, 0,
1046							  IPPROTO_IPV6,
1047							  RT_TOS(iph->tos),
1048							  tunnel->parms.link);
1049
1050		if (!IS_ERR(rt)) {
1051			tdev = rt->dst.dev;
1052			ip_rt_put(rt);
1053		}
1054		dev->flags |= IFF_POINTOPOINT;
1055	}
1056
1057	if (!tdev && tunnel->parms.link)
1058		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1059
1060	if (tdev) {
1061		dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
1062		dev->mtu = tdev->mtu - sizeof(struct iphdr);
1063		if (dev->mtu < IPV6_MIN_MTU)
1064			dev->mtu = IPV6_MIN_MTU;
1065	}
1066	dev->iflink = tunnel->parms.link;
1067}
1068
1069static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p)
1070{
1071	struct net *net = t->net;
1072	struct sit_net *sitn = net_generic(net, sit_net_id);
1073
1074	ipip6_tunnel_unlink(sitn, t);
1075	synchronize_net();
1076	t->parms.iph.saddr = p->iph.saddr;
1077	t->parms.iph.daddr = p->iph.daddr;
1078	memcpy(t->dev->dev_addr, &p->iph.saddr, 4);
1079	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1080	ipip6_tunnel_link(sitn, t);
1081	t->parms.iph.ttl = p->iph.ttl;
1082	t->parms.iph.tos = p->iph.tos;
1083	if (t->parms.link != p->link) {
1084		t->parms.link = p->link;
1085		ipip6_tunnel_bind_dev(t->dev);
1086	}
1087	ip_tunnel_dst_reset_all(t);
1088	netdev_state_change(t->dev);
1089}
1090
1091#ifdef CONFIG_IPV6_SIT_6RD
1092static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1093				   struct ip_tunnel_6rd *ip6rd)
1094{
1095	struct in6_addr prefix;
1096	__be32 relay_prefix;
1097
1098	if (ip6rd->relay_prefixlen > 32 ||
1099	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1100		return -EINVAL;
1101
1102	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1103	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1104		return -EINVAL;
1105	if (ip6rd->relay_prefixlen)
1106		relay_prefix = ip6rd->relay_prefix &
1107			       htonl(0xffffffffUL <<
1108				     (32 - ip6rd->relay_prefixlen));
1109	else
1110		relay_prefix = 0;
1111	if (relay_prefix != ip6rd->relay_prefix)
1112		return -EINVAL;
1113
1114	t->ip6rd.prefix = prefix;
1115	t->ip6rd.relay_prefix = relay_prefix;
1116	t->ip6rd.prefixlen = ip6rd->prefixlen;
1117	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1118	ip_tunnel_dst_reset_all(t);
1119	netdev_state_change(t->dev);
1120	return 0;
1121}
1122#endif
1123
1124static int
1125ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
1126{
1127	int err = 0;
1128	struct ip_tunnel_parm p;
1129	struct ip_tunnel_prl prl;
1130	struct ip_tunnel *t = netdev_priv(dev);
1131	struct net *net = t->net;
1132	struct sit_net *sitn = net_generic(net, sit_net_id);
1133#ifdef CONFIG_IPV6_SIT_6RD
1134	struct ip_tunnel_6rd ip6rd;
1135#endif
1136
1137	switch (cmd) {
1138	case SIOCGETTUNNEL:
1139#ifdef CONFIG_IPV6_SIT_6RD
1140	case SIOCGET6RD:
1141#endif
1142		if (dev == sitn->fb_tunnel_dev) {
1143			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1144				err = -EFAULT;
1145				break;
1146			}
1147			t = ipip6_tunnel_locate(net, &p, 0);
1148			if (t == NULL)
1149				t = netdev_priv(dev);
1150		}
1151
1152		err = -EFAULT;
1153		if (cmd == SIOCGETTUNNEL) {
1154			memcpy(&p, &t->parms, sizeof(p));
1155			if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
1156					 sizeof(p)))
1157				goto done;
1158#ifdef CONFIG_IPV6_SIT_6RD
1159		} else {
1160			ip6rd.prefix = t->ip6rd.prefix;
1161			ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1162			ip6rd.prefixlen = t->ip6rd.prefixlen;
1163			ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1164			if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
1165					 sizeof(ip6rd)))
1166				goto done;
1167#endif
1168		}
1169		err = 0;
1170		break;
1171
1172	case SIOCADDTUNNEL:
1173	case SIOCCHGTUNNEL:
1174		err = -EPERM;
1175		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1176			goto done;
1177
1178		err = -EFAULT;
1179		if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1180			goto done;
1181
1182		err = -EINVAL;
1183		if (p.iph.protocol != IPPROTO_IPV6 &&
1184		    p.iph.protocol != IPPROTO_IPIP &&
1185		    p.iph.protocol != 0)
1186			goto done;
1187		if (p.iph.version != 4 ||
1188		    p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
1189			goto done;
1190		if (p.iph.ttl)
1191			p.iph.frag_off |= htons(IP_DF);
1192
1193		t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
1194
1195		if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
1196			if (t != NULL) {
1197				if (t->dev != dev) {
1198					err = -EEXIST;
1199					break;
1200				}
1201			} else {
1202				if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
1203				    (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
1204					err = -EINVAL;
1205					break;
1206				}
1207				t = netdev_priv(dev);
1208			}
1209
1210			ipip6_tunnel_update(t, &p);
1211		}
1212
1213		if (t) {
1214			err = 0;
1215			if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
1216				err = -EFAULT;
1217		} else
1218			err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1219		break;
1220
1221	case SIOCDELTUNNEL:
1222		err = -EPERM;
1223		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1224			goto done;
1225
1226		if (dev == sitn->fb_tunnel_dev) {
1227			err = -EFAULT;
1228			if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1229				goto done;
1230			err = -ENOENT;
1231			if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1232				goto done;
1233			err = -EPERM;
1234			if (t == netdev_priv(sitn->fb_tunnel_dev))
1235				goto done;
1236			dev = t->dev;
1237		}
1238		unregister_netdevice(dev);
1239		err = 0;
1240		break;
1241
1242	case SIOCGETPRL:
1243		err = -EINVAL;
1244		if (dev == sitn->fb_tunnel_dev)
1245			goto done;
1246		err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1247		break;
1248
1249	case SIOCADDPRL:
1250	case SIOCDELPRL:
1251	case SIOCCHGPRL:
1252		err = -EPERM;
1253		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1254			goto done;
1255		err = -EINVAL;
1256		if (dev == sitn->fb_tunnel_dev)
1257			goto done;
1258		err = -EFAULT;
1259		if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1260			goto done;
1261
1262		switch (cmd) {
1263		case SIOCDELPRL:
1264			err = ipip6_tunnel_del_prl(t, &prl);
1265			break;
1266		case SIOCADDPRL:
1267		case SIOCCHGPRL:
1268			err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1269			break;
1270		}
1271		ip_tunnel_dst_reset_all(t);
1272		netdev_state_change(dev);
1273		break;
1274
1275#ifdef CONFIG_IPV6_SIT_6RD
1276	case SIOCADD6RD:
1277	case SIOCCHG6RD:
1278	case SIOCDEL6RD:
1279		err = -EPERM;
1280		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1281			goto done;
1282
1283		err = -EFAULT;
1284		if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1285				   sizeof(ip6rd)))
1286			goto done;
1287
1288		if (cmd != SIOCDEL6RD) {
1289			err = ipip6_tunnel_update_6rd(t, &ip6rd);
1290			if (err < 0)
1291				goto done;
1292		} else
1293			ipip6_tunnel_clone_6rd(dev, sitn);
1294
1295		err = 0;
1296		break;
1297#endif
1298
1299	default:
1300		err = -EINVAL;
1301	}
1302
1303done:
1304	return err;
1305}
1306
1307static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1308{
1309	if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1310		return -EINVAL;
1311	dev->mtu = new_mtu;
1312	return 0;
1313}
1314
1315static const struct net_device_ops ipip6_netdev_ops = {
1316	.ndo_uninit	= ipip6_tunnel_uninit,
1317	.ndo_start_xmit	= sit_tunnel_xmit,
1318	.ndo_do_ioctl	= ipip6_tunnel_ioctl,
1319	.ndo_change_mtu	= ipip6_tunnel_change_mtu,
1320	.ndo_get_stats64 = ip_tunnel_get_stats64,
1321};
1322
1323static void ipip6_dev_free(struct net_device *dev)
1324{
1325	struct ip_tunnel *tunnel = netdev_priv(dev);
1326
1327	free_percpu(tunnel->dst_cache);
1328	free_percpu(dev->tstats);
1329	free_netdev(dev);
1330}
1331
1332#define SIT_FEATURES (NETIF_F_SG	   | \
1333		      NETIF_F_FRAGLIST	   | \
1334		      NETIF_F_HIGHDMA	   | \
1335		      NETIF_F_GSO_SOFTWARE | \
1336		      NETIF_F_HW_CSUM)
1337
1338static void ipip6_tunnel_setup(struct net_device *dev)
1339{
1340	dev->netdev_ops		= &ipip6_netdev_ops;
1341	dev->destructor 	= ipip6_dev_free;
1342
1343	dev->type		= ARPHRD_SIT;
1344	dev->hard_header_len 	= LL_MAX_HEADER + sizeof(struct iphdr);
1345	dev->mtu		= ETH_DATA_LEN - sizeof(struct iphdr);
1346	dev->flags		= IFF_NOARP;
1347	dev->priv_flags	       &= ~IFF_XMIT_DST_RELEASE;
1348	dev->iflink		= 0;
1349	dev->addr_len		= 4;
1350	dev->features		|= NETIF_F_LLTX;
1351	dev->features		|= SIT_FEATURES;
1352	dev->hw_features	|= SIT_FEATURES;
1353}
1354
1355static int ipip6_tunnel_init(struct net_device *dev)
1356{
1357	struct ip_tunnel *tunnel = netdev_priv(dev);
1358
1359	tunnel->dev = dev;
1360	tunnel->net = dev_net(dev);
1361
1362	memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1363	memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1364
1365	ipip6_tunnel_bind_dev(dev);
1366	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1367	if (!dev->tstats)
1368		return -ENOMEM;
1369
1370	tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1371	if (!tunnel->dst_cache) {
1372		free_percpu(dev->tstats);
1373		return -ENOMEM;
1374	}
1375
1376	return 0;
1377}
1378
1379static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1380{
1381	struct ip_tunnel *tunnel = netdev_priv(dev);
1382	struct iphdr *iph = &tunnel->parms.iph;
1383	struct net *net = dev_net(dev);
1384	struct sit_net *sitn = net_generic(net, sit_net_id);
1385
1386	tunnel->dev = dev;
1387	tunnel->net = dev_net(dev);
1388	strcpy(tunnel->parms.name, dev->name);
1389
1390	iph->version		= 4;
1391	iph->protocol		= IPPROTO_IPV6;
1392	iph->ihl		= 5;
1393	iph->ttl		= 64;
1394
1395	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1396	if (!dev->tstats)
1397		return -ENOMEM;
1398
1399	tunnel->dst_cache = alloc_percpu(struct ip_tunnel_dst);
1400	if (!tunnel->dst_cache) {
1401		free_percpu(dev->tstats);
1402		return -ENOMEM;
1403	}
1404
1405	dev_hold(dev);
1406	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1407	return 0;
1408}
1409
1410static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[])
1411{
1412	u8 proto;
1413
1414	if (!data || !data[IFLA_IPTUN_PROTO])
1415		return 0;
1416
1417	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1418	if (proto != IPPROTO_IPV6 &&
1419	    proto != IPPROTO_IPIP &&
1420	    proto != 0)
1421		return -EINVAL;
1422
1423	return 0;
1424}
1425
1426static void ipip6_netlink_parms(struct nlattr *data[],
1427				struct ip_tunnel_parm *parms)
1428{
1429	memset(parms, 0, sizeof(*parms));
1430
1431	parms->iph.version = 4;
1432	parms->iph.protocol = IPPROTO_IPV6;
1433	parms->iph.ihl = 5;
1434	parms->iph.ttl = 64;
1435
1436	if (!data)
1437		return;
1438
1439	if (data[IFLA_IPTUN_LINK])
1440		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1441
1442	if (data[IFLA_IPTUN_LOCAL])
1443		parms->iph.saddr = nla_get_be32(data[IFLA_IPTUN_LOCAL]);
1444
1445	if (data[IFLA_IPTUN_REMOTE])
1446		parms->iph.daddr = nla_get_be32(data[IFLA_IPTUN_REMOTE]);
1447
1448	if (data[IFLA_IPTUN_TTL]) {
1449		parms->iph.ttl = nla_get_u8(data[IFLA_IPTUN_TTL]);
1450		if (parms->iph.ttl)
1451			parms->iph.frag_off = htons(IP_DF);
1452	}
1453
1454	if (data[IFLA_IPTUN_TOS])
1455		parms->iph.tos = nla_get_u8(data[IFLA_IPTUN_TOS]);
1456
1457	if (!data[IFLA_IPTUN_PMTUDISC] || nla_get_u8(data[IFLA_IPTUN_PMTUDISC]))
1458		parms->iph.frag_off = htons(IP_DF);
1459
1460	if (data[IFLA_IPTUN_FLAGS])
1461		parms->i_flags = nla_get_be16(data[IFLA_IPTUN_FLAGS]);
1462
1463	if (data[IFLA_IPTUN_PROTO])
1464		parms->iph.protocol = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1465
1466}
1467
1468#ifdef CONFIG_IPV6_SIT_6RD
1469/* This function returns true when 6RD attributes are present in the nl msg */
1470static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1471				    struct ip_tunnel_6rd *ip6rd)
1472{
1473	bool ret = false;
1474	memset(ip6rd, 0, sizeof(*ip6rd));
1475
1476	if (!data)
1477		return ret;
1478
1479	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1480		ret = true;
1481		nla_memcpy(&ip6rd->prefix, data[IFLA_IPTUN_6RD_PREFIX],
1482			   sizeof(struct in6_addr));
1483	}
1484
1485	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1486		ret = true;
1487		ip6rd->relay_prefix =
1488			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1489	}
1490
1491	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1492		ret = true;
1493		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1494	}
1495
1496	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1497		ret = true;
1498		ip6rd->relay_prefixlen =
1499			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1500	}
1501
1502	return ret;
1503}
1504#endif
1505
1506static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1507			 struct nlattr *tb[], struct nlattr *data[])
1508{
1509	struct net *net = dev_net(dev);
1510	struct ip_tunnel *nt;
1511#ifdef CONFIG_IPV6_SIT_6RD
1512	struct ip_tunnel_6rd ip6rd;
1513#endif
1514	int err;
1515
1516	nt = netdev_priv(dev);
1517	ipip6_netlink_parms(data, &nt->parms);
1518
1519	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1520		return -EEXIST;
1521
1522	err = ipip6_tunnel_create(dev);
1523	if (err < 0)
1524		return err;
1525
1526#ifdef CONFIG_IPV6_SIT_6RD
1527	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1528		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1529#endif
1530
1531	return err;
1532}
1533
1534static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1535			  struct nlattr *data[])
1536{
1537	struct ip_tunnel *t = netdev_priv(dev);
1538	struct ip_tunnel_parm p;
1539	struct net *net = t->net;
1540	struct sit_net *sitn = net_generic(net, sit_net_id);
1541#ifdef CONFIG_IPV6_SIT_6RD
1542	struct ip_tunnel_6rd ip6rd;
1543#endif
1544
1545	if (dev == sitn->fb_tunnel_dev)
1546		return -EINVAL;
1547
1548	ipip6_netlink_parms(data, &p);
1549
1550	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1551	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1552		return -EINVAL;
1553
1554	t = ipip6_tunnel_locate(net, &p, 0);
1555
1556	if (t) {
1557		if (t->dev != dev)
1558			return -EEXIST;
1559	} else
1560		t = netdev_priv(dev);
1561
1562	ipip6_tunnel_update(t, &p);
1563
1564#ifdef CONFIG_IPV6_SIT_6RD
1565	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1566		return ipip6_tunnel_update_6rd(t, &ip6rd);
1567#endif
1568
1569	return 0;
1570}
1571
1572static size_t ipip6_get_size(const struct net_device *dev)
1573{
1574	return
1575		/* IFLA_IPTUN_LINK */
1576		nla_total_size(4) +
1577		/* IFLA_IPTUN_LOCAL */
1578		nla_total_size(4) +
1579		/* IFLA_IPTUN_REMOTE */
1580		nla_total_size(4) +
1581		/* IFLA_IPTUN_TTL */
1582		nla_total_size(1) +
1583		/* IFLA_IPTUN_TOS */
1584		nla_total_size(1) +
1585		/* IFLA_IPTUN_PMTUDISC */
1586		nla_total_size(1) +
1587		/* IFLA_IPTUN_FLAGS */
1588		nla_total_size(2) +
1589		/* IFLA_IPTUN_PROTO */
1590		nla_total_size(1) +
1591#ifdef CONFIG_IPV6_SIT_6RD
1592		/* IFLA_IPTUN_6RD_PREFIX */
1593		nla_total_size(sizeof(struct in6_addr)) +
1594		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1595		nla_total_size(4) +
1596		/* IFLA_IPTUN_6RD_PREFIXLEN */
1597		nla_total_size(2) +
1598		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1599		nla_total_size(2) +
1600#endif
1601		0;
1602}
1603
1604static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1605{
1606	struct ip_tunnel *tunnel = netdev_priv(dev);
1607	struct ip_tunnel_parm *parm = &tunnel->parms;
1608
1609	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1610	    nla_put_be32(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1611	    nla_put_be32(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1612	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1613	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1614	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1615		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1616	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1617	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags))
1618		goto nla_put_failure;
1619
1620#ifdef CONFIG_IPV6_SIT_6RD
1621	if (nla_put(skb, IFLA_IPTUN_6RD_PREFIX, sizeof(struct in6_addr),
1622		    &tunnel->ip6rd.prefix) ||
1623	    nla_put_be32(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1624			 tunnel->ip6rd.relay_prefix) ||
1625	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1626			tunnel->ip6rd.prefixlen) ||
1627	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1628			tunnel->ip6rd.relay_prefixlen))
1629		goto nla_put_failure;
1630#endif
1631
1632	return 0;
1633
1634nla_put_failure:
1635	return -EMSGSIZE;
1636}
1637
1638static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1639	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1640	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1641	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1642	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1643	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1644	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1645	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1646	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1647#ifdef CONFIG_IPV6_SIT_6RD
1648	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1649	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1650	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1651	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1652#endif
1653};
1654
1655static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1656{
1657	struct net *net = dev_net(dev);
1658	struct sit_net *sitn = net_generic(net, sit_net_id);
1659
1660	if (dev != sitn->fb_tunnel_dev)
1661		unregister_netdevice_queue(dev, head);
1662}
1663
1664static struct rtnl_link_ops sit_link_ops __read_mostly = {
1665	.kind		= "sit",
1666	.maxtype	= IFLA_IPTUN_MAX,
1667	.policy		= ipip6_policy,
1668	.priv_size	= sizeof(struct ip_tunnel),
1669	.setup		= ipip6_tunnel_setup,
1670	.validate	= ipip6_validate,
1671	.newlink	= ipip6_newlink,
1672	.changelink	= ipip6_changelink,
1673	.get_size	= ipip6_get_size,
1674	.fill_info	= ipip6_fill_info,
1675	.dellink	= ipip6_dellink,
1676};
1677
1678static struct xfrm_tunnel sit_handler __read_mostly = {
1679	.handler	=	ipip6_rcv,
1680	.err_handler	=	ipip6_err,
1681	.priority	=	1,
1682};
1683
1684static struct xfrm_tunnel ipip_handler __read_mostly = {
1685	.handler	=	ipip_rcv,
1686	.err_handler	=	ipip6_err,
1687	.priority	=	2,
1688};
1689
1690static void __net_exit sit_destroy_tunnels(struct net *net,
1691					   struct list_head *head)
1692{
1693	struct sit_net *sitn = net_generic(net, sit_net_id);
1694	struct net_device *dev, *aux;
1695	int prio;
1696
1697	for_each_netdev_safe(net, dev, aux)
1698		if (dev->rtnl_link_ops == &sit_link_ops)
1699			unregister_netdevice_queue(dev, head);
1700
1701	for (prio = 1; prio < 4; prio++) {
1702		int h;
1703		for (h = 0; h < HASH_SIZE; h++) {
1704			struct ip_tunnel *t;
1705
1706			t = rtnl_dereference(sitn->tunnels[prio][h]);
1707			while (t != NULL) {
1708				/* If dev is in the same netns, it has already
1709				 * been added to the list by the previous loop.
1710				 */
1711				if (!net_eq(dev_net(t->dev), net))
1712					unregister_netdevice_queue(t->dev,
1713								   head);
1714				t = rtnl_dereference(t->next);
1715			}
1716		}
1717	}
1718}
1719
1720static int __net_init sit_init_net(struct net *net)
1721{
1722	struct sit_net *sitn = net_generic(net, sit_net_id);
1723	struct ip_tunnel *t;
1724	int err;
1725
1726	sitn->tunnels[0] = sitn->tunnels_wc;
1727	sitn->tunnels[1] = sitn->tunnels_l;
1728	sitn->tunnels[2] = sitn->tunnels_r;
1729	sitn->tunnels[3] = sitn->tunnels_r_l;
1730
1731	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1732					   ipip6_tunnel_setup);
1733	if (!sitn->fb_tunnel_dev) {
1734		err = -ENOMEM;
1735		goto err_alloc_dev;
1736	}
1737	dev_net_set(sitn->fb_tunnel_dev, net);
1738	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1739	/* FB netdevice is special: we have one, and only one per netns.
1740	 * Allowing to move it to another netns is clearly unsafe.
1741	 */
1742	sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1743
1744	err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1745	if (err)
1746		goto err_dev_free;
1747
1748	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1749
1750	if ((err = register_netdev(sitn->fb_tunnel_dev)))
1751		goto err_reg_dev;
1752
1753	t = netdev_priv(sitn->fb_tunnel_dev);
1754
1755	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1756	return 0;
1757
1758err_reg_dev:
1759	dev_put(sitn->fb_tunnel_dev);
1760err_dev_free:
1761	ipip6_dev_free(sitn->fb_tunnel_dev);
1762err_alloc_dev:
1763	return err;
1764}
1765
1766static void __net_exit sit_exit_net(struct net *net)
1767{
1768	LIST_HEAD(list);
1769
1770	rtnl_lock();
1771	sit_destroy_tunnels(net, &list);
1772	unregister_netdevice_many(&list);
1773	rtnl_unlock();
1774}
1775
1776static struct pernet_operations sit_net_ops = {
1777	.init = sit_init_net,
1778	.exit = sit_exit_net,
1779	.id   = &sit_net_id,
1780	.size = sizeof(struct sit_net),
1781};
1782
1783static void __exit sit_cleanup(void)
1784{
1785	rtnl_link_unregister(&sit_link_ops);
1786	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1787	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1788
1789	unregister_pernet_device(&sit_net_ops);
1790	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1791}
1792
1793static int __init sit_init(void)
1794{
1795	int err;
1796
1797	pr_info("IPv6 over IPv4 tunneling driver\n");
1798
1799	err = register_pernet_device(&sit_net_ops);
1800	if (err < 0)
1801		return err;
1802	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1803	if (err < 0) {
1804		pr_info("%s: can't register ip6ip4\n", __func__);
1805		goto xfrm_tunnel_failed;
1806	}
1807	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1808	if (err < 0) {
1809		pr_info("%s: can't register ip4ip4\n", __func__);
1810		goto xfrm_tunnel4_failed;
1811	}
1812	err = rtnl_link_register(&sit_link_ops);
1813	if (err < 0)
1814		goto rtnl_link_failed;
1815
1816out:
1817	return err;
1818
1819rtnl_link_failed:
1820	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1821xfrm_tunnel4_failed:
1822	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1823xfrm_tunnel_failed:
1824	unregister_pernet_device(&sit_net_ops);
1825	goto out;
1826}
1827
1828module_init(sit_init);
1829module_exit(sit_cleanup);
1830MODULE_LICENSE("GPL");
1831MODULE_ALIAS_NETDEV("sit0");