ip6_tunnel.c

   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 *	IPv6 tunneling device
   4 *	Linux INET6 implementation
   5 *
   6 *	Authors:
   7 *	Ville Nuorvala		<vnuorval@tcs.hut.fi>
   8 *	Yasuyuki Kozakai	<kozakai@linux-ipv6.org>
   9 *
  10 *      Based on:
  11 *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
  12 *
  13 *      RFC 2473
  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/sockios.h>
  23#include <linux/icmp.h>
  24#include <linux/if.h>
  25#include <linux/in.h>
  26#include <linux/ip.h>
  27#include <linux/net.h>
  28#include <linux/in6.h>
  29#include <linux/netdevice.h>
  30#include <linux/if_arp.h>
  31#include <linux/icmpv6.h>
  32#include <linux/init.h>
  33#include <linux/route.h>
  34#include <linux/rtnetlink.h>
  35#include <linux/netfilter_ipv6.h>
  36#include <linux/slab.h>
  37#include <linux/hash.h>
  38#include <linux/etherdevice.h>
  39
  40#include <linux/uaccess.h>
  41#include <linux/atomic.h>
  42
  43#include <net/icmp.h>
  44#include <net/ip.h>
  45#include <net/ip_tunnels.h>
  46#include <net/ipv6.h>
  47#include <net/ip6_route.h>
  48#include <net/addrconf.h>
  49#include <net/ip6_tunnel.h>
  50#include <net/xfrm.h>
  51#include <net/dsfield.h>
  52#include <net/inet_ecn.h>
  53#include <net/net_namespace.h>
  54#include <net/netns/generic.h>
  55#include <net/netdev_lock.h>
  56#include <net/dst_metadata.h>
  57#include <net/inet_dscp.h>
  58
  59MODULE_AUTHOR("Ville Nuorvala");
  60MODULE_DESCRIPTION("IPv6 tunneling device");
  61MODULE_LICENSE("GPL");
  62MODULE_ALIAS_RTNL_LINK("ip6tnl");
  63MODULE_ALIAS_NETDEV("ip6tnl0");
  64
  65#define IP6_TUNNEL_HASH_SIZE_SHIFT  5
  66#define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
  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 u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
  73{
  74	u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
  75
  76	return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
  77}
  78
  79static int ip6_tnl_dev_init(struct net_device *dev);
  80static void ip6_tnl_dev_setup(struct net_device *dev);
  81static struct rtnl_link_ops ip6_link_ops __read_mostly;
  82
  83static unsigned int ip6_tnl_net_id __read_mostly;
  84struct ip6_tnl_net {
  85	/* the IPv6 tunnel fallback device */
  86	struct net_device *fb_tnl_dev;
  87	/* lists for storing tunnels in use */
  88	struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
  89	struct ip6_tnl __rcu *tnls_wc[1];
  90	struct ip6_tnl __rcu **tnls[2];
  91	struct ip6_tnl __rcu *collect_md_tun;
  92};
  93
  94static inline int ip6_tnl_mpls_supported(void)
  95{
  96	return IS_ENABLED(CONFIG_MPLS);
  97}
  98
  99#define for_each_ip6_tunnel_rcu(start) \
 100	for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
 101
 102/**
 103 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
 104 *   @net: network namespace
 105 *   @link: ifindex of underlying interface
 106 *   @remote: the address of the tunnel exit-point
 107 *   @local: the address of the tunnel entry-point
 108 *
 109 * Return:
 110 *   tunnel matching given end-points if found,
 111 *   else fallback tunnel if its device is up,
 112 *   else %NULL
 113 **/
 114
 115static struct ip6_tnl *
 116ip6_tnl_lookup(struct net *net, int link,
 117	       const struct in6_addr *remote, const struct in6_addr *local)
 118{
 119	unsigned int hash = HASH(remote, local);
 120	struct ip6_tnl *t, *cand = NULL;
 121	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 122	struct in6_addr any;
 123
 124	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 125		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 126		    !ipv6_addr_equal(remote, &t->parms.raddr) ||
 127		    !(t->dev->flags & IFF_UP))
 128			continue;
 129
 130		if (link == t->parms.link)
 131			return t;
 132		else
 133			cand = t;
 134	}
 135
 136	memset(&any, 0, sizeof(any));
 137	hash = HASH(&any, local);
 138	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 139		if (!ipv6_addr_equal(local, &t->parms.laddr) ||
 140		    !ipv6_addr_any(&t->parms.raddr) ||
 141		    !(t->dev->flags & IFF_UP))
 142			continue;
 143
 144		if (link == t->parms.link)
 145			return t;
 146		else if (!cand)
 147			cand = t;
 148	}
 149
 150	hash = HASH(remote, &any);
 151	for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
 152		if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
 153		    !ipv6_addr_any(&t->parms.laddr) ||
 154		    !(t->dev->flags & IFF_UP))
 155			continue;
 156
 157		if (link == t->parms.link)
 158			return t;
 159		else if (!cand)
 160			cand = t;
 161	}
 162
 163	if (cand)
 164		return cand;
 165
 166	t = rcu_dereference(ip6n->collect_md_tun);
 167	if (t && t->dev->flags & IFF_UP)
 168		return t;
 169
 170	t = rcu_dereference(ip6n->tnls_wc[0]);
 171	if (t && (t->dev->flags & IFF_UP))
 172		return t;
 173
 174	return NULL;
 175}
 176
 177/**
 178 * ip6_tnl_bucket - get head of list matching given tunnel parameters
 179 *   @ip6n: the private data for ip6_vti in the netns
 180 *   @p: parameters containing tunnel end-points
 181 *
 182 * Description:
 183 *   ip6_tnl_bucket() returns the head of the list matching the
 184 *   &struct in6_addr entries laddr and raddr in @p.
 185 *
 186 * Return: head of IPv6 tunnel list
 187 **/
 188
 189static struct ip6_tnl __rcu **
 190ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
 191{
 192	const struct in6_addr *remote = &p->raddr;
 193	const struct in6_addr *local = &p->laddr;
 194	unsigned int h = 0;
 195	int prio = 0;
 196
 197	if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
 198		prio = 1;
 199		h = HASH(remote, local);
 200	}
 201	return &ip6n->tnls[prio][h];
 202}
 203
 204/**
 205 * ip6_tnl_link - add tunnel to hash table
 206 *   @ip6n: the private data for ip6_vti in the netns
 207 *   @t: tunnel to be added
 208 **/
 209
 210static void
 211ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 212{
 213	struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
 214
 215	if (t->parms.collect_md)
 216		rcu_assign_pointer(ip6n->collect_md_tun, t);
 217	rcu_assign_pointer(t->next , rtnl_dereference(*tp));
 218	rcu_assign_pointer(*tp, t);
 219}
 220
 221/**
 222 * ip6_tnl_unlink - remove tunnel from hash table
 223 *   @ip6n: the private data for ip6_vti in the netns
 224 *   @t: tunnel to be removed
 225 **/
 226
 227static void
 228ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 229{
 230	struct ip6_tnl __rcu **tp;
 231	struct ip6_tnl *iter;
 232
 233	if (t->parms.collect_md)
 234		rcu_assign_pointer(ip6n->collect_md_tun, NULL);
 235
 236	for (tp = ip6_tnl_bucket(ip6n, &t->parms);
 237	     (iter = rtnl_dereference(*tp)) != NULL;
 238	     tp = &iter->next) {
 239		if (t == iter) {
 240			rcu_assign_pointer(*tp, t->next);
 241			break;
 242		}
 243	}
 244}
 245
 246static void ip6_dev_free(struct net_device *dev)
 247{
 248	struct ip6_tnl *t = netdev_priv(dev);
 249
 250	gro_cells_destroy(&t->gro_cells);
 251	dst_cache_destroy(&t->dst_cache);
 252}
 253
 254static int ip6_tnl_create2(struct net_device *dev)
 255{
 256	struct ip6_tnl *t = netdev_priv(dev);
 257	struct ip6_tnl_net *ip6n = net_generic(t->net, ip6_tnl_net_id);
 258	int err;
 259
 260	dev->rtnl_link_ops = &ip6_link_ops;
 261	err = register_netdevice(dev);
 262	if (err < 0)
 263		goto out;
 264
 265	strcpy(t->parms.name, dev->name);
 266
 267	ip6_tnl_link(ip6n, t);
 268	return 0;
 269
 270out:
 271	return err;
 272}
 273
 274/**
 275 * ip6_tnl_create - create a new tunnel
 276 *   @net: network namespace
 277 *   @p: tunnel parameters
 278 *
 279 * Description:
 280 *   Create tunnel matching given parameters.
 281 *
 282 * Return:
 283 *   created tunnel or error pointer
 284 **/
 285
 286static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
 287{
 288	struct net_device *dev;
 289	struct ip6_tnl *t;
 290	char name[IFNAMSIZ];
 291	int err = -E2BIG;
 292
 293	if (p->name[0]) {
 294		if (!dev_valid_name(p->name))
 295			goto failed;
 296		strscpy(name, p->name, IFNAMSIZ);
 297	} else {
 298		sprintf(name, "ip6tnl%%d");
 299	}
 300	err = -ENOMEM;
 301	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 302			   ip6_tnl_dev_setup);
 303	if (!dev)
 304		goto failed;
 305
 306	dev_net_set(dev, net);
 307
 308	t = netdev_priv(dev);
 309	t->parms = *p;
 310	t->net = dev_net(dev);
 311	err = ip6_tnl_create2(dev);
 312	if (err < 0)
 313		goto failed_free;
 314
 315	return t;
 316
 317failed_free:
 318	free_netdev(dev);
 319failed:
 320	return ERR_PTR(err);
 321}
 322
 323/**
 324 * ip6_tnl_locate - find or create tunnel matching given parameters
 325 *   @net: network namespace
 326 *   @p: tunnel parameters
 327 *   @create: != 0 if allowed to create new tunnel if no match found
 328 *
 329 * Description:
 330 *   ip6_tnl_locate() first tries to locate an existing tunnel
 331 *   based on @parms. If this is unsuccessful, but @create is set a new
 332 *   tunnel device is created and registered for use.
 333 *
 334 * Return:
 335 *   matching tunnel or error pointer
 336 **/
 337
 338static struct ip6_tnl *ip6_tnl_locate(struct net *net,
 339		struct __ip6_tnl_parm *p, int create)
 340{
 341	const struct in6_addr *remote = &p->raddr;
 342	const struct in6_addr *local = &p->laddr;
 343	struct ip6_tnl __rcu **tp;
 344	struct ip6_tnl *t;
 345	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 346
 347	for (tp = ip6_tnl_bucket(ip6n, p);
 348	     (t = rtnl_dereference(*tp)) != NULL;
 349	     tp = &t->next) {
 350		if (ipv6_addr_equal(local, &t->parms.laddr) &&
 351		    ipv6_addr_equal(remote, &t->parms.raddr) &&
 352		    p->link == t->parms.link) {
 353			if (create)
 354				return ERR_PTR(-EEXIST);
 355
 356			return t;
 357		}
 358	}
 359	if (!create)
 360		return ERR_PTR(-ENODEV);
 361	return ip6_tnl_create(net, p);
 362}
 363
 364/**
 365 * ip6_tnl_dev_uninit - tunnel device uninitializer
 366 *   @dev: the device to be destroyed
 367 *
 368 * Description:
 369 *   ip6_tnl_dev_uninit() removes tunnel from its list
 370 **/
 371
 372static void
 373ip6_tnl_dev_uninit(struct net_device *dev)
 374{
 375	struct ip6_tnl *t = netdev_priv(dev);
 376	struct net *net = t->net;
 377	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 378
 379	if (dev == ip6n->fb_tnl_dev)
 380		RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
 381	else
 382		ip6_tnl_unlink(ip6n, t);
 383	dst_cache_reset(&t->dst_cache);
 384	netdev_put(dev, &t->dev_tracker);
 385}
 386
 387/**
 388 * ip6_tnl_parse_tlv_enc_lim - handle encapsulation limit option
 389 *   @skb: received socket buffer
 390 *   @raw: the ICMPv6 error message data
 391 *
 392 * Return:
 393 *   0 if none was found,
 394 *   else index to encapsulation limit
 395 **/
 396
 397__u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
 398{
 399	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
 400	unsigned int nhoff = raw - skb->data;
 401	unsigned int off = nhoff + sizeof(*ipv6h);
 402	u8 nexthdr = ipv6h->nexthdr;
 403
 404	while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
 405		struct ipv6_opt_hdr *hdr;
 406		u16 optlen;
 407
 408		if (!pskb_may_pull(skb, off + sizeof(*hdr)))
 409			break;
 410
 411		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
 412		if (nexthdr == NEXTHDR_FRAGMENT) {
 413			optlen = 8;
 414		} else if (nexthdr == NEXTHDR_AUTH) {
 415			optlen = ipv6_authlen(hdr);
 416		} else {
 417			optlen = ipv6_optlen(hdr);
 418		}
 419
 420		if (!pskb_may_pull(skb, off + optlen))
 421			break;
 422
 423		hdr = (struct ipv6_opt_hdr *)(skb->data + off);
 424		if (nexthdr == NEXTHDR_FRAGMENT) {
 425			struct frag_hdr *frag_hdr = (struct frag_hdr *)hdr;
 426
 427			if (frag_hdr->frag_off)
 428				break;
 429		}
 430		if (nexthdr == NEXTHDR_DEST) {
 431			u16 i = 2;
 432
 433			while (1) {
 434				struct ipv6_tlv_tnl_enc_lim *tel;
 435
 436				/* No more room for encapsulation limit */
 437				if (i + sizeof(*tel) > optlen)
 438					break;
 439
 440				tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
 441				/* return index of option if found and valid */
 442				if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
 443				    tel->length == 1)
 444					return i + off - nhoff;
 445				/* else jump to next option */
 446				if (tel->type)
 447					i += tel->length + 2;
 448				else
 449					i++;
 450			}
 451		}
 452		nexthdr = hdr->nexthdr;
 453		off += optlen;
 454	}
 455	return 0;
 456}
 457EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
 458
 459/* ip6_tnl_err() should handle errors in the tunnel according to the
 460 * specifications in RFC 2473.
 461 */
 462static int
 463ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
 464	    u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 465{
 466	const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
 467	struct net *net = dev_net(skb->dev);
 468	u8 rel_type = ICMPV6_DEST_UNREACH;
 469	u8 rel_code = ICMPV6_ADDR_UNREACH;
 470	__u32 rel_info = 0;
 471	struct ip6_tnl *t;
 472	int err = -ENOENT;
 473	int rel_msg = 0;
 474	u8 tproto;
 475	__u16 len;
 476
 477	/* If the packet doesn't contain the original IPv6 header we are
 478	   in trouble since we might need the source address for further
 479	   processing of the error. */
 480
 481	rcu_read_lock();
 482	t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr);
 483	if (!t)
 484		goto out;
 485
 486	tproto = READ_ONCE(t->parms.proto);
 487	if (tproto != ipproto && tproto != 0)
 488		goto out;
 489
 490	err = 0;
 491
 492	switch (*type) {
 493	case ICMPV6_DEST_UNREACH:
 494		net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
 495				    t->parms.name);
 496		rel_msg = 1;
 497		break;
 498	case ICMPV6_TIME_EXCEED:
 499		if ((*code) == ICMPV6_EXC_HOPLIMIT) {
 500			net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
 501					    t->parms.name);
 502			rel_msg = 1;
 503		}
 504		break;
 505	case ICMPV6_PARAMPROB: {
 506		struct ipv6_tlv_tnl_enc_lim *tel;
 507		__u32 teli;
 508
 509		teli = 0;
 510		if ((*code) == ICMPV6_HDR_FIELD)
 511			teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
 512
 513		if (teli && teli == *info - 2) {
 514			tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
 515			if (tel->encap_limit == 0) {
 516				net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
 517						    t->parms.name);
 518				rel_msg = 1;
 519			}
 520		} else {
 521			net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
 522					    t->parms.name);
 523		}
 524		break;
 525	}
 526	case ICMPV6_PKT_TOOBIG: {
 527		__u32 mtu;
 528
 529		ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
 530				sock_net_uid(net, NULL));
 531		mtu = *info - offset;
 532		if (mtu < IPV6_MIN_MTU)
 533			mtu = IPV6_MIN_MTU;
 534		len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
 535		if (len > mtu) {
 536			rel_type = ICMPV6_PKT_TOOBIG;
 537			rel_code = 0;
 538			rel_info = mtu;
 539			rel_msg = 1;
 540		}
 541		break;
 542	}
 543	case NDISC_REDIRECT:
 544		ip6_redirect(skb, net, skb->dev->ifindex, 0,
 545			     sock_net_uid(net, NULL));
 546		break;
 547	}
 548
 549	*type = rel_type;
 550	*code = rel_code;
 551	*info = rel_info;
 552	*msg = rel_msg;
 553
 554out:
 555	rcu_read_unlock();
 556	return err;
 557}
 558
 559static int
 560ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 561	   u8 type, u8 code, int offset, __be32 info)
 562{
 563	__u32 rel_info = ntohl(info);
 564	const struct iphdr *eiph;
 565	struct sk_buff *skb2;
 566	int err, rel_msg = 0;
 567	u8 rel_type = type;
 568	u8 rel_code = code;
 569	struct rtable *rt;
 570	struct flowi4 fl4;
 571
 572	err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
 573			  &rel_msg, &rel_info, offset);
 574	if (err < 0)
 575		return err;
 576
 577	if (rel_msg == 0)
 578		return 0;
 579
 580	switch (rel_type) {
 581	case ICMPV6_DEST_UNREACH:
 582		if (rel_code != ICMPV6_ADDR_UNREACH)
 583			return 0;
 584		rel_type = ICMP_DEST_UNREACH;
 585		rel_code = ICMP_HOST_UNREACH;
 586		break;
 587	case ICMPV6_PKT_TOOBIG:
 588		if (rel_code != 0)
 589			return 0;
 590		rel_type = ICMP_DEST_UNREACH;
 591		rel_code = ICMP_FRAG_NEEDED;
 592		break;
 593	default:
 594		return 0;
 595	}
 596
 597	if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
 598		return 0;
 599
 600	skb2 = skb_clone(skb, GFP_ATOMIC);
 601	if (!skb2)
 602		return 0;
 603
 604	/* Remove debris left by IPv6 stack. */
 605	memset(IPCB(skb2), 0, sizeof(*IPCB(skb2)));
 606
 607	skb_dst_drop(skb2);
 608
 609	skb_pull(skb2, offset);
 610	skb_reset_network_header(skb2);
 611	eiph = ip_hdr(skb2);
 612	if (eiph->version != 4 || eiph->ihl < 5)
 613		goto out;
 614
 615	/* Try to guess incoming interface */
 616	rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
 617				   0, 0, 0, IPPROTO_IPIP,
 618				   eiph->tos & INET_DSCP_MASK, 0);
 619	if (IS_ERR(rt))
 620		goto out;
 621
 622	skb2->dev = rt->dst.dev;
 623	ip_rt_put(rt);
 624
 625	/* route "incoming" packet */
 626	if (rt->rt_flags & RTCF_LOCAL) {
 627		rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
 628					   eiph->daddr, eiph->saddr, 0, 0,
 629					   IPPROTO_IPIP,
 630					   eiph->tos & INET_DSCP_MASK, 0);
 631		if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) {
 632			if (!IS_ERR(rt))
 633				ip_rt_put(rt);
 634			goto out;
 635		}
 636		skb_dst_set(skb2, &rt->dst);
 637	} else {
 638		if (ip_route_input(skb2, eiph->daddr, eiph->saddr,
 639				   ip4h_dscp(eiph), skb2->dev) ||
 640		    skb_dst_dev(skb2)->type != ARPHRD_TUNNEL6)
 641			goto out;
 642	}
 643
 644	/* change mtu on this route */
 645	if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
 646		if (rel_info > dst_mtu(skb_dst(skb2)))
 647			goto out;
 648
 649		skb_dst_update_pmtu_no_confirm(skb2, rel_info);
 650	}
 651
 652	icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 653
 654out:
 655	kfree_skb(skb2);
 656	return 0;
 657}
 658
 659static int
 660ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 661	   u8 type, u8 code, int offset, __be32 info)
 662{
 663	__u32 rel_info = ntohl(info);
 664	int err, rel_msg = 0;
 665	u8 rel_type = type;
 666	u8 rel_code = code;
 667
 668	err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
 669			  &rel_msg, &rel_info, offset);
 670	if (err < 0)
 671		return err;
 672
 673	if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
 674		struct rt6_info *rt;
 675		struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 676
 677		if (!skb2)
 678			return 0;
 679
 680		skb_dst_drop(skb2);
 681		skb_pull(skb2, offset);
 682		skb_reset_network_header(skb2);
 683
 684		/* Try to guess incoming interface */
 685		rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
 686				NULL, 0, skb2, 0);
 687
 688		if (rt && rt->dst.dev)
 689			skb2->dev = rt->dst.dev;
 690
 691		icmpv6_send(skb2, rel_type, rel_code, rel_info);
 692
 693		ip6_rt_put(rt);
 694
 695		kfree_skb(skb2);
 696	}
 697
 698	return 0;
 699}
 700
 701static int
 702mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 703	    u8 type, u8 code, int offset, __be32 info)
 704{
 705	__u32 rel_info = ntohl(info);
 706	int err, rel_msg = 0;
 707	u8 rel_type = type;
 708	u8 rel_code = code;
 709
 710	err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code,
 711			  &rel_msg, &rel_info, offset);
 712	return err;
 713}
 714
 715static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 716				       const struct ipv6hdr *ipv6h,
 717				       struct sk_buff *skb)
 718{
 719	__u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 720
 721	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 722		ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 723
 724	return IP6_ECN_decapsulate(ipv6h, skb);
 725}
 726
 727static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 728				       const struct ipv6hdr *ipv6h,
 729				       struct sk_buff *skb)
 730{
 731	if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
 732		ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 733
 734	return IP6_ECN_decapsulate(ipv6h, skb);
 735}
 736
 737static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
 738					       const struct ipv6hdr *ipv6h,
 739					       struct sk_buff *skb)
 740{
 741	/* ECN is not supported in AF_MPLS */
 742	return 0;
 743}
 744
 745__u32 ip6_tnl_get_cap(struct ip6_tnl *t,
 746			     const struct in6_addr *laddr,
 747			     const struct in6_addr *raddr)
 748{
 749	struct __ip6_tnl_parm *p = &t->parms;
 750	int ltype = ipv6_addr_type(laddr);
 751	int rtype = ipv6_addr_type(raddr);
 752	__u32 flags = 0;
 753
 754	if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
 755		flags = IP6_TNL_F_CAP_PER_PACKET;
 756	} else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 757		   rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
 758		   !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
 759		   (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
 760		if (ltype&IPV6_ADDR_UNICAST)
 761			flags |= IP6_TNL_F_CAP_XMIT;
 762		if (rtype&IPV6_ADDR_UNICAST)
 763			flags |= IP6_TNL_F_CAP_RCV;
 764	}
 765	return flags;
 766}
 767EXPORT_SYMBOL(ip6_tnl_get_cap);
 768
 769/* called with rcu_read_lock() */
 770int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
 771				  const struct in6_addr *laddr,
 772				  const struct in6_addr *raddr)
 773{
 774	struct __ip6_tnl_parm *p = &t->parms;
 775	int ret = 0;
 776	struct net *net = t->net;
 777
 778	if ((p->flags & IP6_TNL_F_CAP_RCV) ||
 779	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
 780	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
 781		struct net_device *ldev = NULL;
 782
 783		if (p->link)
 784			ldev = dev_get_by_index_rcu(net, p->link);
 785
 786		if ((ipv6_addr_is_multicast(laddr) ||
 787		     likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
 788						    0, IFA_F_TENTATIVE))) &&
 789		    ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
 790		     likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
 791						     0, IFA_F_TENTATIVE))))
 792			ret = 1;
 793	}
 794	return ret;
 795}
 796EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
 797
 798static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
 799			 const struct tnl_ptk_info *tpi,
 800			 struct metadata_dst *tun_dst,
 801			 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 802						const struct ipv6hdr *ipv6h,
 803						struct sk_buff *skb),
 804			 bool log_ecn_err)
 805{
 806	const struct ipv6hdr *ipv6h;
 807	int nh, err;
 808
 809	if (test_bit(IP_TUNNEL_CSUM_BIT, tunnel->parms.i_flags) !=
 810	    test_bit(IP_TUNNEL_CSUM_BIT, tpi->flags)) {
 811		DEV_STATS_INC(tunnel->dev, rx_crc_errors);
 812		DEV_STATS_INC(tunnel->dev, rx_errors);
 813		goto drop;
 814	}
 815
 816	if (test_bit(IP_TUNNEL_SEQ_BIT, tunnel->parms.i_flags)) {
 817		if (!test_bit(IP_TUNNEL_SEQ_BIT, tpi->flags) ||
 818		    (tunnel->i_seqno &&
 819		     (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 820			DEV_STATS_INC(tunnel->dev, rx_fifo_errors);
 821			DEV_STATS_INC(tunnel->dev, rx_errors);
 822			goto drop;
 823		}
 824		tunnel->i_seqno = ntohl(tpi->seq) + 1;
 825	}
 826
 827	skb->protocol = tpi->proto;
 828
 829	/* Warning: All skb pointers will be invalidated! */
 830	if (tunnel->dev->type == ARPHRD_ETHER) {
 831		if (!pskb_may_pull(skb, ETH_HLEN)) {
 832			DEV_STATS_INC(tunnel->dev, rx_length_errors);
 833			DEV_STATS_INC(tunnel->dev, rx_errors);
 834			goto drop;
 835		}
 836
 837		skb->protocol = eth_type_trans(skb, tunnel->dev);
 838		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 839	} else {
 840		skb->dev = tunnel->dev;
 841		skb_reset_mac_header(skb);
 842	}
 843
 844	/* Save offset of outer header relative to skb->head,
 845	 * because we are going to reset the network header to the inner header
 846	 * and might change skb->head.
 847	 */
 848	nh = skb_network_header(skb) - skb->head;
 849
 850	skb_reset_network_header(skb);
 851
 852	if (skb_vlan_inet_prepare(skb, true)) {
 853		DEV_STATS_INC(tunnel->dev, rx_length_errors);
 854		DEV_STATS_INC(tunnel->dev, rx_errors);
 855		goto drop;
 856	}
 857
 858	/* Get the outer header. */
 859	ipv6h = (struct ipv6hdr *)(skb->head + nh);
 860
 861	memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
 862
 863	__skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
 864
 865	err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
 866	if (unlikely(err)) {
 867		if (log_ecn_err)
 868			net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
 869					     &ipv6h->saddr,
 870					     ipv6_get_dsfield(ipv6h));
 871		if (err > 1) {
 872			DEV_STATS_INC(tunnel->dev, rx_frame_errors);
 873			DEV_STATS_INC(tunnel->dev, rx_errors);
 874			goto drop;
 875		}
 876	}
 877
 878	dev_sw_netstats_rx_add(tunnel->dev, skb->len);
 879
 880	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 881
 882	if (tun_dst)
 883		skb_dst_set(skb, (struct dst_entry *)tun_dst);
 884
 885	gro_cells_receive(&tunnel->gro_cells, skb);
 886	return 0;
 887
 888drop:
 889	if (tun_dst)
 890		dst_release((struct dst_entry *)tun_dst);
 891	kfree_skb(skb);
 892	return 0;
 893}
 894
 895int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
 896		const struct tnl_ptk_info *tpi,
 897		struct metadata_dst *tun_dst,
 898		bool log_ecn_err)
 899{
 900	int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 901				    const struct ipv6hdr *ipv6h,
 902				    struct sk_buff *skb);
 903
 904	dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
 905	if (tpi->proto == htons(ETH_P_IP))
 906		dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
 907
 908	return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 909			     log_ecn_err);
 910}
 911EXPORT_SYMBOL(ip6_tnl_rcv);
 912
 913static const struct tnl_ptk_info tpi_v6 = {
 914	/* no tunnel info required for ipxip6. */
 915	.proto = htons(ETH_P_IPV6),
 916};
 917
 918static const struct tnl_ptk_info tpi_v4 = {
 919	/* no tunnel info required for ipxip6. */
 920	.proto = htons(ETH_P_IP),
 921};
 922
 923static const struct tnl_ptk_info tpi_mpls = {
 924	/* no tunnel info required for mplsip6. */
 925	.proto = htons(ETH_P_MPLS_UC),
 926};
 927
 928static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
 929		      const struct tnl_ptk_info *tpi,
 930		      int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
 931						  const struct ipv6hdr *ipv6h,
 932						  struct sk_buff *skb))
 933{
 934	struct ip6_tnl *t;
 935	const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 936	struct metadata_dst *tun_dst = NULL;
 937	int ret = -1;
 938
 939	rcu_read_lock();
 940	t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr);
 941
 942	if (t) {
 943		u8 tproto = READ_ONCE(t->parms.proto);
 944
 945		if (tproto != ipproto && tproto != 0)
 946			goto drop;
 947		if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 948			goto drop;
 949		ipv6h = ipv6_hdr(skb);
 950		if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
 951			goto drop;
 952		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 953			goto drop;
 954		if (t->parms.collect_md) {
 955			IP_TUNNEL_DECLARE_FLAGS(flags) = { };
 956
 957			tun_dst = ipv6_tun_rx_dst(skb, flags, 0, 0);
 958			if (!tun_dst)
 959				goto drop;
 960		}
 961		ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
 962				    log_ecn_error);
 963	}
 964
 965	rcu_read_unlock();
 966
 967	return ret;
 968
 969drop:
 970	rcu_read_unlock();
 971	kfree_skb(skb);
 972	return 0;
 973}
 974
 975static int ip4ip6_rcv(struct sk_buff *skb)
 976{
 977	return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
 978			  ip4ip6_dscp_ecn_decapsulate);
 979}
 980
 981static int ip6ip6_rcv(struct sk_buff *skb)
 982{
 983	return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
 984			  ip6ip6_dscp_ecn_decapsulate);
 985}
 986
 987static int mplsip6_rcv(struct sk_buff *skb)
 988{
 989	return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls,
 990			  mplsip6_dscp_ecn_decapsulate);
 991}
 992
 993struct ipv6_tel_txoption {
 994	struct ipv6_txoptions ops;
 995	__u8 dst_opt[8];
 996};
 997
 998static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
 999{
1000	memset(opt, 0, sizeof(struct ipv6_tel_txoption));
1001
1002	opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
1003	opt->dst_opt[3] = 1;
1004	opt->dst_opt[4] = encap_limit;
1005	opt->dst_opt[5] = IPV6_TLV_PADN;
1006	opt->dst_opt[6] = 1;
1007
1008	opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
1009	opt->ops.opt_nflen = 8;
1010}
1011
1012/**
1013 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
1014 *   @t: the outgoing tunnel device
1015 *   @hdr: IPv6 header from the incoming packet
1016 *
1017 * Description:
1018 *   Avoid trivial tunneling loop by checking that tunnel exit-point
1019 *   doesn't match source of incoming packet.
1020 *
1021 * Return:
1022 *   1 if conflict,
1023 *   0 else
1024 **/
1025
1026static inline bool
1027ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
1028{
1029	return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
1030}
1031
1032int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
1033		     const struct in6_addr *laddr,
1034		     const struct in6_addr *raddr)
1035{
1036	struct __ip6_tnl_parm *p = &t->parms;
1037	int ret = 0;
1038	struct net *net = t->net;
1039
1040	if (t->parms.collect_md)
1041		return 1;
1042
1043	if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1044	    ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1045	     (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1046		struct net_device *ldev = NULL;
1047
1048		rcu_read_lock();
1049		if (p->link)
1050			ldev = dev_get_by_index_rcu(net, p->link);
1051
1052		if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1053						      0, IFA_F_TENTATIVE)))
1054			pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1055					    p->name);
1056		else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1057			 !ipv6_addr_is_multicast(raddr) &&
1058			 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1059							  true, 0, IFA_F_TENTATIVE)))
1060			pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n",
1061					    p->name);
1062		else
1063			ret = 1;
1064		rcu_read_unlock();
1065	}
1066	return ret;
1067}
1068EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1069
1070/**
1071 * ip6_tnl_xmit - encapsulate packet and send
1072 *   @skb: the outgoing socket buffer
1073 *   @dev: the outgoing tunnel device
1074 *   @dsfield: dscp code for outer header
1075 *   @fl6: flow of tunneled packet
1076 *   @encap_limit: encapsulation limit
1077 *   @pmtu: Path MTU is stored if packet is too big
1078 *   @proto: next header value
1079 *
1080 * Description:
1081 *   Build new header and do some sanity checks on the packet before sending
1082 *   it.
1083 *
1084 * Return:
1085 *   0 on success
1086 *   -1 fail
1087 *   %-EMSGSIZE message too big. return mtu in this case.
1088 **/
1089
1090int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1091		 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1092		 __u8 proto)
1093{
1094	struct ip6_tnl *t = netdev_priv(dev);
1095	struct net *net = t->net;
1096	struct ipv6hdr *ipv6h;
1097	struct ipv6_tel_txoption opt;
1098	struct dst_entry *dst = NULL, *ndst = NULL;
1099	struct net_device *tdev;
1100	int mtu;
1101	unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1102	unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1103	unsigned int max_headroom = psh_hlen;
1104	__be16 payload_protocol;
1105	bool use_cache = false;
1106	u8 hop_limit;
1107	int err = -1;
1108
1109	payload_protocol = skb_protocol(skb, true);
1110
1111	if (t->parms.collect_md) {
1112		hop_limit = skb_tunnel_info(skb)->key.ttl;
1113		goto route_lookup;
1114	} else {
1115		hop_limit = t->parms.hop_limit;
1116	}
1117
1118	/* NBMA tunnel */
1119	if (ipv6_addr_any(&t->parms.raddr)) {
1120		if (payload_protocol == htons(ETH_P_IPV6)) {
1121			struct in6_addr *addr6;
1122			struct neighbour *neigh;
1123			int addr_type;
1124
1125			if (!skb_dst(skb))
1126				goto tx_err_link_failure;
1127
1128			neigh = dst_neigh_lookup(skb_dst(skb),
1129						 &ipv6_hdr(skb)->daddr);
1130			if (!neigh)
1131				goto tx_err_link_failure;
1132
1133			addr6 = (struct in6_addr *)&neigh->primary_key;
1134			addr_type = ipv6_addr_type(addr6);
1135
1136			if (addr_type == IPV6_ADDR_ANY)
1137				addr6 = &ipv6_hdr(skb)->daddr;
1138
1139			memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1140			neigh_release(neigh);
1141		} else if (payload_protocol == htons(ETH_P_IP)) {
1142			const struct rtable *rt = skb_rtable(skb);
1143
1144			if (!rt)
1145				goto tx_err_link_failure;
1146
1147			if (rt->rt_gw_family == AF_INET6)
1148				memcpy(&fl6->daddr, &rt->rt_gw6, sizeof(fl6->daddr));
1149		}
1150	} else if (t->parms.proto != 0 && !(t->parms.flags &
1151					    (IP6_TNL_F_USE_ORIG_TCLASS |
1152					     IP6_TNL_F_USE_ORIG_FWMARK))) {
1153		/* enable the cache only if neither the outer protocol nor the
1154		 * routing decision depends on the current inner header value
1155		 */
1156		use_cache = true;
1157	}
1158
1159	if (use_cache)
1160		dst = dst_cache_get(&t->dst_cache);
1161
1162	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1163		goto tx_err_link_failure;
1164
1165	if (!dst) {
1166route_lookup:
1167		/* add dsfield to flowlabel for route lookup */
1168		fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1169
1170		dst = ip6_route_output(net, NULL, fl6);
1171
1172		if (dst->error)
1173			goto tx_err_link_failure;
1174		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1175		if (IS_ERR(dst)) {
1176			err = PTR_ERR(dst);
1177			dst = NULL;
1178			goto tx_err_link_failure;
1179		}
1180		if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1181		    ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1182				       &fl6->daddr, 0, &fl6->saddr))
1183			goto tx_err_link_failure;
1184		ndst = dst;
1185	}
1186
1187	tdev = dst_dev(dst);
1188
1189	if (tdev == dev) {
1190		DEV_STATS_INC(dev, collisions);
1191		net_warn_ratelimited("%s: Local routing loop detected!\n",
1192				     t->parms.name);
1193		goto tx_err_dst_release;
1194	}
1195	mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1196	if (encap_limit >= 0) {
1197		max_headroom += 8;
1198		mtu -= 8;
1199	}
1200	mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1201		       IPV6_MIN_MTU : IPV4_MIN_MTU);
1202
1203	skb_dst_update_pmtu_no_confirm(skb, mtu);
1204	if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1205		*pmtu = mtu;
1206		err = -EMSGSIZE;
1207		goto tx_err_dst_release;
1208	}
1209
1210	if (t->err_count > 0) {
1211		if (time_before(jiffies,
1212				t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1213			t->err_count--;
1214
1215			dst_link_failure(skb);
1216		} else {
1217			t->err_count = 0;
1218		}
1219	}
1220
1221	skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1222
1223	/*
1224	 * Okay, now see if we can stuff it in the buffer as-is.
1225	 */
1226	max_headroom += LL_RESERVED_SPACE(tdev);
1227
1228	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1229	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1230		struct sk_buff *new_skb;
1231
1232		new_skb = skb_realloc_headroom(skb, max_headroom);
1233		if (!new_skb)
1234			goto tx_err_dst_release;
1235
1236		if (skb->sk)
1237			skb_set_owner_w(new_skb, skb->sk);
1238		consume_skb(skb);
1239		skb = new_skb;
1240	}
1241
1242	if (t->parms.collect_md) {
1243		if (t->encap.type != TUNNEL_ENCAP_NONE)
1244			goto tx_err_dst_release;
1245	} else {
1246		if (use_cache && ndst)
1247			dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1248	}
1249	skb_dst_set(skb, dst);
1250
1251	if (hop_limit == 0) {
1252		if (payload_protocol == htons(ETH_P_IP))
1253			hop_limit = ip_hdr(skb)->ttl;
1254		else if (payload_protocol == htons(ETH_P_IPV6))
1255			hop_limit = ipv6_hdr(skb)->hop_limit;
1256		else
1257			hop_limit = ip6_dst_hoplimit(dst);
1258	}
1259
1260	/* Calculate max headroom for all the headers and adjust
1261	 * needed_headroom if necessary.
1262	 */
1263	max_headroom = LL_RESERVED_SPACE(tdev) + sizeof(struct ipv6hdr)
1264			+ dst->header_len + t->hlen;
1265	ip_tunnel_adj_headroom(dev, max_headroom);
1266
1267	err = ip6_tnl_encap(skb, t, &proto, fl6);
1268	if (err)
1269		return err;
1270
1271	if (encap_limit >= 0) {
1272		init_tel_txopt(&opt, encap_limit);
1273		ipv6_push_frag_opts(skb, &opt.ops, &proto);
1274	}
1275
1276	skb_push(skb, sizeof(struct ipv6hdr));
1277	skb_reset_network_header(skb);
1278	ipv6h = ipv6_hdr(skb);
1279	ip6_flow_hdr(ipv6h, dsfield,
1280		     ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1281	ipv6h->hop_limit = hop_limit;
1282	ipv6h->nexthdr = proto;
1283	ipv6h->saddr = fl6->saddr;
1284	ipv6h->daddr = fl6->daddr;
1285	ip6tunnel_xmit(NULL, skb, dev, 0);
1286	return 0;
1287tx_err_link_failure:
1288	DEV_STATS_INC(dev, tx_carrier_errors);
1289	dst_link_failure(skb);
1290tx_err_dst_release:
1291	dst_release(dst);
1292	return err;
1293}
1294EXPORT_SYMBOL(ip6_tnl_xmit);
1295
1296static inline int
1297ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev,
1298		u8 protocol)
1299{
1300	struct ip6_tnl *t = netdev_priv(dev);
1301	struct ipv6hdr *ipv6h;
1302	const struct iphdr  *iph;
1303	int encap_limit = -1;
1304	__u16 offset;
1305	struct flowi6 fl6;
1306	__u8 dsfield, orig_dsfield;
1307	__u32 mtu;
1308	u8 tproto;
1309	int err;
1310
1311	tproto = READ_ONCE(t->parms.proto);
1312	if (tproto != protocol && tproto != 0)
1313		return -1;
1314
1315	if (t->parms.collect_md) {
1316		struct ip_tunnel_info *tun_info;
1317		const struct ip_tunnel_key *key;
1318
1319		tun_info = skb_tunnel_info(skb);
1320		if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1321			     ip_tunnel_info_af(tun_info) != AF_INET6))
1322			return -1;
1323		key = &tun_info->key;
1324		memset(&fl6, 0, sizeof(fl6));
1325		fl6.flowi6_proto = protocol;
1326		fl6.saddr = key->u.ipv6.src;
1327		fl6.daddr = key->u.ipv6.dst;
1328		fl6.flowlabel = key->label;
1329		dsfield =  key->tos;
1330		switch (protocol) {
1331		case IPPROTO_IPIP:
1332			iph = ip_hdr(skb);
1333			orig_dsfield = ipv4_get_dsfield(iph);
1334			break;
1335		case IPPROTO_IPV6:
1336			ipv6h = ipv6_hdr(skb);
1337			orig_dsfield = ipv6_get_dsfield(ipv6h);
1338			break;
1339		default:
1340			orig_dsfield = dsfield;
1341			break;
1342		}
1343	} else {
1344		if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1345			encap_limit = t->parms.encap_limit;
1346		if (protocol == IPPROTO_IPV6) {
1347			offset = ip6_tnl_parse_tlv_enc_lim(skb,
1348						skb_network_header(skb));
1349			/* ip6_tnl_parse_tlv_enc_lim() might have
1350			 * reallocated skb->head
1351			 */
1352			if (offset > 0) {
1353				struct ipv6_tlv_tnl_enc_lim *tel;
1354
1355				tel = (void *)&skb_network_header(skb)[offset];
1356				if (tel->encap_limit == 0) {
1357					icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
1358							ICMPV6_HDR_FIELD, offset + 2);
1359					return -1;
1360				}
1361				encap_limit = tel->encap_limit - 1;
1362			}
1363		}
1364
1365		memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1366		fl6.flowi6_proto = protocol;
1367
1368		if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1369			fl6.flowi6_mark = skb->mark;
1370		else
1371			fl6.flowi6_mark = t->parms.fwmark;
1372		switch (protocol) {
1373		case IPPROTO_IPIP:
1374			iph = ip_hdr(skb);
1375			orig_dsfield = ipv4_get_dsfield(iph);
1376			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1377				dsfield = orig_dsfield;
1378			else
1379				dsfield = ip6_tclass(t->parms.flowinfo);
1380			break;
1381		case IPPROTO_IPV6:
1382			ipv6h = ipv6_hdr(skb);
1383			orig_dsfield = ipv6_get_dsfield(ipv6h);
1384			if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1385				dsfield = orig_dsfield;
1386			else
1387				dsfield = ip6_tclass(t->parms.flowinfo);
1388			if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1389				fl6.flowlabel |= ip6_flowlabel(ipv6h);
1390			break;
1391		default:
1392			orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo);
1393			break;
1394		}
1395	}
1396
1397	fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1398	dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield);
1399
1400	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1401		return -1;
1402
1403	skb_set_inner_ipproto(skb, protocol);
1404
1405	err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1406			   protocol);
1407	if (err != 0) {
1408		/* XXX: send ICMP error even if DF is not set. */
1409		if (err == -EMSGSIZE)
1410			switch (protocol) {
1411			case IPPROTO_IPIP:
1412				icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1413					      ICMP_FRAG_NEEDED, htonl(mtu));
1414				break;
1415			case IPPROTO_IPV6:
1416				icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1417				break;
1418			default:
1419				break;
1420			}
1421		return -1;
1422	}
1423
1424	return 0;
1425}
1426
1427static netdev_tx_t
1428ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1429{
1430	struct ip6_tnl *t = netdev_priv(dev);
1431	u8 ipproto;
1432	int ret;
1433
1434	if (!pskb_inet_may_pull(skb))
1435		goto tx_err;
1436
1437	switch (skb->protocol) {
1438	case htons(ETH_P_IP):
1439		ipproto = IPPROTO_IPIP;
1440		break;
1441	case htons(ETH_P_IPV6):
1442		if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb)))
1443			goto tx_err;
1444		ipproto = IPPROTO_IPV6;
1445		break;
1446	case htons(ETH_P_MPLS_UC):
1447		ipproto = IPPROTO_MPLS;
1448		break;
1449	default:
1450		goto tx_err;
1451	}
1452
1453	ret = ipxip6_tnl_xmit(skb, dev, ipproto);
1454	if (ret < 0)
1455		goto tx_err;
1456
1457	return NETDEV_TX_OK;
1458
1459tx_err:
1460	DEV_STATS_INC(dev, tx_errors);
1461	DEV_STATS_INC(dev, tx_dropped);
1462	kfree_skb(skb);
1463	return NETDEV_TX_OK;
1464}
1465
1466static void ip6_tnl_link_config(struct ip6_tnl *t)
1467{
1468	struct net_device *dev = t->dev;
1469	struct net_device *tdev = NULL;
1470	struct __ip6_tnl_parm *p = &t->parms;
1471	struct flowi6 *fl6 = &t->fl.u.ip6;
1472	int t_hlen;
1473	int mtu;
1474
1475	__dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1476	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1477
1478	/* Set up flowi template */
1479	fl6->saddr = p->laddr;
1480	fl6->daddr = p->raddr;
1481	fl6->flowi6_oif = p->link;
1482	fl6->flowlabel = 0;
1483
1484	if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1485		fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1486	if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1487		fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1488
1489	p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1490	p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1491
1492	if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1493		dev->flags |= IFF_POINTOPOINT;
1494	else
1495		dev->flags &= ~IFF_POINTOPOINT;
1496
1497	t->tun_hlen = 0;
1498	t->hlen = t->encap_hlen + t->tun_hlen;
1499	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1500
1501	if (p->flags & IP6_TNL_F_CAP_XMIT) {
1502		int strict = (ipv6_addr_type(&p->raddr) &
1503			      (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1504
1505		struct rt6_info *rt = rt6_lookup(t->net,
1506						 &p->raddr, &p->laddr,
1507						 p->link, NULL, strict);
1508		if (rt) {
1509			tdev = rt->dst.dev;
1510			ip6_rt_put(rt);
1511		}
1512
1513		if (!tdev && p->link)
1514			tdev = __dev_get_by_index(t->net, p->link);
1515
1516		if (tdev) {
1517			dev->needed_headroom = tdev->hard_header_len +
1518				tdev->needed_headroom + t_hlen;
1519			mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU);
1520
1521			mtu = mtu - t_hlen;
1522			if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1523				mtu -= 8;
1524
1525			if (mtu < IPV6_MIN_MTU)
1526				mtu = IPV6_MIN_MTU;
1527			WRITE_ONCE(dev->mtu, mtu);
1528		}
1529	}
1530}
1531
1532/**
1533 * ip6_tnl_change - update the tunnel parameters
1534 *   @t: tunnel to be changed
1535 *   @p: tunnel configuration parameters
1536 *
1537 * Description:
1538 *   ip6_tnl_change() updates the tunnel parameters
1539 **/
1540
1541static void
1542ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1543{
1544	t->parms.laddr = p->laddr;
1545	t->parms.raddr = p->raddr;
1546	t->parms.flags = p->flags;
1547	t->parms.hop_limit = p->hop_limit;
1548	t->parms.encap_limit = p->encap_limit;
1549	t->parms.flowinfo = p->flowinfo;
1550	t->parms.link = p->link;
1551	t->parms.proto = p->proto;
1552	t->parms.fwmark = p->fwmark;
1553	dst_cache_reset(&t->dst_cache);
1554	ip6_tnl_link_config(t);
1555}
1556
1557static void ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1558{
1559	struct net *net = t->net;
1560	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1561
1562	ip6_tnl_unlink(ip6n, t);
1563	synchronize_net();
1564	ip6_tnl_change(t, p);
1565	ip6_tnl_link(ip6n, t);
1566	netdev_state_change(t->dev);
1567}
1568
1569static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
1570			   bool strict)
1571{
1572	/* For the default ip6tnl0 device, allow changing only the protocol
1573	 * (the IP6_TNL_F_CAP_PER_PACKET flag is set on ip6tnl0, and all other
1574	 * parameters are 0).
1575	 */
1576	if (strict &&
1577	    (!ipv6_addr_any(&p->laddr) || !ipv6_addr_any(&p->raddr) ||
1578	     p->flags != t->parms.flags || p->hop_limit || p->encap_limit ||
1579	     p->flowinfo || p->link || p->fwmark || p->collect_md))
1580		return -EINVAL;
1581
1582	t->parms.proto = p->proto;
1583	netdev_state_change(t->dev);
1584	return 0;
1585}
1586
1587static void
1588ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1589{
1590	p->laddr = u->laddr;
1591	p->raddr = u->raddr;
1592	p->flags = u->flags;
1593	p->hop_limit = u->hop_limit;
1594	p->encap_limit = u->encap_limit;
1595	p->flowinfo = u->flowinfo;
1596	p->link = u->link;
1597	p->proto = u->proto;
1598	memcpy(p->name, u->name, sizeof(u->name));
1599}
1600
1601static void
1602ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1603{
1604	u->laddr = p->laddr;
1605	u->raddr = p->raddr;
1606	u->flags = p->flags;
1607	u->hop_limit = p->hop_limit;
1608	u->encap_limit = p->encap_limit;
1609	u->flowinfo = p->flowinfo;
1610	u->link = p->link;
1611	u->proto = p->proto;
1612	memcpy(u->name, p->name, sizeof(u->name));
1613}
1614
1615/**
1616 * ip6_tnl_siocdevprivate - configure ipv6 tunnels from userspace
1617 *   @dev: virtual device associated with tunnel
1618 *   @ifr: unused
1619 *   @data: parameters passed from userspace
1620 *   @cmd: command to be performed
1621 *
1622 * Description:
1623 *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1624 *   from userspace.
1625 *
1626 *   The possible commands are the following:
1627 *     %SIOCGETTUNNEL: get tunnel parameters for device
1628 *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1629 *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1630 *     %SIOCDELTUNNEL: delete tunnel
1631 *
1632 *   The fallback device "ip6tnl0", created during module
1633 *   initialization, can be used for creating other tunnel devices.
1634 *
1635 * Return:
1636 *   0 on success,
1637 *   %-EFAULT if unable to copy data to or from userspace,
1638 *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1639 *   %-EINVAL if passed tunnel parameters are invalid,
1640 *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1641 *   %-ENODEV if attempting to change or delete a nonexisting device
1642 **/
1643
1644static int
1645ip6_tnl_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1646		       void __user *data, int cmd)
1647{
1648	int err = 0;
1649	struct ip6_tnl_parm p;
1650	struct __ip6_tnl_parm p1;
1651	struct ip6_tnl *t = netdev_priv(dev);
1652	struct net *net = t->net;
1653	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1654
1655	memset(&p1, 0, sizeof(p1));
1656
1657	switch (cmd) {
1658	case SIOCGETTUNNEL:
1659		if (dev == ip6n->fb_tnl_dev) {
1660			if (copy_from_user(&p, data, sizeof(p))) {
1661				err = -EFAULT;
1662				break;
1663			}
1664			ip6_tnl_parm_from_user(&p1, &p);
1665			t = ip6_tnl_locate(net, &p1, 0);
1666			if (IS_ERR(t))
1667				t = netdev_priv(dev);
1668		} else {
1669			memset(&p, 0, sizeof(p));
1670		}
1671		ip6_tnl_parm_to_user(&p, &t->parms);
1672		if (copy_to_user(data, &p, sizeof(p)))
1673			err = -EFAULT;
1674		break;
1675	case SIOCADDTUNNEL:
1676	case SIOCCHGTUNNEL:
1677		err = -EPERM;
1678		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1679			break;
1680		err = -EFAULT;
1681		if (copy_from_user(&p, data, sizeof(p)))
1682			break;
1683		err = -EINVAL;
1684		if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1685		    p.proto != 0)
1686			break;
1687		ip6_tnl_parm_from_user(&p1, &p);
1688		t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1689		if (cmd == SIOCCHGTUNNEL) {
1690			if (!IS_ERR(t)) {
1691				if (t->dev != dev) {
1692					err = -EEXIST;
1693					break;
1694				}
1695			} else
1696				t = netdev_priv(dev);
1697			if (dev == ip6n->fb_tnl_dev)
1698				ip6_tnl0_update(t, &p1, false);
1699			else
1700				ip6_tnl_update(t, &p1);
1701		}
1702		if (!IS_ERR(t)) {
1703			err = 0;
1704			ip6_tnl_parm_to_user(&p, &t->parms);
1705			if (copy_to_user(data, &p, sizeof(p)))
1706				err = -EFAULT;
1707
1708		} else {
1709			err = PTR_ERR(t);
1710		}
1711		break;
1712	case SIOCDELTUNNEL:
1713		err = -EPERM;
1714		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1715			break;
1716
1717		if (dev == ip6n->fb_tnl_dev) {
1718			err = -EFAULT;
1719			if (copy_from_user(&p, data, sizeof(p)))
1720				break;
1721			err = -ENOENT;
1722			ip6_tnl_parm_from_user(&p1, &p);
1723			t = ip6_tnl_locate(net, &p1, 0);
1724			if (IS_ERR(t))
1725				break;
1726			err = -EPERM;
1727			if (t->dev == ip6n->fb_tnl_dev)
1728				break;
1729			dev = t->dev;
1730		}
1731		err = 0;
1732		unregister_netdevice(dev);
1733		break;
1734	default:
1735		err = -EINVAL;
1736	}
1737	return err;
1738}
1739
1740/**
1741 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1742 *   @dev: virtual device associated with tunnel
1743 *   @new_mtu: the new mtu
1744 *
1745 * Return:
1746 *   0 on success,
1747 *   %-EINVAL if mtu too small
1748 **/
1749
1750int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1751{
1752	struct ip6_tnl *tnl = netdev_priv(dev);
1753	int t_hlen;
1754
1755	t_hlen = tnl->hlen + sizeof(struct ipv6hdr);
1756	if (tnl->parms.proto == IPPROTO_IPV6) {
1757		if (new_mtu < IPV6_MIN_MTU)
1758			return -EINVAL;
1759	} else {
1760		if (new_mtu < ETH_MIN_MTU)
1761			return -EINVAL;
1762	}
1763	if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1764		if (new_mtu > IP6_MAX_MTU - dev->hard_header_len - t_hlen)
1765			return -EINVAL;
1766	} else {
1767		if (new_mtu > IP_MAX_MTU - dev->hard_header_len - t_hlen)
1768			return -EINVAL;
1769	}
1770	WRITE_ONCE(dev->mtu, new_mtu);
1771	return 0;
1772}
1773EXPORT_SYMBOL(ip6_tnl_change_mtu);
1774
1775int ip6_tnl_get_iflink(const struct net_device *dev)
1776{
1777	struct ip6_tnl *t = netdev_priv(dev);
1778
1779	return READ_ONCE(t->parms.link);
1780}
1781EXPORT_SYMBOL(ip6_tnl_get_iflink);
1782
1783int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1784			  unsigned int num)
1785{
1786	if (num >= MAX_IPTUN_ENCAP_OPS)
1787		return -ERANGE;
1788
1789	return !cmpxchg((const struct ip6_tnl_encap_ops **)
1790			&ip6tun_encaps[num],
1791			NULL, ops) ? 0 : -1;
1792}
1793EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1794
1795int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1796			  unsigned int num)
1797{
1798	int ret;
1799
1800	if (num >= MAX_IPTUN_ENCAP_OPS)
1801		return -ERANGE;
1802
1803	ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1804		       &ip6tun_encaps[num],
1805		       ops, NULL) == ops) ? 0 : -1;
1806
1807	synchronize_net();
1808
1809	return ret;
1810}
1811EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1812
1813int ip6_tnl_encap_setup(struct ip6_tnl *t,
1814			struct ip_tunnel_encap *ipencap)
1815{
1816	int hlen;
1817
1818	memset(&t->encap, 0, sizeof(t->encap));
1819
1820	hlen = ip6_encap_hlen(ipencap);
1821	if (hlen < 0)
1822		return hlen;
1823
1824	t->encap.type = ipencap->type;
1825	t->encap.sport = ipencap->sport;
1826	t->encap.dport = ipencap->dport;
1827	t->encap.flags = ipencap->flags;
1828
1829	t->encap_hlen = hlen;
1830	t->hlen = t->encap_hlen + t->tun_hlen;
1831
1832	return 0;
1833}
1834EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1835
1836static const struct net_device_ops ip6_tnl_netdev_ops = {
1837	.ndo_init	= ip6_tnl_dev_init,
1838	.ndo_uninit	= ip6_tnl_dev_uninit,
1839	.ndo_start_xmit = ip6_tnl_start_xmit,
1840	.ndo_siocdevprivate = ip6_tnl_siocdevprivate,
1841	.ndo_change_mtu = ip6_tnl_change_mtu,
1842	.ndo_get_stats64 = dev_get_tstats64,
1843	.ndo_get_iflink = ip6_tnl_get_iflink,
1844};
1845
1846#define IPXIPX_FEATURES (NETIF_F_SG |		\
1847			 NETIF_F_FRAGLIST |	\
1848			 NETIF_F_HIGHDMA |	\
1849			 NETIF_F_GSO_SOFTWARE |	\
1850			 NETIF_F_HW_CSUM)
1851
1852/**
1853 * ip6_tnl_dev_setup - setup virtual tunnel device
1854 *   @dev: virtual device associated with tunnel
1855 *
1856 * Description:
1857 *   Initialize function pointers and device parameters
1858 **/
1859
1860static void ip6_tnl_dev_setup(struct net_device *dev)
1861{
1862	dev->netdev_ops = &ip6_tnl_netdev_ops;
1863	dev->header_ops = &ip_tunnel_header_ops;
1864	dev->needs_free_netdev = true;
1865	dev->priv_destructor = ip6_dev_free;
1866
1867	dev->type = ARPHRD_TUNNEL6;
1868	dev->flags |= IFF_NOARP;
1869	dev->addr_len = sizeof(struct in6_addr);
1870	dev->lltx = true;
1871	dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
1872	netif_keep_dst(dev);
1873
1874	dev->features		|= IPXIPX_FEATURES;
1875	dev->hw_features	|= IPXIPX_FEATURES;
1876
1877	/* This perm addr will be used as interface identifier by IPv6 */
1878	dev->addr_assign_type = NET_ADDR_RANDOM;
1879	eth_random_addr(dev->perm_addr);
1880}
1881
1882
1883/**
1884 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1885 *   @dev: virtual device associated with tunnel
1886 **/
1887
1888static inline int
1889ip6_tnl_dev_init_gen(struct net_device *dev)
1890{
1891	struct ip6_tnl *t = netdev_priv(dev);
1892	int ret;
1893	int t_hlen;
1894
1895	t->dev = dev;
1896
1897	ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1898	if (ret)
1899		return ret;
1900
1901	ret = gro_cells_init(&t->gro_cells, dev);
1902	if (ret)
1903		goto destroy_dst;
1904
1905	t->tun_hlen = 0;
1906	t->hlen = t->encap_hlen + t->tun_hlen;
1907	t_hlen = t->hlen + sizeof(struct ipv6hdr);
1908
1909	dev->type = ARPHRD_TUNNEL6;
1910	dev->mtu = ETH_DATA_LEN - t_hlen;
1911	if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1912		dev->mtu -= 8;
1913	dev->min_mtu = ETH_MIN_MTU;
1914	dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len - t_hlen;
1915
1916	netdev_hold(dev, &t->dev_tracker, GFP_KERNEL);
1917	netdev_lockdep_set_classes(dev);
1918	return 0;
1919
1920destroy_dst:
1921	dst_cache_destroy(&t->dst_cache);
1922
1923	return ret;
1924}
1925
1926/**
1927 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1928 *   @dev: virtual device associated with tunnel
1929 **/
1930
1931static int ip6_tnl_dev_init(struct net_device *dev)
1932{
1933	struct ip6_tnl *t = netdev_priv(dev);
1934	int err = ip6_tnl_dev_init_gen(dev);
1935
1936	if (err)
1937		return err;
1938	ip6_tnl_link_config(t);
1939	if (t->parms.collect_md)
1940		netif_keep_dst(dev);
1941	return 0;
1942}
1943
1944/**
1945 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1946 *   @dev: fallback device
1947 *
1948 * Return: 0
1949 **/
1950
1951static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1952{
1953	struct ip6_tnl *t = netdev_priv(dev);
1954	struct net *net = dev_net(dev);
1955	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1956
1957	t->net = net;
1958	t->parms.proto = IPPROTO_IPV6;
1959
1960	rcu_assign_pointer(ip6n->tnls_wc[0], t);
1961	return 0;
1962}
1963
1964static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1965			    struct netlink_ext_ack *extack)
1966{
1967	u8 proto;
1968
1969	if (!data || !data[IFLA_IPTUN_PROTO])
1970		return 0;
1971
1972	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1973	if (proto != IPPROTO_IPV6 &&
1974	    proto != IPPROTO_IPIP &&
1975	    proto != 0)
1976		return -EINVAL;
1977
1978	return 0;
1979}
1980
1981static void ip6_tnl_netlink_parms(struct nlattr *data[],
1982				  struct __ip6_tnl_parm *parms)
1983{
1984	memset(parms, 0, sizeof(*parms));
1985
1986	if (!data)
1987		return;
1988
1989	if (data[IFLA_IPTUN_LINK])
1990		parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1991
1992	if (data[IFLA_IPTUN_LOCAL])
1993		parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1994
1995	if (data[IFLA_IPTUN_REMOTE])
1996		parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1997
1998	if (data[IFLA_IPTUN_TTL])
1999		parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
2000
2001	if (data[IFLA_IPTUN_ENCAP_LIMIT])
2002		parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
2003
2004	if (data[IFLA_IPTUN_FLOWINFO])
2005		parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
2006
2007	if (data[IFLA_IPTUN_FLAGS])
2008		parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
2009
2010	if (data[IFLA_IPTUN_PROTO])
2011		parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
2012
2013	if (data[IFLA_IPTUN_COLLECT_METADATA])
2014		parms->collect_md = true;
2015
2016	if (data[IFLA_IPTUN_FWMARK])
2017		parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
2018}
2019
2020static int ip6_tnl_newlink(struct net_device *dev,
2021			   struct rtnl_newlink_params *params,
2022			   struct netlink_ext_ack *extack)
2023{
2024	struct nlattr **data = params->data;
2025	struct nlattr **tb = params->tb;
2026	struct ip_tunnel_encap ipencap;
2027	struct ip6_tnl_net *ip6n;
2028	struct ip6_tnl *nt, *t;
2029	struct net *net;
2030	int err;
2031
2032	net = params->link_net ? : dev_net(dev);
2033	ip6n = net_generic(net, ip6_tnl_net_id);
2034	nt = netdev_priv(dev);
2035	nt->net = net;
2036
2037	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
2038		err = ip6_tnl_encap_setup(nt, &ipencap);
2039		if (err < 0)
2040			return err;
2041	}
2042
2043	ip6_tnl_netlink_parms(data, &nt->parms);
2044
2045	if (nt->parms.collect_md) {
2046		if (rtnl_dereference(ip6n->collect_md_tun))
2047			return -EEXIST;
2048	} else {
2049		t = ip6_tnl_locate(net, &nt->parms, 0);
2050		if (!IS_ERR(t))
2051			return -EEXIST;
2052	}
2053
2054	err = ip6_tnl_create2(dev);
2055	if (!err && tb[IFLA_MTU])
2056		ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2057
2058	return err;
2059}
2060
2061static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2062			      struct nlattr *data[],
2063			      struct netlink_ext_ack *extack)
2064{
2065	struct ip6_tnl *t = netdev_priv(dev);
2066	struct __ip6_tnl_parm p;
2067	struct net *net = t->net;
2068	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2069	struct ip_tunnel_encap ipencap;
2070
2071	if (dev == ip6n->fb_tnl_dev) {
2072		if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
2073			/* iproute2 always sets TUNNEL_ENCAP_FLAG_CSUM6, so
2074			 * let's ignore this flag.
2075			 */
2076			ipencap.flags &= ~TUNNEL_ENCAP_FLAG_CSUM6;
2077			if (memchr_inv(&ipencap, 0, sizeof(ipencap))) {
2078				NL_SET_ERR_MSG(extack,
2079					       "Only protocol can be changed for fallback tunnel, not encap params");
2080				return -EINVAL;
2081			}
2082		}
2083
2084		ip6_tnl_netlink_parms(data, &p);
2085		if (ip6_tnl0_update(t, &p, true) < 0) {
2086			NL_SET_ERR_MSG(extack,
2087				       "Only protocol can be changed for fallback tunnel");
2088			return -EINVAL;
2089		}
2090
2091		return 0;
2092	}
2093
2094	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
2095		int err = ip6_tnl_encap_setup(t, &ipencap);
2096
2097		if (err < 0)
2098			return err;
2099	}
2100	ip6_tnl_netlink_parms(data, &p);
2101	if (p.collect_md)
2102		return -EINVAL;
2103
2104	t = ip6_tnl_locate(net, &p, 0);
2105	if (!IS_ERR(t)) {
2106		if (t->dev != dev)
2107			return -EEXIST;
2108	} else
2109		t = netdev_priv(dev);
2110
2111	ip6_tnl_update(t, &p);
2112	return 0;
2113}
2114
2115static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2116{
2117	struct net *net = dev_net(dev);
2118	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2119
2120	if (dev != ip6n->fb_tnl_dev)
2121		unregister_netdevice_queue(dev, head);
2122}
2123
2124static size_t ip6_tnl_get_size(const struct net_device *dev)
2125{
2126	return
2127		/* IFLA_IPTUN_LINK */
2128		nla_total_size(4) +
2129		/* IFLA_IPTUN_LOCAL */
2130		nla_total_size(sizeof(struct in6_addr)) +
2131		/* IFLA_IPTUN_REMOTE */
2132		nla_total_size(sizeof(struct in6_addr)) +
2133		/* IFLA_IPTUN_TTL */
2134		nla_total_size(1) +
2135		/* IFLA_IPTUN_ENCAP_LIMIT */
2136		nla_total_size(1) +
2137		/* IFLA_IPTUN_FLOWINFO */
2138		nla_total_size(4) +
2139		/* IFLA_IPTUN_FLAGS */
2140		nla_total_size(4) +
2141		/* IFLA_IPTUN_PROTO */
2142		nla_total_size(1) +
2143		/* IFLA_IPTUN_ENCAP_TYPE */
2144		nla_total_size(2) +
2145		/* IFLA_IPTUN_ENCAP_FLAGS */
2146		nla_total_size(2) +
2147		/* IFLA_IPTUN_ENCAP_SPORT */
2148		nla_total_size(2) +
2149		/* IFLA_IPTUN_ENCAP_DPORT */
2150		nla_total_size(2) +
2151		/* IFLA_IPTUN_COLLECT_METADATA */
2152		nla_total_size(0) +
2153		/* IFLA_IPTUN_FWMARK */
2154		nla_total_size(4) +
2155		0;
2156}
2157
2158static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2159{
2160	struct ip6_tnl *tunnel = netdev_priv(dev);
2161	struct __ip6_tnl_parm *parm = &tunnel->parms;
2162
2163	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2164	    nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2165	    nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2166	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2167	    nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2168	    nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2169	    nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2170	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2171	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2172		goto nla_put_failure;
2173
2174	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2175	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2176	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2177	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2178		goto nla_put_failure;
2179
2180	if (parm->collect_md)
2181		if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2182			goto nla_put_failure;
2183
2184	return 0;
2185
2186nla_put_failure:
2187	return -EMSGSIZE;
2188}
2189
2190struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2191{
2192	struct ip6_tnl *tunnel = netdev_priv(dev);
2193
2194	return READ_ONCE(tunnel->net);
2195}
2196EXPORT_SYMBOL(ip6_tnl_get_link_net);
2197
2198static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2199	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
2200	[IFLA_IPTUN_LOCAL]		= { .len = sizeof(struct in6_addr) },
2201	[IFLA_IPTUN_REMOTE]		= { .len = sizeof(struct in6_addr) },
2202	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
2203	[IFLA_IPTUN_ENCAP_LIMIT]	= { .type = NLA_U8 },
2204	[IFLA_IPTUN_FLOWINFO]		= { .type = NLA_U32 },
2205	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U32 },
2206	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
2207	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
2208	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
2209	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
2210	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
2211	[IFLA_IPTUN_COLLECT_METADATA]	= { .type = NLA_FLAG },
2212	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
2213};
2214
2215static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2216	.kind		= "ip6tnl",
2217	.maxtype	= IFLA_IPTUN_MAX,
2218	.policy		= ip6_tnl_policy,
2219	.priv_size	= sizeof(struct ip6_tnl),
2220	.setup		= ip6_tnl_dev_setup,
2221	.validate	= ip6_tnl_validate,
2222	.newlink	= ip6_tnl_newlink,
2223	.changelink	= ip6_tnl_changelink,
2224	.dellink	= ip6_tnl_dellink,
2225	.get_size	= ip6_tnl_get_size,
2226	.fill_info	= ip6_tnl_fill_info,
2227	.get_link_net	= ip6_tnl_get_link_net,
2228};
2229
2230static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2231	.handler	= ip4ip6_rcv,
2232	.err_handler	= ip4ip6_err,
2233	.priority	=	1,
2234};
2235
2236static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2237	.handler	= ip6ip6_rcv,
2238	.err_handler	= ip6ip6_err,
2239	.priority	=	1,
2240};
2241
2242static struct xfrm6_tunnel mplsip6_handler __read_mostly = {
2243	.handler	= mplsip6_rcv,
2244	.err_handler	= mplsip6_err,
2245	.priority	=	1,
2246};
2247
2248static void __net_exit ip6_tnl_exit_rtnl_net(struct net *net, struct list_head *list)
2249{
2250	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2251	struct net_device *dev, *aux;
2252	int h;
2253	struct ip6_tnl *t;
2254
2255	for_each_netdev_safe(net, dev, aux)
2256		if (dev->rtnl_link_ops == &ip6_link_ops)
2257			unregister_netdevice_queue(dev, list);
2258
2259	for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2260		t = rtnl_net_dereference(net, ip6n->tnls_r_l[h]);
2261		while (t) {
2262			/* If dev is in the same netns, it has already
2263			 * been added to the list by the previous loop.
2264			 */
2265			if (!net_eq(dev_net(t->dev), net))
2266				unregister_netdevice_queue(t->dev, list);
2267
2268			t = rtnl_net_dereference(net, t->next);
2269		}
2270	}
2271
2272	t = rtnl_net_dereference(net, ip6n->tnls_wc[0]);
2273	while (t) {
2274		/* If dev is in the same netns, it has already
2275		 * been added to the list by the previous loop.
2276		 */
2277		if (!net_eq(dev_net(t->dev), net))
2278			unregister_netdevice_queue(t->dev, list);
2279
2280		t = rtnl_net_dereference(net, t->next);
2281	}
2282}
2283
2284static int __net_init ip6_tnl_init_net(struct net *net)
2285{
2286	struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2287	struct ip6_tnl *t = NULL;
2288	int err;
2289
2290	ip6n->tnls[0] = ip6n->tnls_wc;
2291	ip6n->tnls[1] = ip6n->tnls_r_l;
2292
2293	if (!net_has_fallback_tunnels(net))
2294		return 0;
2295	err = -ENOMEM;
2296	ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2297					NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2298
2299	if (!ip6n->fb_tnl_dev)
2300		goto err_alloc_dev;
2301	dev_net_set(ip6n->fb_tnl_dev, net);
2302	ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2303	/* FB netdevice is special: we have one, and only one per netns.
2304	 * Allowing to move it to another netns is clearly unsafe.
2305	 */
2306	ip6n->fb_tnl_dev->netns_immutable = true;
2307
2308	err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2309	if (err < 0)
2310		goto err_register;
2311
2312	err = register_netdev(ip6n->fb_tnl_dev);
2313	if (err < 0)
2314		goto err_register;
2315
2316	t = netdev_priv(ip6n->fb_tnl_dev);
2317
2318	strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2319	return 0;
2320
2321err_register:
2322	free_netdev(ip6n->fb_tnl_dev);
2323err_alloc_dev:
2324	return err;
2325}
2326
2327static struct pernet_operations ip6_tnl_net_ops = {
2328	.init = ip6_tnl_init_net,
2329	.exit_rtnl = ip6_tnl_exit_rtnl_net,
2330	.id   = &ip6_tnl_net_id,
2331	.size = sizeof(struct ip6_tnl_net),
2332};
2333
2334/**
2335 * ip6_tunnel_init - register protocol and reserve needed resources
2336 *
2337 * Return: 0 on success
2338 **/
2339
2340static int __init ip6_tunnel_init(void)
2341{
2342	int  err;
2343
2344	if (!ipv6_mod_enabled())
2345		return -EOPNOTSUPP;
2346
2347	err = register_pernet_device(&ip6_tnl_net_ops);
2348	if (err < 0)
2349		goto out_pernet;
2350
2351	err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2352	if (err < 0) {
2353		pr_err("%s: can't register ip4ip6\n", __func__);
2354		goto out_ip4ip6;
2355	}
2356
2357	err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2358	if (err < 0) {
2359		pr_err("%s: can't register ip6ip6\n", __func__);
2360		goto out_ip6ip6;
2361	}
2362
2363	if (ip6_tnl_mpls_supported()) {
2364		err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS);
2365		if (err < 0) {
2366			pr_err("%s: can't register mplsip6\n", __func__);
2367			goto out_mplsip6;
2368		}
2369	}
2370
2371	err = rtnl_link_register(&ip6_link_ops);
2372	if (err < 0)
2373		goto rtnl_link_failed;
2374
2375	return 0;
2376
2377rtnl_link_failed:
2378	if (ip6_tnl_mpls_supported())
2379		xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS);
2380out_mplsip6:
2381	xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2382out_ip6ip6:
2383	xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2384out_ip4ip6:
2385	unregister_pernet_device(&ip6_tnl_net_ops);
2386out_pernet:
2387	return err;
2388}
2389
2390/**
2391 * ip6_tunnel_cleanup - free resources and unregister protocol
2392 **/
2393
2394static void __exit ip6_tunnel_cleanup(void)
2395{
2396	rtnl_link_unregister(&ip6_link_ops);
2397	if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2398		pr_info("%s: can't deregister ip4ip6\n", __func__);
2399
2400	if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2401		pr_info("%s: can't deregister ip6ip6\n", __func__);
2402
2403	if (ip6_tnl_mpls_supported() &&
2404	    xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS))
2405		pr_info("%s: can't deregister mplsip6\n", __func__);
2406	unregister_pernet_device(&ip6_tnl_net_ops);
2407}
2408
2409module_init(ip6_tunnel_init);
2410module_exit(ip6_tunnel_cleanup);