Linux Audio

Check our new training course

Loading...
v6.8
   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
  86static inline struct sit_net *dev_to_sit_net(struct net_device *dev)
  87{
  88	struct ip_tunnel *t = netdev_priv(dev);
  89
  90	return net_generic(t->net, sit_net_id);
  91}
  92
  93/*
  94 * Must be invoked with rcu_read_lock
  95 */
  96static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
  97					     struct net_device *dev,
  98					     __be32 remote, __be32 local,
  99					     int sifindex)
 100{
 101	unsigned int h0 = HASH(remote);
 102	unsigned int h1 = HASH(local);
 103	struct ip_tunnel *t;
 104	struct sit_net *sitn = net_generic(net, sit_net_id);
 105	int ifindex = dev ? dev->ifindex : 0;
 106
 107	for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 108		if (local == t->parms.iph.saddr &&
 109		    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_r[h0]) {
 116		if (remote == t->parms.iph.daddr &&
 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	for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 123		if (local == t->parms.iph.saddr &&
 124		    (!dev || !t->parms.link || ifindex == t->parms.link ||
 125		     sifindex == t->parms.link) &&
 126		    (t->dev->flags & IFF_UP))
 127			return t;
 128	}
 129	t = rcu_dereference(sitn->tunnels_wc[0]);
 130	if (t && (t->dev->flags & IFF_UP))
 131		return t;
 132	return NULL;
 133}
 134
 135static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
 136		struct ip_tunnel_parm *parms)
 137{
 138	__be32 remote = parms->iph.daddr;
 139	__be32 local = parms->iph.saddr;
 140	unsigned int h = 0;
 141	int prio = 0;
 142
 143	if (remote) {
 144		prio |= 2;
 145		h ^= HASH(remote);
 146	}
 147	if (local) {
 148		prio |= 1;
 149		h ^= HASH(local);
 150	}
 151	return &sitn->tunnels[prio][h];
 152}
 153
 154static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 155		struct ip_tunnel *t)
 156{
 157	return __ipip6_bucket(sitn, &t->parms);
 158}
 159
 160static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 161{
 162	struct ip_tunnel __rcu **tp;
 163	struct ip_tunnel *iter;
 164
 165	for (tp = ipip6_bucket(sitn, t);
 166	     (iter = rtnl_dereference(*tp)) != NULL;
 167	     tp = &iter->next) {
 168		if (t == iter) {
 169			rcu_assign_pointer(*tp, t->next);
 170			break;
 171		}
 172	}
 173}
 174
 175static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 176{
 177	struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 178
 179	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 180	rcu_assign_pointer(*tp, t);
 181}
 182
 183static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 184{
 185#ifdef CONFIG_IPV6_SIT_6RD
 186	struct ip_tunnel *t = netdev_priv(dev);
 187
 188	if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
 189		ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 190		t->ip6rd.relay_prefix = 0;
 191		t->ip6rd.prefixlen = 16;
 192		t->ip6rd.relay_prefixlen = 0;
 193	} else {
 194		struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 195		memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 196	}
 197#endif
 198}
 199
 200static int ipip6_tunnel_create(struct net_device *dev)
 201{
 202	struct ip_tunnel *t = netdev_priv(dev);
 203	struct net *net = dev_net(dev);
 204	struct sit_net *sitn = net_generic(net, sit_net_id);
 205	int err;
 206
 207	__dev_addr_set(dev, &t->parms.iph.saddr, 4);
 208	memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
 209
 210	if ((__force u16)t->parms.i_flags & SIT_ISATAP)
 211		dev->priv_flags |= IFF_ISATAP;
 212
 213	dev->rtnl_link_ops = &sit_link_ops;
 214
 215	err = register_netdevice(dev);
 216	if (err < 0)
 217		goto out;
 218
 219	ipip6_tunnel_clone_6rd(dev, sitn);
 220
 221	ipip6_tunnel_link(sitn, t);
 222	return 0;
 223
 224out:
 225	return err;
 226}
 227
 228static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 229		struct ip_tunnel_parm *parms, int create)
 
 230{
 231	__be32 remote = parms->iph.daddr;
 232	__be32 local = parms->iph.saddr;
 233	struct ip_tunnel *t, *nt;
 234	struct ip_tunnel __rcu **tp;
 235	struct net_device *dev;
 236	char name[IFNAMSIZ];
 237	struct sit_net *sitn = net_generic(net, sit_net_id);
 238
 239	for (tp = __ipip6_bucket(sitn, parms);
 240	    (t = rtnl_dereference(*tp)) != NULL;
 241	     tp = &t->next) {
 242		if (local == t->parms.iph.saddr &&
 243		    remote == t->parms.iph.daddr &&
 244		    parms->link == t->parms.link) {
 245			if (create)
 246				return NULL;
 247			else
 248				return t;
 249		}
 250	}
 251	if (!create)
 252		goto failed;
 253
 254	if (parms->name[0]) {
 255		if (!dev_valid_name(parms->name))
 256			goto failed;
 257		strscpy(name, parms->name, IFNAMSIZ);
 258	} else {
 259		strcpy(name, "sit%d");
 260	}
 261	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 262			   ipip6_tunnel_setup);
 263	if (!dev)
 264		return NULL;
 265
 266	dev_net_set(dev, net);
 267
 268	nt = netdev_priv(dev);
 269
 270	nt->parms = *parms;
 271	if (ipip6_tunnel_create(dev) < 0)
 272		goto failed_free;
 273
 274	if (!parms->name[0])
 275		strcpy(parms->name, dev->name);
 276
 277	return nt;
 278
 279failed_free:
 280	free_netdev(dev);
 281failed:
 282	return NULL;
 283}
 284
 285#define for_each_prl_rcu(start)			\
 286	for (prl = rcu_dereference(start);	\
 287	     prl;				\
 288	     prl = rcu_dereference(prl->next))
 289
 290static struct ip_tunnel_prl_entry *
 291__ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 292{
 293	struct ip_tunnel_prl_entry *prl;
 294
 295	for_each_prl_rcu(t->prl)
 296		if (prl->addr == addr)
 297			break;
 298	return prl;
 299
 300}
 301
 302static int ipip6_tunnel_get_prl(struct net_device *dev, struct ip_tunnel_prl __user *a)
 303{
 304	struct ip_tunnel *t = netdev_priv(dev);
 305	struct ip_tunnel_prl kprl, *kp;
 306	struct ip_tunnel_prl_entry *prl;
 307	unsigned int cmax, c = 0, ca, len;
 308	int ret = 0;
 309
 310	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
 311		return -EINVAL;
 312
 313	if (copy_from_user(&kprl, a, sizeof(kprl)))
 314		return -EFAULT;
 315	cmax = kprl.datalen / sizeof(kprl);
 316	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 317		cmax = 1;
 318
 319	/* For simple GET or for root users,
 320	 * we try harder to allocate.
 321	 */
 322	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 323		kcalloc(cmax, sizeof(*kp), GFP_KERNEL_ACCOUNT | __GFP_NOWARN) :
 324		NULL;
 325
 326	ca = min(t->prl_count, cmax);
 327
 328	if (!kp) {
 329		/* We don't try hard to allocate much memory for
 330		 * non-root users.
 331		 * For root users, retry allocating enough memory for
 332		 * the answer.
 333		 */
 334		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC | __GFP_ACCOUNT |
 335					      __GFP_NOWARN);
 336		if (!kp) {
 337			ret = -ENOMEM;
 338			goto out;
 339		}
 340	}
 341
 342	rcu_read_lock();
 343	for_each_prl_rcu(t->prl) {
 344		if (c >= cmax)
 345			break;
 346		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 347			continue;
 348		kp[c].addr = prl->addr;
 349		kp[c].flags = prl->flags;
 350		c++;
 351		if (kprl.addr != htonl(INADDR_ANY))
 352			break;
 353	}
 354
 355	rcu_read_unlock();
 356
 357	len = sizeof(*kp) * c;
 358	ret = 0;
 359	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 360		ret = -EFAULT;
 361
 362	kfree(kp);
 363out:
 364	return ret;
 365}
 366
 367static int
 368ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 369{
 370	struct ip_tunnel_prl_entry *p;
 371	int err = 0;
 372
 373	if (a->addr == htonl(INADDR_ANY))
 374		return -EINVAL;
 375
 376	ASSERT_RTNL();
 377
 378	for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 379		if (p->addr == a->addr) {
 380			if (chg) {
 381				p->flags = a->flags;
 382				goto out;
 383			}
 384			err = -EEXIST;
 385			goto out;
 386		}
 387	}
 388
 389	if (chg) {
 390		err = -ENXIO;
 391		goto out;
 392	}
 393
 394	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 395	if (!p) {
 396		err = -ENOBUFS;
 397		goto out;
 398	}
 399
 400	p->next = t->prl;
 401	p->addr = a->addr;
 402	p->flags = a->flags;
 403	t->prl_count++;
 404	rcu_assign_pointer(t->prl, p);
 405out:
 406	return err;
 407}
 408
 409static void prl_list_destroy_rcu(struct rcu_head *head)
 410{
 411	struct ip_tunnel_prl_entry *p, *n;
 412
 413	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 414	do {
 415		n = rcu_dereference_protected(p->next, 1);
 416		kfree(p);
 417		p = n;
 418	} while (p);
 419}
 420
 421static int
 422ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 423{
 424	struct ip_tunnel_prl_entry *x;
 425	struct ip_tunnel_prl_entry __rcu **p;
 426	int err = 0;
 427
 428	ASSERT_RTNL();
 429
 430	if (a && a->addr != htonl(INADDR_ANY)) {
 431		for (p = &t->prl;
 432		     (x = rtnl_dereference(*p)) != NULL;
 433		     p = &x->next) {
 434			if (x->addr == a->addr) {
 435				*p = x->next;
 436				kfree_rcu(x, rcu_head);
 437				t->prl_count--;
 438				goto out;
 439			}
 440		}
 441		err = -ENXIO;
 442	} else {
 443		x = rtnl_dereference(t->prl);
 444		if (x) {
 445			t->prl_count = 0;
 446			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 447			t->prl = NULL;
 448		}
 449	}
 450out:
 451	return err;
 452}
 453
 454static int ipip6_tunnel_prl_ctl(struct net_device *dev,
 455				struct ip_tunnel_prl __user *data, int cmd)
 456{
 457	struct ip_tunnel *t = netdev_priv(dev);
 458	struct ip_tunnel_prl prl;
 459	int err;
 460
 461	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
 462		return -EPERM;
 463	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
 464		return -EINVAL;
 465
 466	if (copy_from_user(&prl, data, sizeof(prl)))
 467		return -EFAULT;
 468
 469	switch (cmd) {
 470	case SIOCDELPRL:
 471		err = ipip6_tunnel_del_prl(t, &prl);
 472		break;
 473	case SIOCADDPRL:
 474	case SIOCCHGPRL:
 475		err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 476		break;
 477	}
 478	dst_cache_reset(&t->dst_cache);
 479	netdev_state_change(dev);
 480	return err;
 481}
 482
 483static int
 484isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 485{
 486	struct ip_tunnel_prl_entry *p;
 487	int ok = 1;
 488
 489	rcu_read_lock();
 490	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 491	if (p) {
 492		if (p->flags & PRL_DEFAULT)
 493			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 494		else
 495			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 496	} else {
 497		const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 498
 499		if (ipv6_addr_is_isatap(addr6) &&
 500		    (addr6->s6_addr32[3] == iph->saddr) &&
 501		    ipv6_chk_prefix(addr6, t->dev))
 502			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 503		else
 504			ok = 0;
 505	}
 506	rcu_read_unlock();
 507	return ok;
 508}
 509
 510static void ipip6_tunnel_uninit(struct net_device *dev)
 511{
 512	struct ip_tunnel *tunnel = netdev_priv(dev);
 513	struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
 514
 515	if (dev == sitn->fb_tunnel_dev) {
 516		RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 517	} else {
 518		ipip6_tunnel_unlink(sitn, tunnel);
 519		ipip6_tunnel_del_prl(tunnel, NULL);
 520	}
 521	dst_cache_reset(&tunnel->dst_cache);
 522	netdev_put(dev, &tunnel->dev_tracker);
 523}
 524
 525static int ipip6_err(struct sk_buff *skb, u32 info)
 526{
 527	const struct iphdr *iph = (const struct iphdr *)skb->data;
 528	const int type = icmp_hdr(skb)->type;
 529	const int code = icmp_hdr(skb)->code;
 530	unsigned int data_len = 0;
 531	struct ip_tunnel *t;
 532	int sifindex;
 533	int err;
 534
 535	switch (type) {
 536	default:
 537	case ICMP_PARAMETERPROB:
 538		return 0;
 539
 540	case ICMP_DEST_UNREACH:
 541		switch (code) {
 542		case ICMP_SR_FAILED:
 543			/* Impossible event. */
 544			return 0;
 545		default:
 546			/* All others are translated to HOST_UNREACH.
 547			   rfc2003 contains "deep thoughts" about NET_UNREACH,
 548			   I believe they are just ether pollution. --ANK
 549			 */
 550			break;
 551		}
 552		break;
 553	case ICMP_TIME_EXCEEDED:
 554		if (code != ICMP_EXC_TTL)
 555			return 0;
 556		data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
 557		break;
 558	case ICMP_REDIRECT:
 559		break;
 560	}
 561
 562	err = -ENOENT;
 563
 564	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 565	t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 566				iph->daddr, iph->saddr, sifindex);
 567	if (!t)
 568		goto out;
 569
 570	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 571		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 572				 t->parms.link, iph->protocol);
 573		err = 0;
 574		goto out;
 575	}
 576	if (type == ICMP_REDIRECT) {
 577		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
 578			      iph->protocol);
 579		err = 0;
 580		goto out;
 581	}
 582
 583	err = 0;
 584	if (__in6_dev_get(skb->dev) &&
 585	    !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
 586		goto out;
 587
 588	if (t->parms.iph.daddr == 0)
 589		goto out;
 590
 591	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 592		goto out;
 593
 594	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 595		t->err_count++;
 596	else
 597		t->err_count = 1;
 598	t->err_time = jiffies;
 599out:
 600	return err;
 601}
 602
 603static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 604				  const struct in6_addr *v6addr)
 605{
 606	__be32 v4embed = 0;
 607	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 608		return true;
 609	return false;
 610}
 611
 612/* Checks if an address matches an address on the tunnel interface.
 613 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 614 * Long story:
 615 * This function is called after we considered the packet as spoofed
 616 * in is_spoofed_6rd.
 617 * We may have a router that is doing NAT for proto 41 packets
 618 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 619 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 620 * function will return true, dropping the packet.
 621 * But, we can still check if is spoofed against the IP
 622 * addresses associated with the interface.
 623 */
 624static bool only_dnatted(const struct ip_tunnel *tunnel,
 625	const struct in6_addr *v6dst)
 626{
 627	int prefix_len;
 628
 629#ifdef CONFIG_IPV6_SIT_6RD
 630	prefix_len = tunnel->ip6rd.prefixlen + 32
 631		- tunnel->ip6rd.relay_prefixlen;
 632#else
 633	prefix_len = 48;
 634#endif
 635	return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 636}
 637
 638/* Returns true if a packet is spoofed */
 639static bool packet_is_spoofed(struct sk_buff *skb,
 640			      const struct iphdr *iph,
 641			      struct ip_tunnel *tunnel)
 642{
 643	const struct ipv6hdr *ipv6h;
 644
 645	if (tunnel->dev->priv_flags & IFF_ISATAP) {
 646		if (!isatap_chksrc(skb, iph, tunnel))
 647			return true;
 648
 649		return false;
 650	}
 651
 652	if (tunnel->dev->flags & IFF_POINTOPOINT)
 653		return false;
 654
 655	ipv6h = ipv6_hdr(skb);
 656
 657	if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 658		net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 659				     &iph->saddr, &ipv6h->saddr,
 660				     &iph->daddr, &ipv6h->daddr);
 661		return true;
 662	}
 663
 664	if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 665		return false;
 666
 667	if (only_dnatted(tunnel, &ipv6h->daddr))
 668		return false;
 669
 670	net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 671			     &iph->saddr, &ipv6h->saddr,
 672			     &iph->daddr, &ipv6h->daddr);
 673	return true;
 674}
 675
 676static int ipip6_rcv(struct sk_buff *skb)
 677{
 678	const struct iphdr *iph = ip_hdr(skb);
 679	struct ip_tunnel *tunnel;
 680	int sifindex;
 681	int err;
 682
 683	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 684	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 685				     iph->saddr, iph->daddr, sifindex);
 686	if (tunnel) {
 687		if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 688		    tunnel->parms.iph.protocol != 0)
 689			goto out;
 690
 691		skb->mac_header = skb->network_header;
 692		skb_reset_network_header(skb);
 693		IPCB(skb)->flags = 0;
 694		skb->dev = tunnel->dev;
 695
 696		if (packet_is_spoofed(skb, iph, tunnel)) {
 697			DEV_STATS_INC(tunnel->dev, rx_errors);
 698			goto out;
 699		}
 700
 701		if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
 702		    !net_eq(tunnel->net, dev_net(tunnel->dev))))
 703			goto out;
 704
 705		/* skb can be uncloned in iptunnel_pull_header, so
 706		 * old iph is no longer valid
 707		 */
 708		iph = (const struct iphdr *)skb_mac_header(skb);
 709		skb_reset_mac_header(skb);
 710
 711		err = IP_ECN_decapsulate(iph, skb);
 712		if (unlikely(err)) {
 713			if (log_ecn_error)
 714				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 715						     &iph->saddr, iph->tos);
 716			if (err > 1) {
 717				DEV_STATS_INC(tunnel->dev, rx_frame_errors);
 718				DEV_STATS_INC(tunnel->dev, rx_errors);
 719				goto out;
 720			}
 721		}
 722
 723		dev_sw_netstats_rx_add(tunnel->dev, skb->len);
 724
 725		netif_rx(skb);
 726
 727		return 0;
 728	}
 729
 730	/* no tunnel matched,  let upstream know, ipsec may handle it */
 731	return 1;
 732out:
 733	kfree_skb(skb);
 734	return 0;
 735}
 736
 737static const struct tnl_ptk_info ipip_tpi = {
 738	/* no tunnel info required for ipip. */
 739	.proto = htons(ETH_P_IP),
 740};
 741
 742#if IS_ENABLED(CONFIG_MPLS)
 743static const struct tnl_ptk_info mplsip_tpi = {
 744	/* no tunnel info required for mplsip. */
 745	.proto = htons(ETH_P_MPLS_UC),
 746};
 747#endif
 748
 749static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
 750{
 751	const struct iphdr *iph;
 752	struct ip_tunnel *tunnel;
 753	int sifindex;
 754
 755	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 756
 757	iph = ip_hdr(skb);
 758	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 759				     iph->saddr, iph->daddr, sifindex);
 760	if (tunnel) {
 761		const struct tnl_ptk_info *tpi;
 762
 763		if (tunnel->parms.iph.protocol != ipproto &&
 764		    tunnel->parms.iph.protocol != 0)
 765			goto drop;
 766
 767		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 768			goto drop;
 769#if IS_ENABLED(CONFIG_MPLS)
 770		if (ipproto == IPPROTO_MPLS)
 771			tpi = &mplsip_tpi;
 772		else
 773#endif
 774			tpi = &ipip_tpi;
 775		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 776			goto drop;
 777		skb_reset_mac_header(skb);
 778
 779		return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 780	}
 781
 782	return 1;
 783
 784drop:
 785	kfree_skb(skb);
 786	return 0;
 787}
 788
 789static int ipip_rcv(struct sk_buff *skb)
 790{
 791	return sit_tunnel_rcv(skb, IPPROTO_IPIP);
 792}
 793
 794#if IS_ENABLED(CONFIG_MPLS)
 795static int mplsip_rcv(struct sk_buff *skb)
 796{
 797	return sit_tunnel_rcv(skb, IPPROTO_MPLS);
 798}
 799#endif
 800
 801/*
 802 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 803 * stores the embedded IPv4 address in v4dst and returns true.
 804 */
 805static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 806		      __be32 *v4dst)
 807{
 808#ifdef CONFIG_IPV6_SIT_6RD
 809	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 810			      tunnel->ip6rd.prefixlen)) {
 811		unsigned int pbw0, pbi0;
 812		int pbi1;
 813		u32 d;
 814
 815		pbw0 = tunnel->ip6rd.prefixlen >> 5;
 816		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 817
 818		d = tunnel->ip6rd.relay_prefixlen < 32 ?
 819			(ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 820		    tunnel->ip6rd.relay_prefixlen : 0;
 821
 822		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 823		if (pbi1 > 0)
 824			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 825			     (32 - pbi1);
 826
 827		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 828		return true;
 829	}
 830#else
 831	if (v6dst->s6_addr16[0] == htons(0x2002)) {
 832		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 833		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 834		return true;
 835	}
 836#endif
 837	return false;
 838}
 839
 840static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 841			     const struct in6_addr *v6dst)
 842{
 843	__be32 dst = 0;
 844	check_6rd(tunnel, v6dst, &dst);
 845	return dst;
 846}
 847
 848/*
 849 *	This function assumes it is being called from dev_queue_xmit()
 850 *	and that skb is filled properly by that function.
 851 */
 852
 853static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 854				     struct net_device *dev)
 855{
 856	struct ip_tunnel *tunnel = netdev_priv(dev);
 857	const struct iphdr  *tiph = &tunnel->parms.iph;
 858	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 859	u8     tos = tunnel->parms.iph.tos;
 860	__be16 df = tiph->frag_off;
 861	struct rtable *rt;		/* Route to the other host */
 862	struct net_device *tdev;	/* Device to other host */
 863	unsigned int max_headroom;	/* The extra header space needed */
 864	__be32 dst = tiph->daddr;
 865	struct flowi4 fl4;
 866	int    mtu;
 867	const struct in6_addr *addr6;
 868	int addr_type;
 869	u8 ttl;
 870	u8 protocol = IPPROTO_IPV6;
 871	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 872
 873	if (tos == 1)
 874		tos = ipv6_get_dsfield(iph6);
 875
 876	/* ISATAP (RFC4214) - must come before 6to4 */
 877	if (dev->priv_flags & IFF_ISATAP) {
 878		struct neighbour *neigh = NULL;
 879		bool do_tx_error = false;
 880
 881		if (skb_dst(skb))
 882			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 883
 884		if (!neigh) {
 885			net_dbg_ratelimited("nexthop == NULL\n");
 886			goto tx_error;
 887		}
 888
 889		addr6 = (const struct in6_addr *)&neigh->primary_key;
 890		addr_type = ipv6_addr_type(addr6);
 891
 892		if ((addr_type & IPV6_ADDR_UNICAST) &&
 893		     ipv6_addr_is_isatap(addr6))
 894			dst = addr6->s6_addr32[3];
 895		else
 896			do_tx_error = true;
 897
 898		neigh_release(neigh);
 899		if (do_tx_error)
 900			goto tx_error;
 901	}
 902
 903	if (!dst)
 904		dst = try_6rd(tunnel, &iph6->daddr);
 905
 906	if (!dst) {
 907		struct neighbour *neigh = NULL;
 908		bool do_tx_error = false;
 909
 910		if (skb_dst(skb))
 911			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 912
 913		if (!neigh) {
 914			net_dbg_ratelimited("nexthop == NULL\n");
 915			goto tx_error;
 916		}
 917
 918		addr6 = (const struct in6_addr *)&neigh->primary_key;
 919		addr_type = ipv6_addr_type(addr6);
 920
 921		if (addr_type == IPV6_ADDR_ANY) {
 922			addr6 = &ipv6_hdr(skb)->daddr;
 923			addr_type = ipv6_addr_type(addr6);
 924		}
 925
 926		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 927			dst = addr6->s6_addr32[3];
 928		else
 929			do_tx_error = true;
 930
 931		neigh_release(neigh);
 932		if (do_tx_error)
 933			goto tx_error;
 934	}
 935
 936	flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
 937			   RT_TOS(tos), RT_SCOPE_UNIVERSE, IPPROTO_IPV6,
 938			   0, dst, tiph->saddr, 0, 0,
 939			   sock_net_uid(tunnel->net, NULL));
 940
 941	rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
 942	if (!rt) {
 943		rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
 944		if (IS_ERR(rt)) {
 945			DEV_STATS_INC(dev, tx_carrier_errors);
 946			goto tx_error_icmp;
 947		}
 948		dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
 949	}
 950
 951	if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
 952		ip_rt_put(rt);
 953		DEV_STATS_INC(dev, tx_carrier_errors);
 954		goto tx_error_icmp;
 955	}
 956	tdev = rt->dst.dev;
 957
 958	if (tdev == dev) {
 959		ip_rt_put(rt);
 960		DEV_STATS_INC(dev, collisions);
 961		goto tx_error;
 962	}
 963
 964	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
 965		ip_rt_put(rt);
 966		goto tx_error;
 967	}
 968
 969	if (df) {
 970		mtu = dst_mtu(&rt->dst) - t_hlen;
 971
 972		if (mtu < IPV4_MIN_MTU) {
 973			DEV_STATS_INC(dev, collisions);
 974			ip_rt_put(rt);
 975			goto tx_error;
 976		}
 977
 978		if (mtu < IPV6_MIN_MTU) {
 979			mtu = IPV6_MIN_MTU;
 980			df = 0;
 981		}
 982
 983		if (tunnel->parms.iph.daddr)
 984			skb_dst_update_pmtu_no_confirm(skb, mtu);
 985
 986		if (skb->len > mtu && !skb_is_gso(skb)) {
 987			icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 988			ip_rt_put(rt);
 989			goto tx_error;
 990		}
 991	}
 992
 993	if (tunnel->err_count > 0) {
 994		if (time_before(jiffies,
 995				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 996			tunnel->err_count--;
 997			dst_link_failure(skb);
 998		} else
 999			tunnel->err_count = 0;
1000	}
1001
1002	/*
1003	 * Okay, now see if we can stuff it in the buffer as-is.
1004	 */
1005	max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
1006
1007	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1008	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1009		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
1010		if (!new_skb) {
1011			ip_rt_put(rt);
1012			DEV_STATS_INC(dev, tx_dropped);
1013			kfree_skb(skb);
1014			return NETDEV_TX_OK;
1015		}
1016		if (skb->sk)
1017			skb_set_owner_w(new_skb, skb->sk);
1018		dev_kfree_skb(skb);
1019		skb = new_skb;
1020		iph6 = ipv6_hdr(skb);
1021	}
1022	ttl = tiph->ttl;
1023	if (ttl == 0)
1024		ttl = iph6->hop_limit;
1025	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
1026
1027	if (ip_tunnel_encap(skb, &tunnel->encap, &protocol, &fl4) < 0) {
1028		ip_rt_put(rt);
1029		goto tx_error;
1030	}
1031
1032	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1033
1034	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
1035		      df, !net_eq(tunnel->net, dev_net(dev)));
1036	return NETDEV_TX_OK;
1037
1038tx_error_icmp:
1039	dst_link_failure(skb);
1040tx_error:
1041	kfree_skb(skb);
1042	DEV_STATS_INC(dev, tx_errors);
1043	return NETDEV_TX_OK;
1044}
1045
1046static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1047				     struct net_device *dev, u8 ipproto)
1048{
1049	struct ip_tunnel *tunnel = netdev_priv(dev);
1050	const struct iphdr  *tiph = &tunnel->parms.iph;
1051
1052	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1053		goto tx_error;
1054
1055	skb_set_inner_ipproto(skb, ipproto);
1056
1057	ip_tunnel_xmit(skb, dev, tiph, ipproto);
1058	return NETDEV_TX_OK;
1059tx_error:
1060	kfree_skb(skb);
1061	DEV_STATS_INC(dev, tx_errors);
1062	return NETDEV_TX_OK;
1063}
1064
1065static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1066				   struct net_device *dev)
1067{
1068	if (!pskb_inet_may_pull(skb))
1069		goto tx_err;
1070
1071	switch (skb->protocol) {
1072	case htons(ETH_P_IP):
1073		sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1074		break;
1075	case htons(ETH_P_IPV6):
1076		ipip6_tunnel_xmit(skb, dev);
1077		break;
1078#if IS_ENABLED(CONFIG_MPLS)
1079	case htons(ETH_P_MPLS_UC):
1080		sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1081		break;
1082#endif
1083	default:
1084		goto tx_err;
1085	}
1086
1087	return NETDEV_TX_OK;
1088
1089tx_err:
1090	DEV_STATS_INC(dev, tx_errors);
1091	kfree_skb(skb);
1092	return NETDEV_TX_OK;
1093
1094}
1095
1096static void ipip6_tunnel_bind_dev(struct net_device *dev)
1097{
1098	struct ip_tunnel *tunnel = netdev_priv(dev);
1099	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1100	struct net_device *tdev = NULL;
1101	int hlen = LL_MAX_HEADER;
1102	const struct iphdr *iph;
1103	struct flowi4 fl4;
1104
1105	iph = &tunnel->parms.iph;
1106
1107	if (iph->daddr) {
1108		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1109							  NULL,
1110							  iph->daddr, iph->saddr,
1111							  0, 0,
1112							  IPPROTO_IPV6,
1113							  RT_TOS(iph->tos),
1114							  tunnel->parms.link);
1115
1116		if (!IS_ERR(rt)) {
1117			tdev = rt->dst.dev;
1118			ip_rt_put(rt);
1119		}
1120		dev->flags |= IFF_POINTOPOINT;
1121	}
1122
1123	if (!tdev && tunnel->parms.link)
1124		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1125
1126	if (tdev && !netif_is_l3_master(tdev)) {
1127		int mtu;
1128
1129		mtu = tdev->mtu - t_hlen;
1130		if (mtu < IPV6_MIN_MTU)
1131			mtu = IPV6_MIN_MTU;
1132		WRITE_ONCE(dev->mtu, mtu);
1133		hlen = tdev->hard_header_len + tdev->needed_headroom;
1134	}
1135	dev->needed_headroom = t_hlen + hlen;
1136}
1137
1138static void ipip6_tunnel_update(struct ip_tunnel *t, struct ip_tunnel_parm *p,
 
1139				__u32 fwmark)
1140{
1141	struct net *net = t->net;
1142	struct sit_net *sitn = net_generic(net, sit_net_id);
1143
1144	ipip6_tunnel_unlink(sitn, t);
1145	synchronize_net();
1146	t->parms.iph.saddr = p->iph.saddr;
1147	t->parms.iph.daddr = p->iph.daddr;
1148	__dev_addr_set(t->dev, &p->iph.saddr, 4);
1149	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1150	ipip6_tunnel_link(sitn, t);
1151	t->parms.iph.ttl = p->iph.ttl;
1152	t->parms.iph.tos = p->iph.tos;
1153	t->parms.iph.frag_off = p->iph.frag_off;
1154	if (t->parms.link != p->link || t->fwmark != fwmark) {
1155		t->parms.link = p->link;
1156		t->fwmark = fwmark;
1157		ipip6_tunnel_bind_dev(t->dev);
1158	}
1159	dst_cache_reset(&t->dst_cache);
1160	netdev_state_change(t->dev);
1161}
1162
1163#ifdef CONFIG_IPV6_SIT_6RD
1164static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1165				   struct ip_tunnel_6rd *ip6rd)
1166{
1167	struct in6_addr prefix;
1168	__be32 relay_prefix;
1169
1170	if (ip6rd->relay_prefixlen > 32 ||
1171	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1172		return -EINVAL;
1173
1174	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1175	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1176		return -EINVAL;
1177	if (ip6rd->relay_prefixlen)
1178		relay_prefix = ip6rd->relay_prefix &
1179			       htonl(0xffffffffUL <<
1180				     (32 - ip6rd->relay_prefixlen));
1181	else
1182		relay_prefix = 0;
1183	if (relay_prefix != ip6rd->relay_prefix)
1184		return -EINVAL;
1185
1186	t->ip6rd.prefix = prefix;
1187	t->ip6rd.relay_prefix = relay_prefix;
1188	t->ip6rd.prefixlen = ip6rd->prefixlen;
1189	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1190	dst_cache_reset(&t->dst_cache);
1191	netdev_state_change(t->dev);
1192	return 0;
1193}
1194
1195static int
1196ipip6_tunnel_get6rd(struct net_device *dev, struct ip_tunnel_parm __user *data)
1197{
1198	struct ip_tunnel *t = netdev_priv(dev);
 
1199	struct ip_tunnel_6rd ip6rd;
1200	struct ip_tunnel_parm p;
1201
1202	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1203		if (copy_from_user(&p, data, sizeof(p)))
1204			return -EFAULT;
1205		t = ipip6_tunnel_locate(t->net, &p, 0);
1206	}
1207	if (!t)
1208		t = netdev_priv(dev);
1209
1210	ip6rd.prefix = t->ip6rd.prefix;
1211	ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1212	ip6rd.prefixlen = t->ip6rd.prefixlen;
1213	ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1214	if (copy_to_user(data, &ip6rd, sizeof(ip6rd)))
1215		return -EFAULT;
1216	return 0;
1217}
1218
1219static int
1220ipip6_tunnel_6rdctl(struct net_device *dev, struct ip_tunnel_6rd __user *data,
1221		    int cmd)
1222{
1223	struct ip_tunnel *t = netdev_priv(dev);
1224	struct ip_tunnel_6rd ip6rd;
1225	int err;
1226
1227	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1228		return -EPERM;
1229	if (copy_from_user(&ip6rd, data, sizeof(ip6rd)))
1230		return -EFAULT;
1231
1232	if (cmd != SIOCDEL6RD) {
1233		err = ipip6_tunnel_update_6rd(t, &ip6rd);
1234		if (err < 0)
1235			return err;
1236	} else
1237		ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev));
1238	return 0;
1239}
1240
1241#endif /* CONFIG_IPV6_SIT_6RD */
1242
1243static bool ipip6_valid_ip_proto(u8 ipproto)
1244{
1245	return ipproto == IPPROTO_IPV6 ||
1246		ipproto == IPPROTO_IPIP ||
1247#if IS_ENABLED(CONFIG_MPLS)
1248		ipproto == IPPROTO_MPLS ||
1249#endif
1250		ipproto == 0;
1251}
1252
1253static int
1254__ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm *p)
1255{
1256	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1257		return -EPERM;
1258
1259	if (!ipip6_valid_ip_proto(p->iph.protocol))
1260		return -EINVAL;
1261	if (p->iph.version != 4 ||
1262	    p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
1263		return -EINVAL;
1264
1265	if (p->iph.ttl)
1266		p->iph.frag_off |= htons(IP_DF);
1267	return 0;
1268}
1269
1270static int
1271ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm *p)
1272{
1273	struct ip_tunnel *t = netdev_priv(dev);
1274
1275	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
1276		t = ipip6_tunnel_locate(t->net, p, 0);
1277	if (!t)
1278		t = netdev_priv(dev);
1279	memcpy(p, &t->parms, sizeof(*p));
1280	return 0;
1281}
1282
1283static int
1284ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm *p)
1285{
1286	struct ip_tunnel *t = netdev_priv(dev);
1287	int err;
1288
1289	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1290	if (err)
1291		return err;
1292
1293	t = ipip6_tunnel_locate(t->net, p, 1);
1294	if (!t)
1295		return -ENOBUFS;
1296	return 0;
1297}
1298
1299static int
1300ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm *p)
1301{
1302	struct ip_tunnel *t = netdev_priv(dev);
1303	int err;
1304
1305	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1306	if (err)
1307		return err;
1308
1309	t = ipip6_tunnel_locate(t->net, p, 0);
1310	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1311		if (!t)
1312			return -ENOENT;
1313	} else {
1314		if (t) {
1315			if (t->dev != dev)
1316				return -EEXIST;
1317		} else {
1318			if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) ||
1319			    (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr))
1320				return -EINVAL;
1321			t = netdev_priv(dev);
1322		}
1323
1324		ipip6_tunnel_update(t, p, t->fwmark);
1325	}
1326
1327	return 0;
1328}
1329
1330static int
1331ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm *p)
1332{
1333	struct ip_tunnel *t = netdev_priv(dev);
1334
1335	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1336		return -EPERM;
1337
1338	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1339		t = ipip6_tunnel_locate(t->net, p, 0);
1340		if (!t)
1341			return -ENOENT;
1342		if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev))
1343			return -EPERM;
1344		dev = t->dev;
1345	}
1346	unregister_netdevice(dev);
1347	return 0;
1348}
1349
1350static int
1351ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
 
1352{
1353	switch (cmd) {
1354	case SIOCGETTUNNEL:
1355		return ipip6_tunnel_get(dev, p);
1356	case SIOCADDTUNNEL:
1357		return ipip6_tunnel_add(dev, p);
1358	case SIOCCHGTUNNEL:
1359		return ipip6_tunnel_change(dev, p);
1360	case SIOCDELTUNNEL:
1361		return ipip6_tunnel_del(dev, p);
1362	default:
1363		return -EINVAL;
1364	}
1365}
1366
1367static int
1368ipip6_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1369			    void __user *data, int cmd)
1370{
1371	switch (cmd) {
1372	case SIOCGETTUNNEL:
1373	case SIOCADDTUNNEL:
1374	case SIOCCHGTUNNEL:
1375	case SIOCDELTUNNEL:
1376		return ip_tunnel_siocdevprivate(dev, ifr, data, cmd);
1377	case SIOCGETPRL:
1378		return ipip6_tunnel_get_prl(dev, data);
1379	case SIOCADDPRL:
1380	case SIOCDELPRL:
1381	case SIOCCHGPRL:
1382		return ipip6_tunnel_prl_ctl(dev, data, cmd);
1383#ifdef CONFIG_IPV6_SIT_6RD
1384	case SIOCGET6RD:
1385		return ipip6_tunnel_get6rd(dev, data);
1386	case SIOCADD6RD:
1387	case SIOCCHG6RD:
1388	case SIOCDEL6RD:
1389		return ipip6_tunnel_6rdctl(dev, data, cmd);
1390#endif
1391	default:
1392		return -EINVAL;
1393	}
1394}
1395
1396static const struct net_device_ops ipip6_netdev_ops = {
1397	.ndo_init	= ipip6_tunnel_init,
1398	.ndo_uninit	= ipip6_tunnel_uninit,
1399	.ndo_start_xmit	= sit_tunnel_xmit,
1400	.ndo_siocdevprivate = ipip6_tunnel_siocdevprivate,
1401	.ndo_get_stats64 = dev_get_tstats64,
1402	.ndo_get_iflink = ip_tunnel_get_iflink,
1403	.ndo_tunnel_ctl = ipip6_tunnel_ctl,
1404};
1405
1406static void ipip6_dev_free(struct net_device *dev)
1407{
1408	struct ip_tunnel *tunnel = netdev_priv(dev);
1409
1410	dst_cache_destroy(&tunnel->dst_cache);
1411	free_percpu(dev->tstats);
1412}
1413
1414#define SIT_FEATURES (NETIF_F_SG	   | \
1415		      NETIF_F_FRAGLIST	   | \
1416		      NETIF_F_HIGHDMA	   | \
1417		      NETIF_F_GSO_SOFTWARE | \
1418		      NETIF_F_HW_CSUM)
1419
1420static void ipip6_tunnel_setup(struct net_device *dev)
1421{
1422	struct ip_tunnel *tunnel = netdev_priv(dev);
1423	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1424
1425	dev->netdev_ops		= &ipip6_netdev_ops;
1426	dev->header_ops		= &ip_tunnel_header_ops;
1427	dev->needs_free_netdev	= true;
1428	dev->priv_destructor	= ipip6_dev_free;
1429
1430	dev->type		= ARPHRD_SIT;
1431	dev->mtu		= ETH_DATA_LEN - t_hlen;
1432	dev->min_mtu		= IPV6_MIN_MTU;
1433	dev->max_mtu		= IP6_MAX_MTU - t_hlen;
1434	dev->flags		= IFF_NOARP;
1435	netif_keep_dst(dev);
1436	dev->addr_len		= 4;
1437	dev->features		|= NETIF_F_LLTX;
1438	dev->features		|= SIT_FEATURES;
1439	dev->hw_features	|= SIT_FEATURES;
 
 
1440}
1441
1442static int ipip6_tunnel_init(struct net_device *dev)
1443{
1444	struct ip_tunnel *tunnel = netdev_priv(dev);
1445	int err;
1446
1447	tunnel->dev = dev;
1448	tunnel->net = dev_net(dev);
1449	strcpy(tunnel->parms.name, dev->name);
1450
1451	ipip6_tunnel_bind_dev(dev);
1452	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1453	if (!dev->tstats)
1454		return -ENOMEM;
1455
1456	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1457	if (err) {
1458		free_percpu(dev->tstats);
1459		dev->tstats = NULL;
1460		return err;
1461	}
1462	netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
 
1463	return 0;
1464}
1465
1466static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1467{
1468	struct ip_tunnel *tunnel = netdev_priv(dev);
1469	struct iphdr *iph = &tunnel->parms.iph;
1470	struct net *net = dev_net(dev);
1471	struct sit_net *sitn = net_generic(net, sit_net_id);
1472
1473	iph->version		= 4;
1474	iph->protocol		= IPPROTO_IPV6;
1475	iph->ihl		= 5;
1476	iph->ttl		= 64;
1477
1478	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1479}
1480
1481static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1482			  struct netlink_ext_ack *extack)
1483{
1484	u8 proto;
1485
1486	if (!data || !data[IFLA_IPTUN_PROTO])
1487		return 0;
1488
1489	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1490	if (!ipip6_valid_ip_proto(proto))
1491		return -EINVAL;
1492
1493	return 0;
1494}
1495
1496static void ipip6_netlink_parms(struct nlattr *data[],
1497				struct ip_tunnel_parm *parms,
1498				__u32 *fwmark)
1499{
1500	memset(parms, 0, sizeof(*parms));
1501
1502	parms->iph.version = 4;
1503	parms->iph.protocol = IPPROTO_IPV6;
1504	parms->iph.ihl = 5;
1505	parms->iph.ttl = 64;
1506
1507	if (!data)
1508		return;
1509
1510	ip_tunnel_netlink_parms(data, parms);
1511
1512	if (data[IFLA_IPTUN_FWMARK])
1513		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1514}
1515
1516#ifdef CONFIG_IPV6_SIT_6RD
1517/* This function returns true when 6RD attributes are present in the nl msg */
1518static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1519				    struct ip_tunnel_6rd *ip6rd)
1520{
1521	bool ret = false;
1522	memset(ip6rd, 0, sizeof(*ip6rd));
1523
1524	if (!data)
1525		return ret;
1526
1527	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1528		ret = true;
1529		ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1530	}
1531
1532	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1533		ret = true;
1534		ip6rd->relay_prefix =
1535			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1536	}
1537
1538	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1539		ret = true;
1540		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1541	}
1542
1543	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1544		ret = true;
1545		ip6rd->relay_prefixlen =
1546			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1547	}
1548
1549	return ret;
1550}
1551#endif
1552
1553static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1554			 struct nlattr *tb[], struct nlattr *data[],
1555			 struct netlink_ext_ack *extack)
1556{
1557	struct net *net = dev_net(dev);
1558	struct ip_tunnel *nt;
1559	struct ip_tunnel_encap ipencap;
1560#ifdef CONFIG_IPV6_SIT_6RD
1561	struct ip_tunnel_6rd ip6rd;
1562#endif
1563	int err;
1564
1565	nt = netdev_priv(dev);
1566
1567	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1568		err = ip_tunnel_encap_setup(nt, &ipencap);
1569		if (err < 0)
1570			return err;
1571	}
1572
1573	ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1574
1575	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1576		return -EEXIST;
1577
1578	err = ipip6_tunnel_create(dev);
1579	if (err < 0)
1580		return err;
1581
1582	if (tb[IFLA_MTU]) {
1583		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1584
1585		if (mtu >= IPV6_MIN_MTU &&
1586		    mtu <= IP6_MAX_MTU - dev->hard_header_len)
1587			dev->mtu = mtu;
1588	}
1589
1590#ifdef CONFIG_IPV6_SIT_6RD
1591	if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1592		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1593		if (err < 0)
1594			unregister_netdevice_queue(dev, NULL);
1595	}
1596#endif
1597
1598	return err;
1599}
1600
1601static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1602			    struct nlattr *data[],
1603			    struct netlink_ext_ack *extack)
1604{
1605	struct ip_tunnel *t = netdev_priv(dev);
1606	struct ip_tunnel_parm p;
1607	struct ip_tunnel_encap ipencap;
 
1608	struct net *net = t->net;
1609	struct sit_net *sitn = net_generic(net, sit_net_id);
1610#ifdef CONFIG_IPV6_SIT_6RD
1611	struct ip_tunnel_6rd ip6rd;
1612#endif
1613	__u32 fwmark = t->fwmark;
1614	int err;
1615
1616	if (dev == sitn->fb_tunnel_dev)
1617		return -EINVAL;
1618
1619	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1620		err = ip_tunnel_encap_setup(t, &ipencap);
1621		if (err < 0)
1622			return err;
1623	}
1624
1625	ipip6_netlink_parms(data, &p, &fwmark);
1626
1627	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1628	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1629		return -EINVAL;
1630
1631	t = ipip6_tunnel_locate(net, &p, 0);
1632
1633	if (t) {
1634		if (t->dev != dev)
1635			return -EEXIST;
1636	} else
1637		t = netdev_priv(dev);
1638
1639	ipip6_tunnel_update(t, &p, fwmark);
1640
1641#ifdef CONFIG_IPV6_SIT_6RD
1642	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1643		return ipip6_tunnel_update_6rd(t, &ip6rd);
1644#endif
1645
1646	return 0;
1647}
1648
1649static size_t ipip6_get_size(const struct net_device *dev)
1650{
1651	return
1652		/* IFLA_IPTUN_LINK */
1653		nla_total_size(4) +
1654		/* IFLA_IPTUN_LOCAL */
1655		nla_total_size(4) +
1656		/* IFLA_IPTUN_REMOTE */
1657		nla_total_size(4) +
1658		/* IFLA_IPTUN_TTL */
1659		nla_total_size(1) +
1660		/* IFLA_IPTUN_TOS */
1661		nla_total_size(1) +
1662		/* IFLA_IPTUN_PMTUDISC */
1663		nla_total_size(1) +
1664		/* IFLA_IPTUN_FLAGS */
1665		nla_total_size(2) +
1666		/* IFLA_IPTUN_PROTO */
1667		nla_total_size(1) +
1668#ifdef CONFIG_IPV6_SIT_6RD
1669		/* IFLA_IPTUN_6RD_PREFIX */
1670		nla_total_size(sizeof(struct in6_addr)) +
1671		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1672		nla_total_size(4) +
1673		/* IFLA_IPTUN_6RD_PREFIXLEN */
1674		nla_total_size(2) +
1675		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1676		nla_total_size(2) +
1677#endif
1678		/* IFLA_IPTUN_ENCAP_TYPE */
1679		nla_total_size(2) +
1680		/* IFLA_IPTUN_ENCAP_FLAGS */
1681		nla_total_size(2) +
1682		/* IFLA_IPTUN_ENCAP_SPORT */
1683		nla_total_size(2) +
1684		/* IFLA_IPTUN_ENCAP_DPORT */
1685		nla_total_size(2) +
1686		/* IFLA_IPTUN_FWMARK */
1687		nla_total_size(4) +
1688		0;
1689}
1690
1691static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1692{
1693	struct ip_tunnel *tunnel = netdev_priv(dev);
1694	struct ip_tunnel_parm *parm = &tunnel->parms;
1695
1696	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1697	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1698	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1699	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1700	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1701	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1702		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1703	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1704	    nla_put_be16(skb, IFLA_IPTUN_FLAGS, parm->i_flags) ||
 
1705	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1706		goto nla_put_failure;
1707
1708#ifdef CONFIG_IPV6_SIT_6RD
1709	if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1710			     &tunnel->ip6rd.prefix) ||
1711	    nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1712			    tunnel->ip6rd.relay_prefix) ||
1713	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1714			tunnel->ip6rd.prefixlen) ||
1715	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1716			tunnel->ip6rd.relay_prefixlen))
1717		goto nla_put_failure;
1718#endif
1719
1720	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1721			tunnel->encap.type) ||
1722	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1723			tunnel->encap.sport) ||
1724	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1725			tunnel->encap.dport) ||
1726	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1727			tunnel->encap.flags))
1728		goto nla_put_failure;
1729
1730	return 0;
1731
1732nla_put_failure:
1733	return -EMSGSIZE;
1734}
1735
1736static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1737	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1738	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1739	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1740	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1741	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1742	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1743	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1744	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1745#ifdef CONFIG_IPV6_SIT_6RD
1746	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1747	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1748	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1749	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1750#endif
1751	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1752	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1753	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1754	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1755	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
1756};
1757
1758static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1759{
1760	struct net *net = dev_net(dev);
1761	struct sit_net *sitn = net_generic(net, sit_net_id);
1762
1763	if (dev != sitn->fb_tunnel_dev)
1764		unregister_netdevice_queue(dev, head);
1765}
1766
1767static struct rtnl_link_ops sit_link_ops __read_mostly = {
1768	.kind		= "sit",
1769	.maxtype	= IFLA_IPTUN_MAX,
1770	.policy		= ipip6_policy,
1771	.priv_size	= sizeof(struct ip_tunnel),
1772	.setup		= ipip6_tunnel_setup,
1773	.validate	= ipip6_validate,
1774	.newlink	= ipip6_newlink,
1775	.changelink	= ipip6_changelink,
1776	.get_size	= ipip6_get_size,
1777	.fill_info	= ipip6_fill_info,
1778	.dellink	= ipip6_dellink,
1779	.get_link_net	= ip_tunnel_get_link_net,
1780};
1781
1782static struct xfrm_tunnel sit_handler __read_mostly = {
1783	.handler	=	ipip6_rcv,
1784	.err_handler	=	ipip6_err,
1785	.priority	=	1,
1786};
1787
1788static struct xfrm_tunnel ipip_handler __read_mostly = {
1789	.handler	=	ipip_rcv,
1790	.err_handler	=	ipip6_err,
1791	.priority	=	2,
1792};
1793
1794#if IS_ENABLED(CONFIG_MPLS)
1795static struct xfrm_tunnel mplsip_handler __read_mostly = {
1796	.handler	=	mplsip_rcv,
1797	.err_handler	=	ipip6_err,
1798	.priority	=	2,
1799};
1800#endif
1801
1802static void __net_exit sit_destroy_tunnels(struct net *net,
1803					   struct list_head *head)
1804{
1805	struct sit_net *sitn = net_generic(net, sit_net_id);
1806	struct net_device *dev, *aux;
1807	int prio;
1808
1809	for_each_netdev_safe(net, dev, aux)
1810		if (dev->rtnl_link_ops == &sit_link_ops)
1811			unregister_netdevice_queue(dev, head);
1812
1813	for (prio = 0; prio < 4; prio++) {
1814		int h;
1815		for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1816			struct ip_tunnel *t;
1817
1818			t = rtnl_dereference(sitn->tunnels[prio][h]);
1819			while (t) {
1820				/* If dev is in the same netns, it has already
1821				 * been added to the list by the previous loop.
1822				 */
1823				if (!net_eq(dev_net(t->dev), net))
1824					unregister_netdevice_queue(t->dev,
1825								   head);
1826				t = rtnl_dereference(t->next);
1827			}
1828		}
1829	}
1830}
1831
1832static int __net_init sit_init_net(struct net *net)
1833{
1834	struct sit_net *sitn = net_generic(net, sit_net_id);
1835	struct ip_tunnel *t;
1836	int err;
1837
1838	sitn->tunnels[0] = sitn->tunnels_wc;
1839	sitn->tunnels[1] = sitn->tunnels_l;
1840	sitn->tunnels[2] = sitn->tunnels_r;
1841	sitn->tunnels[3] = sitn->tunnels_r_l;
1842
1843	if (!net_has_fallback_tunnels(net))
1844		return 0;
1845
1846	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1847					   NET_NAME_UNKNOWN,
1848					   ipip6_tunnel_setup);
1849	if (!sitn->fb_tunnel_dev) {
1850		err = -ENOMEM;
1851		goto err_alloc_dev;
1852	}
1853	dev_net_set(sitn->fb_tunnel_dev, net);
1854	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1855	/* FB netdevice is special: we have one, and only one per netns.
1856	 * Allowing to move it to another netns is clearly unsafe.
1857	 */
1858	sitn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1859
1860	err = register_netdev(sitn->fb_tunnel_dev);
1861	if (err)
1862		goto err_reg_dev;
1863
1864	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1865	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1866
1867	t = netdev_priv(sitn->fb_tunnel_dev);
1868
1869	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1870	return 0;
1871
1872err_reg_dev:
1873	free_netdev(sitn->fb_tunnel_dev);
1874err_alloc_dev:
1875	return err;
1876}
1877
1878static void __net_exit sit_exit_batch_net(struct list_head *net_list)
 
1879{
1880	LIST_HEAD(list);
1881	struct net *net;
1882
1883	rtnl_lock();
1884	list_for_each_entry(net, net_list, exit_list)
1885		sit_destroy_tunnels(net, &list);
1886
1887	unregister_netdevice_many(&list);
1888	rtnl_unlock();
1889}
1890
1891static struct pernet_operations sit_net_ops = {
1892	.init = sit_init_net,
1893	.exit_batch = sit_exit_batch_net,
1894	.id   = &sit_net_id,
1895	.size = sizeof(struct sit_net),
1896};
1897
1898static void __exit sit_cleanup(void)
1899{
1900	rtnl_link_unregister(&sit_link_ops);
1901	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1902	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1903#if IS_ENABLED(CONFIG_MPLS)
1904	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1905#endif
1906
1907	unregister_pernet_device(&sit_net_ops);
1908	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1909}
1910
1911static int __init sit_init(void)
1912{
1913	int err;
1914
1915	pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1916
1917	err = register_pernet_device(&sit_net_ops);
1918	if (err < 0)
1919		return err;
1920	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1921	if (err < 0) {
1922		pr_info("%s: can't register ip6ip4\n", __func__);
1923		goto xfrm_tunnel_failed;
1924	}
1925	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1926	if (err < 0) {
1927		pr_info("%s: can't register ip4ip4\n", __func__);
1928		goto xfrm_tunnel4_failed;
1929	}
1930#if IS_ENABLED(CONFIG_MPLS)
1931	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1932	if (err < 0) {
1933		pr_info("%s: can't register mplsip\n", __func__);
1934		goto xfrm_tunnel_mpls_failed;
1935	}
1936#endif
1937	err = rtnl_link_register(&sit_link_ops);
1938	if (err < 0)
1939		goto rtnl_link_failed;
1940
1941out:
1942	return err;
1943
1944rtnl_link_failed:
1945#if IS_ENABLED(CONFIG_MPLS)
1946	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1947xfrm_tunnel_mpls_failed:
1948#endif
1949	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1950xfrm_tunnel4_failed:
1951	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1952xfrm_tunnel_failed:
1953	unregister_pernet_device(&sit_net_ops);
1954	goto out;
1955}
1956
1957module_init(sit_init);
1958module_exit(sit_cleanup);
1959MODULE_DESCRIPTION("IPv6-in-IPv4 tunnel SIT driver");
1960MODULE_LICENSE("GPL");
1961MODULE_ALIAS_RTNL_LINK("sit");
1962MODULE_ALIAS_NETDEV("sit0");
v6.13.7
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 *	IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
   4 *	Linux INET6 implementation
   5 *
   6 *	Authors:
   7 *	Pedro Roque		<roque@di.fc.ul.pt>
   8 *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
   9 *
  10 *	Changes:
  11 * Roger Venning <r.venning@telstra.com>:	6to4 support
  12 * Nate Thompson <nate@thebog.net>:		6to4 support
  13 * Fred Templin <fred.l.templin@boeing.com>:	isatap support
  14 */
  15
  16#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  17
  18#include <linux/module.h>
  19#include <linux/capability.h>
  20#include <linux/errno.h>
  21#include <linux/types.h>
  22#include <linux/socket.h>
  23#include <linux/sockios.h>
  24#include <linux/net.h>
  25#include <linux/in6.h>
  26#include <linux/netdevice.h>
  27#include <linux/if_arp.h>
  28#include <linux/icmp.h>
  29#include <linux/slab.h>
  30#include <linux/uaccess.h>
  31#include <linux/init.h>
  32#include <linux/netfilter_ipv4.h>
  33#include <linux/if_ether.h>
  34
  35#include <net/sock.h>
  36#include <net/snmp.h>
  37
  38#include <net/ipv6.h>
  39#include <net/protocol.h>
  40#include <net/transp_v6.h>
  41#include <net/ip6_fib.h>
  42#include <net/ip6_route.h>
  43#include <net/ndisc.h>
  44#include <net/addrconf.h>
  45#include <net/ip.h>
  46#include <net/udp.h>
  47#include <net/icmp.h>
  48#include <net/ip_tunnels.h>
  49#include <net/inet_ecn.h>
  50#include <net/xfrm.h>
  51#include <net/dsfield.h>
  52#include <net/net_namespace.h>
  53#include <net/netns/generic.h>
  54#include <net/inet_dscp.h>
  55
  56/*
  57   This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  58
  59   For comments look at net/ipv4/ip_gre.c --ANK
  60 */
  61
  62#define IP6_SIT_HASH_SIZE  16
  63#define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  64
  65static bool log_ecn_error = true;
  66module_param(log_ecn_error, bool, 0644);
  67MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  68
  69static int ipip6_tunnel_init(struct net_device *dev);
  70static void ipip6_tunnel_setup(struct net_device *dev);
  71static void ipip6_dev_free(struct net_device *dev);
  72static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
  73		      __be32 *v4dst);
  74static struct rtnl_link_ops sit_link_ops __read_mostly;
  75
  76static unsigned int sit_net_id __read_mostly;
  77struct sit_net {
  78	struct ip_tunnel __rcu *tunnels_r_l[IP6_SIT_HASH_SIZE];
  79	struct ip_tunnel __rcu *tunnels_r[IP6_SIT_HASH_SIZE];
  80	struct ip_tunnel __rcu *tunnels_l[IP6_SIT_HASH_SIZE];
  81	struct ip_tunnel __rcu *tunnels_wc[1];
  82	struct ip_tunnel __rcu **tunnels[4];
  83
  84	struct net_device *fb_tunnel_dev;
  85};
  86
  87static inline struct sit_net *dev_to_sit_net(struct net_device *dev)
  88{
  89	struct ip_tunnel *t = netdev_priv(dev);
  90
  91	return net_generic(t->net, sit_net_id);
  92}
  93
  94/*
  95 * Must be invoked with rcu_read_lock
  96 */
  97static struct ip_tunnel *ipip6_tunnel_lookup(struct net *net,
  98					     struct net_device *dev,
  99					     __be32 remote, __be32 local,
 100					     int sifindex)
 101{
 102	unsigned int h0 = HASH(remote);
 103	unsigned int h1 = HASH(local);
 104	struct ip_tunnel *t;
 105	struct sit_net *sitn = net_generic(net, sit_net_id);
 106	int ifindex = dev ? dev->ifindex : 0;
 107
 108	for_each_ip_tunnel_rcu(t, sitn->tunnels_r_l[h0 ^ h1]) {
 109		if (local == t->parms.iph.saddr &&
 110		    remote == t->parms.iph.daddr &&
 111		    (!dev || !t->parms.link || ifindex == t->parms.link ||
 112		     sifindex == t->parms.link) &&
 113		    (t->dev->flags & IFF_UP))
 114			return t;
 115	}
 116	for_each_ip_tunnel_rcu(t, sitn->tunnels_r[h0]) {
 117		if (remote == t->parms.iph.daddr &&
 118		    (!dev || !t->parms.link || ifindex == t->parms.link ||
 119		     sifindex == t->parms.link) &&
 120		    (t->dev->flags & IFF_UP))
 121			return t;
 122	}
 123	for_each_ip_tunnel_rcu(t, sitn->tunnels_l[h1]) {
 124		if (local == t->parms.iph.saddr &&
 125		    (!dev || !t->parms.link || ifindex == t->parms.link ||
 126		     sifindex == t->parms.link) &&
 127		    (t->dev->flags & IFF_UP))
 128			return t;
 129	}
 130	t = rcu_dereference(sitn->tunnels_wc[0]);
 131	if (t && (t->dev->flags & IFF_UP))
 132		return t;
 133	return NULL;
 134}
 135
 136static struct ip_tunnel __rcu **
 137__ipip6_bucket(struct sit_net *sitn, struct ip_tunnel_parm_kern *parms)
 138{
 139	__be32 remote = parms->iph.daddr;
 140	__be32 local = parms->iph.saddr;
 141	unsigned int h = 0;
 142	int prio = 0;
 143
 144	if (remote) {
 145		prio |= 2;
 146		h ^= HASH(remote);
 147	}
 148	if (local) {
 149		prio |= 1;
 150		h ^= HASH(local);
 151	}
 152	return &sitn->tunnels[prio][h];
 153}
 154
 155static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
 156		struct ip_tunnel *t)
 157{
 158	return __ipip6_bucket(sitn, &t->parms);
 159}
 160
 161static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
 162{
 163	struct ip_tunnel __rcu **tp;
 164	struct ip_tunnel *iter;
 165
 166	for (tp = ipip6_bucket(sitn, t);
 167	     (iter = rtnl_dereference(*tp)) != NULL;
 168	     tp = &iter->next) {
 169		if (t == iter) {
 170			rcu_assign_pointer(*tp, t->next);
 171			break;
 172		}
 173	}
 174}
 175
 176static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
 177{
 178	struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
 179
 180	rcu_assign_pointer(t->next, rtnl_dereference(*tp));
 181	rcu_assign_pointer(*tp, t);
 182}
 183
 184static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
 185{
 186#ifdef CONFIG_IPV6_SIT_6RD
 187	struct ip_tunnel *t = netdev_priv(dev);
 188
 189	if (dev == sitn->fb_tunnel_dev || !sitn->fb_tunnel_dev) {
 190		ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
 191		t->ip6rd.relay_prefix = 0;
 192		t->ip6rd.prefixlen = 16;
 193		t->ip6rd.relay_prefixlen = 0;
 194	} else {
 195		struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
 196		memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
 197	}
 198#endif
 199}
 200
 201static int ipip6_tunnel_create(struct net_device *dev)
 202{
 203	struct ip_tunnel *t = netdev_priv(dev);
 204	struct net *net = dev_net(dev);
 205	struct sit_net *sitn = net_generic(net, sit_net_id);
 206	int err;
 207
 208	__dev_addr_set(dev, &t->parms.iph.saddr, 4);
 209	memcpy(dev->broadcast, &t->parms.iph.daddr, 4);
 210
 211	if (test_bit(IP_TUNNEL_SIT_ISATAP_BIT, t->parms.i_flags))
 212		dev->priv_flags |= IFF_ISATAP;
 213
 214	dev->rtnl_link_ops = &sit_link_ops;
 215
 216	err = register_netdevice(dev);
 217	if (err < 0)
 218		goto out;
 219
 220	ipip6_tunnel_clone_6rd(dev, sitn);
 221
 222	ipip6_tunnel_link(sitn, t);
 223	return 0;
 224
 225out:
 226	return err;
 227}
 228
 229static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
 230					     struct ip_tunnel_parm_kern *parms,
 231					     int create)
 232{
 233	__be32 remote = parms->iph.daddr;
 234	__be32 local = parms->iph.saddr;
 235	struct ip_tunnel *t, *nt;
 236	struct ip_tunnel __rcu **tp;
 237	struct net_device *dev;
 238	char name[IFNAMSIZ];
 239	struct sit_net *sitn = net_generic(net, sit_net_id);
 240
 241	for (tp = __ipip6_bucket(sitn, parms);
 242	    (t = rtnl_dereference(*tp)) != NULL;
 243	     tp = &t->next) {
 244		if (local == t->parms.iph.saddr &&
 245		    remote == t->parms.iph.daddr &&
 246		    parms->link == t->parms.link) {
 247			if (create)
 248				return NULL;
 249			else
 250				return t;
 251		}
 252	}
 253	if (!create)
 254		goto failed;
 255
 256	if (parms->name[0]) {
 257		if (!dev_valid_name(parms->name))
 258			goto failed;
 259		strscpy(name, parms->name, IFNAMSIZ);
 260	} else {
 261		strcpy(name, "sit%d");
 262	}
 263	dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
 264			   ipip6_tunnel_setup);
 265	if (!dev)
 266		return NULL;
 267
 268	dev_net_set(dev, net);
 269
 270	nt = netdev_priv(dev);
 271
 272	nt->parms = *parms;
 273	if (ipip6_tunnel_create(dev) < 0)
 274		goto failed_free;
 275
 276	if (!parms->name[0])
 277		strcpy(parms->name, dev->name);
 278
 279	return nt;
 280
 281failed_free:
 282	free_netdev(dev);
 283failed:
 284	return NULL;
 285}
 286
 287#define for_each_prl_rcu(start)			\
 288	for (prl = rcu_dereference(start);	\
 289	     prl;				\
 290	     prl = rcu_dereference(prl->next))
 291
 292static struct ip_tunnel_prl_entry *
 293__ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 294{
 295	struct ip_tunnel_prl_entry *prl;
 296
 297	for_each_prl_rcu(t->prl)
 298		if (prl->addr == addr)
 299			break;
 300	return prl;
 301
 302}
 303
 304static int ipip6_tunnel_get_prl(struct net_device *dev, struct ip_tunnel_prl __user *a)
 305{
 306	struct ip_tunnel *t = netdev_priv(dev);
 307	struct ip_tunnel_prl kprl, *kp;
 308	struct ip_tunnel_prl_entry *prl;
 309	unsigned int cmax, c = 0, ca, len;
 310	int ret = 0;
 311
 312	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
 313		return -EINVAL;
 314
 315	if (copy_from_user(&kprl, a, sizeof(kprl)))
 316		return -EFAULT;
 317	cmax = kprl.datalen / sizeof(kprl);
 318	if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
 319		cmax = 1;
 320
 321	/* For simple GET or for root users,
 322	 * we try harder to allocate.
 323	 */
 324	kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 325		kcalloc(cmax, sizeof(*kp), GFP_KERNEL_ACCOUNT | __GFP_NOWARN) :
 326		NULL;
 327
 328	ca = min(t->prl_count, cmax);
 329
 330	if (!kp) {
 331		/* We don't try hard to allocate much memory for
 332		 * non-root users.
 333		 * For root users, retry allocating enough memory for
 334		 * the answer.
 335		 */
 336		kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC | __GFP_ACCOUNT |
 337					      __GFP_NOWARN);
 338		if (!kp) {
 339			ret = -ENOMEM;
 340			goto out;
 341		}
 342	}
 343
 344	rcu_read_lock();
 345	for_each_prl_rcu(t->prl) {
 346		if (c >= cmax)
 347			break;
 348		if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
 349			continue;
 350		kp[c].addr = prl->addr;
 351		kp[c].flags = prl->flags;
 352		c++;
 353		if (kprl.addr != htonl(INADDR_ANY))
 354			break;
 355	}
 356
 357	rcu_read_unlock();
 358
 359	len = sizeof(*kp) * c;
 360	ret = 0;
 361	if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
 362		ret = -EFAULT;
 363
 364	kfree(kp);
 365out:
 366	return ret;
 367}
 368
 369static int
 370ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 371{
 372	struct ip_tunnel_prl_entry *p;
 373	int err = 0;
 374
 375	if (a->addr == htonl(INADDR_ANY))
 376		return -EINVAL;
 377
 378	ASSERT_RTNL();
 379
 380	for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
 381		if (p->addr == a->addr) {
 382			if (chg) {
 383				p->flags = a->flags;
 384				goto out;
 385			}
 386			err = -EEXIST;
 387			goto out;
 388		}
 389	}
 390
 391	if (chg) {
 392		err = -ENXIO;
 393		goto out;
 394	}
 395
 396	p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 397	if (!p) {
 398		err = -ENOBUFS;
 399		goto out;
 400	}
 401
 402	p->next = t->prl;
 403	p->addr = a->addr;
 404	p->flags = a->flags;
 405	t->prl_count++;
 406	rcu_assign_pointer(t->prl, p);
 407out:
 408	return err;
 409}
 410
 411static void prl_list_destroy_rcu(struct rcu_head *head)
 412{
 413	struct ip_tunnel_prl_entry *p, *n;
 414
 415	p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
 416	do {
 417		n = rcu_dereference_protected(p->next, 1);
 418		kfree(p);
 419		p = n;
 420	} while (p);
 421}
 422
 423static int
 424ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 425{
 426	struct ip_tunnel_prl_entry *x;
 427	struct ip_tunnel_prl_entry __rcu **p;
 428	int err = 0;
 429
 430	ASSERT_RTNL();
 431
 432	if (a && a->addr != htonl(INADDR_ANY)) {
 433		for (p = &t->prl;
 434		     (x = rtnl_dereference(*p)) != NULL;
 435		     p = &x->next) {
 436			if (x->addr == a->addr) {
 437				*p = x->next;
 438				kfree_rcu(x, rcu_head);
 439				t->prl_count--;
 440				goto out;
 441			}
 442		}
 443		err = -ENXIO;
 444	} else {
 445		x = rtnl_dereference(t->prl);
 446		if (x) {
 447			t->prl_count = 0;
 448			call_rcu(&x->rcu_head, prl_list_destroy_rcu);
 449			t->prl = NULL;
 450		}
 451	}
 452out:
 453	return err;
 454}
 455
 456static int ipip6_tunnel_prl_ctl(struct net_device *dev,
 457				struct ip_tunnel_prl __user *data, int cmd)
 458{
 459	struct ip_tunnel *t = netdev_priv(dev);
 460	struct ip_tunnel_prl prl;
 461	int err;
 462
 463	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
 464		return -EPERM;
 465	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
 466		return -EINVAL;
 467
 468	if (copy_from_user(&prl, data, sizeof(prl)))
 469		return -EFAULT;
 470
 471	switch (cmd) {
 472	case SIOCDELPRL:
 473		err = ipip6_tunnel_del_prl(t, &prl);
 474		break;
 475	case SIOCADDPRL:
 476	case SIOCCHGPRL:
 477		err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 478		break;
 479	}
 480	dst_cache_reset(&t->dst_cache);
 481	netdev_state_change(dev);
 482	return err;
 483}
 484
 485static int
 486isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
 487{
 488	struct ip_tunnel_prl_entry *p;
 489	int ok = 1;
 490
 491	rcu_read_lock();
 492	p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 493	if (p) {
 494		if (p->flags & PRL_DEFAULT)
 495			skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 496		else
 497			skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 498	} else {
 499		const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 500
 501		if (ipv6_addr_is_isatap(addr6) &&
 502		    (addr6->s6_addr32[3] == iph->saddr) &&
 503		    ipv6_chk_prefix(addr6, t->dev))
 504			skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 505		else
 506			ok = 0;
 507	}
 508	rcu_read_unlock();
 509	return ok;
 510}
 511
 512static void ipip6_tunnel_uninit(struct net_device *dev)
 513{
 514	struct ip_tunnel *tunnel = netdev_priv(dev);
 515	struct sit_net *sitn = net_generic(tunnel->net, sit_net_id);
 516
 517	if (dev == sitn->fb_tunnel_dev) {
 518		RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
 519	} else {
 520		ipip6_tunnel_unlink(sitn, tunnel);
 521		ipip6_tunnel_del_prl(tunnel, NULL);
 522	}
 523	dst_cache_reset(&tunnel->dst_cache);
 524	netdev_put(dev, &tunnel->dev_tracker);
 525}
 526
 527static int ipip6_err(struct sk_buff *skb, u32 info)
 528{
 529	const struct iphdr *iph = (const struct iphdr *)skb->data;
 530	const int type = icmp_hdr(skb)->type;
 531	const int code = icmp_hdr(skb)->code;
 532	unsigned int data_len = 0;
 533	struct ip_tunnel *t;
 534	int sifindex;
 535	int err;
 536
 537	switch (type) {
 538	default:
 539	case ICMP_PARAMETERPROB:
 540		return 0;
 541
 542	case ICMP_DEST_UNREACH:
 543		switch (code) {
 544		case ICMP_SR_FAILED:
 545			/* Impossible event. */
 546			return 0;
 547		default:
 548			/* All others are translated to HOST_UNREACH.
 549			   rfc2003 contains "deep thoughts" about NET_UNREACH,
 550			   I believe they are just ether pollution. --ANK
 551			 */
 552			break;
 553		}
 554		break;
 555	case ICMP_TIME_EXCEEDED:
 556		if (code != ICMP_EXC_TTL)
 557			return 0;
 558		data_len = icmp_hdr(skb)->un.reserved[1] * 4; /* RFC 4884 4.1 */
 559		break;
 560	case ICMP_REDIRECT:
 561		break;
 562	}
 563
 564	err = -ENOENT;
 565
 566	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 567	t = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 568				iph->daddr, iph->saddr, sifindex);
 569	if (!t)
 570		goto out;
 571
 572	if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED) {
 573		ipv4_update_pmtu(skb, dev_net(skb->dev), info,
 574				 t->parms.link, iph->protocol);
 575		err = 0;
 576		goto out;
 577	}
 578	if (type == ICMP_REDIRECT) {
 579		ipv4_redirect(skb, dev_net(skb->dev), t->parms.link,
 580			      iph->protocol);
 581		err = 0;
 582		goto out;
 583	}
 584
 585	err = 0;
 586	if (__in6_dev_get(skb->dev) &&
 587	    !ip6_err_gen_icmpv6_unreach(skb, iph->ihl * 4, type, data_len))
 588		goto out;
 589
 590	if (t->parms.iph.daddr == 0)
 591		goto out;
 592
 593	if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 594		goto out;
 595
 596	if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
 597		t->err_count++;
 598	else
 599		t->err_count = 1;
 600	t->err_time = jiffies;
 601out:
 602	return err;
 603}
 604
 605static inline bool is_spoofed_6rd(struct ip_tunnel *tunnel, const __be32 v4addr,
 606				  const struct in6_addr *v6addr)
 607{
 608	__be32 v4embed = 0;
 609	if (check_6rd(tunnel, v6addr, &v4embed) && v4addr != v4embed)
 610		return true;
 611	return false;
 612}
 613
 614/* Checks if an address matches an address on the tunnel interface.
 615 * Used to detect the NAT of proto 41 packets and let them pass spoofing test.
 616 * Long story:
 617 * This function is called after we considered the packet as spoofed
 618 * in is_spoofed_6rd.
 619 * We may have a router that is doing NAT for proto 41 packets
 620 * for an internal station. Destination a.a.a.a/PREFIX:bbbb:bbbb
 621 * will be translated to n.n.n.n/PREFIX:bbbb:bbbb. And is_spoofed_6rd
 622 * function will return true, dropping the packet.
 623 * But, we can still check if is spoofed against the IP
 624 * addresses associated with the interface.
 625 */
 626static bool only_dnatted(const struct ip_tunnel *tunnel,
 627	const struct in6_addr *v6dst)
 628{
 629	int prefix_len;
 630
 631#ifdef CONFIG_IPV6_SIT_6RD
 632	prefix_len = tunnel->ip6rd.prefixlen + 32
 633		- tunnel->ip6rd.relay_prefixlen;
 634#else
 635	prefix_len = 48;
 636#endif
 637	return ipv6_chk_custom_prefix(v6dst, prefix_len, tunnel->dev);
 638}
 639
 640/* Returns true if a packet is spoofed */
 641static bool packet_is_spoofed(struct sk_buff *skb,
 642			      const struct iphdr *iph,
 643			      struct ip_tunnel *tunnel)
 644{
 645	const struct ipv6hdr *ipv6h;
 646
 647	if (tunnel->dev->priv_flags & IFF_ISATAP) {
 648		if (!isatap_chksrc(skb, iph, tunnel))
 649			return true;
 650
 651		return false;
 652	}
 653
 654	if (tunnel->dev->flags & IFF_POINTOPOINT)
 655		return false;
 656
 657	ipv6h = ipv6_hdr(skb);
 658
 659	if (unlikely(is_spoofed_6rd(tunnel, iph->saddr, &ipv6h->saddr))) {
 660		net_warn_ratelimited("Src spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 661				     &iph->saddr, &ipv6h->saddr,
 662				     &iph->daddr, &ipv6h->daddr);
 663		return true;
 664	}
 665
 666	if (likely(!is_spoofed_6rd(tunnel, iph->daddr, &ipv6h->daddr)))
 667		return false;
 668
 669	if (only_dnatted(tunnel, &ipv6h->daddr))
 670		return false;
 671
 672	net_warn_ratelimited("Dst spoofed %pI4/%pI6c -> %pI4/%pI6c\n",
 673			     &iph->saddr, &ipv6h->saddr,
 674			     &iph->daddr, &ipv6h->daddr);
 675	return true;
 676}
 677
 678static int ipip6_rcv(struct sk_buff *skb)
 679{
 680	const struct iphdr *iph = ip_hdr(skb);
 681	struct ip_tunnel *tunnel;
 682	int sifindex;
 683	int err;
 684
 685	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 686	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 687				     iph->saddr, iph->daddr, sifindex);
 688	if (tunnel) {
 689		if (tunnel->parms.iph.protocol != IPPROTO_IPV6 &&
 690		    tunnel->parms.iph.protocol != 0)
 691			goto out;
 692
 693		skb->mac_header = skb->network_header;
 694		skb_reset_network_header(skb);
 695		IPCB(skb)->flags = 0;
 696		skb->dev = tunnel->dev;
 697
 698		if (packet_is_spoofed(skb, iph, tunnel)) {
 699			DEV_STATS_INC(tunnel->dev, rx_errors);
 700			goto out;
 701		}
 702
 703		if (iptunnel_pull_header(skb, 0, htons(ETH_P_IPV6),
 704		    !net_eq(tunnel->net, dev_net(tunnel->dev))))
 705			goto out;
 706
 707		/* skb can be uncloned in iptunnel_pull_header, so
 708		 * old iph is no longer valid
 709		 */
 710		iph = (const struct iphdr *)skb_mac_header(skb);
 711		skb_reset_mac_header(skb);
 712
 713		err = IP_ECN_decapsulate(iph, skb);
 714		if (unlikely(err)) {
 715			if (log_ecn_error)
 716				net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 717						     &iph->saddr, iph->tos);
 718			if (err > 1) {
 719				DEV_STATS_INC(tunnel->dev, rx_frame_errors);
 720				DEV_STATS_INC(tunnel->dev, rx_errors);
 721				goto out;
 722			}
 723		}
 724
 725		dev_sw_netstats_rx_add(tunnel->dev, skb->len);
 726
 727		netif_rx(skb);
 728
 729		return 0;
 730	}
 731
 732	/* no tunnel matched,  let upstream know, ipsec may handle it */
 733	return 1;
 734out:
 735	kfree_skb(skb);
 736	return 0;
 737}
 738
 739static const struct tnl_ptk_info ipip_tpi = {
 740	/* no tunnel info required for ipip. */
 741	.proto = htons(ETH_P_IP),
 742};
 743
 744#if IS_ENABLED(CONFIG_MPLS)
 745static const struct tnl_ptk_info mplsip_tpi = {
 746	/* no tunnel info required for mplsip. */
 747	.proto = htons(ETH_P_MPLS_UC),
 748};
 749#endif
 750
 751static int sit_tunnel_rcv(struct sk_buff *skb, u8 ipproto)
 752{
 753	const struct iphdr *iph;
 754	struct ip_tunnel *tunnel;
 755	int sifindex;
 756
 757	sifindex = netif_is_l3_master(skb->dev) ? IPCB(skb)->iif : 0;
 758
 759	iph = ip_hdr(skb);
 760	tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
 761				     iph->saddr, iph->daddr, sifindex);
 762	if (tunnel) {
 763		const struct tnl_ptk_info *tpi;
 764
 765		if (tunnel->parms.iph.protocol != ipproto &&
 766		    tunnel->parms.iph.protocol != 0)
 767			goto drop;
 768
 769		if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
 770			goto drop;
 771#if IS_ENABLED(CONFIG_MPLS)
 772		if (ipproto == IPPROTO_MPLS)
 773			tpi = &mplsip_tpi;
 774		else
 775#endif
 776			tpi = &ipip_tpi;
 777		if (iptunnel_pull_header(skb, 0, tpi->proto, false))
 778			goto drop;
 779		skb_reset_mac_header(skb);
 780
 781		return ip_tunnel_rcv(tunnel, skb, tpi, NULL, log_ecn_error);
 782	}
 783
 784	return 1;
 785
 786drop:
 787	kfree_skb(skb);
 788	return 0;
 789}
 790
 791static int ipip_rcv(struct sk_buff *skb)
 792{
 793	return sit_tunnel_rcv(skb, IPPROTO_IPIP);
 794}
 795
 796#if IS_ENABLED(CONFIG_MPLS)
 797static int mplsip_rcv(struct sk_buff *skb)
 798{
 799	return sit_tunnel_rcv(skb, IPPROTO_MPLS);
 800}
 801#endif
 802
 803/*
 804 * If the IPv6 address comes from 6rd / 6to4 (RFC 3056) addr space this function
 805 * stores the embedded IPv4 address in v4dst and returns true.
 806 */
 807static bool check_6rd(struct ip_tunnel *tunnel, const struct in6_addr *v6dst,
 808		      __be32 *v4dst)
 809{
 810#ifdef CONFIG_IPV6_SIT_6RD
 811	if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
 812			      tunnel->ip6rd.prefixlen)) {
 813		unsigned int pbw0, pbi0;
 814		int pbi1;
 815		u32 d;
 816
 817		pbw0 = tunnel->ip6rd.prefixlen >> 5;
 818		pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
 819
 820		d = tunnel->ip6rd.relay_prefixlen < 32 ?
 821			(ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
 822		    tunnel->ip6rd.relay_prefixlen : 0;
 823
 824		pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
 825		if (pbi1 > 0)
 826			d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
 827			     (32 - pbi1);
 828
 829		*v4dst = tunnel->ip6rd.relay_prefix | htonl(d);
 830		return true;
 831	}
 832#else
 833	if (v6dst->s6_addr16[0] == htons(0x2002)) {
 834		/* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 835		memcpy(v4dst, &v6dst->s6_addr16[1], 4);
 836		return true;
 837	}
 838#endif
 839	return false;
 840}
 841
 842static inline __be32 try_6rd(struct ip_tunnel *tunnel,
 843			     const struct in6_addr *v6dst)
 844{
 845	__be32 dst = 0;
 846	check_6rd(tunnel, v6dst, &dst);
 847	return dst;
 848}
 849
 850/*
 851 *	This function assumes it is being called from dev_queue_xmit()
 852 *	and that skb is filled properly by that function.
 853 */
 854
 855static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 856				     struct net_device *dev)
 857{
 858	struct ip_tunnel *tunnel = netdev_priv(dev);
 859	const struct iphdr  *tiph = &tunnel->parms.iph;
 860	const struct ipv6hdr *iph6 = ipv6_hdr(skb);
 861	u8     tos = tunnel->parms.iph.tos;
 862	__be16 df = tiph->frag_off;
 863	struct rtable *rt;		/* Route to the other host */
 864	struct net_device *tdev;	/* Device to other host */
 865	unsigned int max_headroom;	/* The extra header space needed */
 866	__be32 dst = tiph->daddr;
 867	struct flowi4 fl4;
 868	int    mtu;
 869	const struct in6_addr *addr6;
 870	int addr_type;
 871	u8 ttl;
 872	u8 protocol = IPPROTO_IPV6;
 873	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 874
 875	if (tos == 1)
 876		tos = ipv6_get_dsfield(iph6);
 877
 878	/* ISATAP (RFC4214) - must come before 6to4 */
 879	if (dev->priv_flags & IFF_ISATAP) {
 880		struct neighbour *neigh = NULL;
 881		bool do_tx_error = false;
 882
 883		if (skb_dst(skb))
 884			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 885
 886		if (!neigh) {
 887			net_dbg_ratelimited("nexthop == NULL\n");
 888			goto tx_error;
 889		}
 890
 891		addr6 = (const struct in6_addr *)&neigh->primary_key;
 892		addr_type = ipv6_addr_type(addr6);
 893
 894		if ((addr_type & IPV6_ADDR_UNICAST) &&
 895		     ipv6_addr_is_isatap(addr6))
 896			dst = addr6->s6_addr32[3];
 897		else
 898			do_tx_error = true;
 899
 900		neigh_release(neigh);
 901		if (do_tx_error)
 902			goto tx_error;
 903	}
 904
 905	if (!dst)
 906		dst = try_6rd(tunnel, &iph6->daddr);
 907
 908	if (!dst) {
 909		struct neighbour *neigh = NULL;
 910		bool do_tx_error = false;
 911
 912		if (skb_dst(skb))
 913			neigh = dst_neigh_lookup(skb_dst(skb), &iph6->daddr);
 914
 915		if (!neigh) {
 916			net_dbg_ratelimited("nexthop == NULL\n");
 917			goto tx_error;
 918		}
 919
 920		addr6 = (const struct in6_addr *)&neigh->primary_key;
 921		addr_type = ipv6_addr_type(addr6);
 922
 923		if (addr_type == IPV6_ADDR_ANY) {
 924			addr6 = &ipv6_hdr(skb)->daddr;
 925			addr_type = ipv6_addr_type(addr6);
 926		}
 927
 928		if ((addr_type & IPV6_ADDR_COMPATv4) != 0)
 929			dst = addr6->s6_addr32[3];
 930		else
 931			do_tx_error = true;
 932
 933		neigh_release(neigh);
 934		if (do_tx_error)
 935			goto tx_error;
 936	}
 937
 938	flowi4_init_output(&fl4, tunnel->parms.link, tunnel->fwmark,
 939			   tos & INET_DSCP_MASK, RT_SCOPE_UNIVERSE,
 940			   IPPROTO_IPV6, 0, dst, tiph->saddr, 0, 0,
 941			   sock_net_uid(tunnel->net, NULL));
 942
 943	rt = dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr);
 944	if (!rt) {
 945		rt = ip_route_output_flow(tunnel->net, &fl4, NULL);
 946		if (IS_ERR(rt)) {
 947			DEV_STATS_INC(dev, tx_carrier_errors);
 948			goto tx_error_icmp;
 949		}
 950		dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst, fl4.saddr);
 951	}
 952
 953	if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
 954		ip_rt_put(rt);
 955		DEV_STATS_INC(dev, tx_carrier_errors);
 956		goto tx_error_icmp;
 957	}
 958	tdev = rt->dst.dev;
 959
 960	if (tdev == dev) {
 961		ip_rt_put(rt);
 962		DEV_STATS_INC(dev, collisions);
 963		goto tx_error;
 964	}
 965
 966	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4)) {
 967		ip_rt_put(rt);
 968		goto tx_error;
 969	}
 970
 971	if (df) {
 972		mtu = dst_mtu(&rt->dst) - t_hlen;
 973
 974		if (mtu < IPV4_MIN_MTU) {
 975			DEV_STATS_INC(dev, collisions);
 976			ip_rt_put(rt);
 977			goto tx_error;
 978		}
 979
 980		if (mtu < IPV6_MIN_MTU) {
 981			mtu = IPV6_MIN_MTU;
 982			df = 0;
 983		}
 984
 985		if (tunnel->parms.iph.daddr)
 986			skb_dst_update_pmtu_no_confirm(skb, mtu);
 987
 988		if (skb->len > mtu && !skb_is_gso(skb)) {
 989			icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 990			ip_rt_put(rt);
 991			goto tx_error;
 992		}
 993	}
 994
 995	if (tunnel->err_count > 0) {
 996		if (time_before(jiffies,
 997				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 998			tunnel->err_count--;
 999			dst_link_failure(skb);
1000		} else
1001			tunnel->err_count = 0;
1002	}
1003
1004	/*
1005	 * Okay, now see if we can stuff it in the buffer as-is.
1006	 */
1007	max_headroom = LL_RESERVED_SPACE(tdev) + t_hlen;
1008
1009	if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1010	    (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1011		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
1012		if (!new_skb) {
1013			ip_rt_put(rt);
1014			DEV_STATS_INC(dev, tx_dropped);
1015			kfree_skb(skb);
1016			return NETDEV_TX_OK;
1017		}
1018		if (skb->sk)
1019			skb_set_owner_w(new_skb, skb->sk);
1020		dev_kfree_skb(skb);
1021		skb = new_skb;
1022		iph6 = ipv6_hdr(skb);
1023	}
1024	ttl = tiph->ttl;
1025	if (ttl == 0)
1026		ttl = iph6->hop_limit;
1027	tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
1028
1029	if (ip_tunnel_encap(skb, &tunnel->encap, &protocol, &fl4) < 0) {
1030		ip_rt_put(rt);
1031		goto tx_error;
1032	}
1033
1034	skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1035
1036	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
1037		      df, !net_eq(tunnel->net, dev_net(dev)));
1038	return NETDEV_TX_OK;
1039
1040tx_error_icmp:
1041	dst_link_failure(skb);
1042tx_error:
1043	kfree_skb(skb);
1044	DEV_STATS_INC(dev, tx_errors);
1045	return NETDEV_TX_OK;
1046}
1047
1048static netdev_tx_t sit_tunnel_xmit__(struct sk_buff *skb,
1049				     struct net_device *dev, u8 ipproto)
1050{
1051	struct ip_tunnel *tunnel = netdev_priv(dev);
1052	const struct iphdr  *tiph = &tunnel->parms.iph;
1053
1054	if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP4))
1055		goto tx_error;
1056
1057	skb_set_inner_ipproto(skb, ipproto);
1058
1059	ip_tunnel_xmit(skb, dev, tiph, ipproto);
1060	return NETDEV_TX_OK;
1061tx_error:
1062	kfree_skb(skb);
1063	DEV_STATS_INC(dev, tx_errors);
1064	return NETDEV_TX_OK;
1065}
1066
1067static netdev_tx_t sit_tunnel_xmit(struct sk_buff *skb,
1068				   struct net_device *dev)
1069{
1070	if (!pskb_inet_may_pull(skb))
1071		goto tx_err;
1072
1073	switch (skb->protocol) {
1074	case htons(ETH_P_IP):
1075		sit_tunnel_xmit__(skb, dev, IPPROTO_IPIP);
1076		break;
1077	case htons(ETH_P_IPV6):
1078		ipip6_tunnel_xmit(skb, dev);
1079		break;
1080#if IS_ENABLED(CONFIG_MPLS)
1081	case htons(ETH_P_MPLS_UC):
1082		sit_tunnel_xmit__(skb, dev, IPPROTO_MPLS);
1083		break;
1084#endif
1085	default:
1086		goto tx_err;
1087	}
1088
1089	return NETDEV_TX_OK;
1090
1091tx_err:
1092	DEV_STATS_INC(dev, tx_errors);
1093	kfree_skb(skb);
1094	return NETDEV_TX_OK;
1095
1096}
1097
1098static void ipip6_tunnel_bind_dev(struct net_device *dev)
1099{
1100	struct ip_tunnel *tunnel = netdev_priv(dev);
1101	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1102	struct net_device *tdev = NULL;
1103	int hlen = LL_MAX_HEADER;
1104	const struct iphdr *iph;
1105	struct flowi4 fl4;
1106
1107	iph = &tunnel->parms.iph;
1108
1109	if (iph->daddr) {
1110		struct rtable *rt = ip_route_output_ports(tunnel->net, &fl4,
1111							  NULL,
1112							  iph->daddr, iph->saddr,
1113							  0, 0,
1114							  IPPROTO_IPV6,
1115							  iph->tos & INET_DSCP_MASK,
1116							  tunnel->parms.link);
1117
1118		if (!IS_ERR(rt)) {
1119			tdev = rt->dst.dev;
1120			ip_rt_put(rt);
1121		}
1122		dev->flags |= IFF_POINTOPOINT;
1123	}
1124
1125	if (!tdev && tunnel->parms.link)
1126		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
1127
1128	if (tdev && !netif_is_l3_master(tdev)) {
1129		int mtu;
1130
1131		mtu = tdev->mtu - t_hlen;
1132		if (mtu < IPV6_MIN_MTU)
1133			mtu = IPV6_MIN_MTU;
1134		WRITE_ONCE(dev->mtu, mtu);
1135		hlen = tdev->hard_header_len + tdev->needed_headroom;
1136	}
1137	dev->needed_headroom = t_hlen + hlen;
1138}
1139
1140static void ipip6_tunnel_update(struct ip_tunnel *t,
1141				struct ip_tunnel_parm_kern *p,
1142				__u32 fwmark)
1143{
1144	struct net *net = t->net;
1145	struct sit_net *sitn = net_generic(net, sit_net_id);
1146
1147	ipip6_tunnel_unlink(sitn, t);
1148	synchronize_net();
1149	t->parms.iph.saddr = p->iph.saddr;
1150	t->parms.iph.daddr = p->iph.daddr;
1151	__dev_addr_set(t->dev, &p->iph.saddr, 4);
1152	memcpy(t->dev->broadcast, &p->iph.daddr, 4);
1153	ipip6_tunnel_link(sitn, t);
1154	t->parms.iph.ttl = p->iph.ttl;
1155	t->parms.iph.tos = p->iph.tos;
1156	t->parms.iph.frag_off = p->iph.frag_off;
1157	if (t->parms.link != p->link || t->fwmark != fwmark) {
1158		t->parms.link = p->link;
1159		t->fwmark = fwmark;
1160		ipip6_tunnel_bind_dev(t->dev);
1161	}
1162	dst_cache_reset(&t->dst_cache);
1163	netdev_state_change(t->dev);
1164}
1165
1166#ifdef CONFIG_IPV6_SIT_6RD
1167static int ipip6_tunnel_update_6rd(struct ip_tunnel *t,
1168				   struct ip_tunnel_6rd *ip6rd)
1169{
1170	struct in6_addr prefix;
1171	__be32 relay_prefix;
1172
1173	if (ip6rd->relay_prefixlen > 32 ||
1174	    ip6rd->prefixlen + (32 - ip6rd->relay_prefixlen) > 64)
1175		return -EINVAL;
1176
1177	ipv6_addr_prefix(&prefix, &ip6rd->prefix, ip6rd->prefixlen);
1178	if (!ipv6_addr_equal(&prefix, &ip6rd->prefix))
1179		return -EINVAL;
1180	if (ip6rd->relay_prefixlen)
1181		relay_prefix = ip6rd->relay_prefix &
1182			       htonl(0xffffffffUL <<
1183				     (32 - ip6rd->relay_prefixlen));
1184	else
1185		relay_prefix = 0;
1186	if (relay_prefix != ip6rd->relay_prefix)
1187		return -EINVAL;
1188
1189	t->ip6rd.prefix = prefix;
1190	t->ip6rd.relay_prefix = relay_prefix;
1191	t->ip6rd.prefixlen = ip6rd->prefixlen;
1192	t->ip6rd.relay_prefixlen = ip6rd->relay_prefixlen;
1193	dst_cache_reset(&t->dst_cache);
1194	netdev_state_change(t->dev);
1195	return 0;
1196}
1197
1198static int
1199ipip6_tunnel_get6rd(struct net_device *dev, struct ip_tunnel_parm __user *data)
1200{
1201	struct ip_tunnel *t = netdev_priv(dev);
1202	struct ip_tunnel_parm_kern p;
1203	struct ip_tunnel_6rd ip6rd;
 
1204
1205	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1206		if (!ip_tunnel_parm_from_user(&p, data))
1207			return -EFAULT;
1208		t = ipip6_tunnel_locate(t->net, &p, 0);
1209	}
1210	if (!t)
1211		t = netdev_priv(dev);
1212
1213	ip6rd.prefix = t->ip6rd.prefix;
1214	ip6rd.relay_prefix = t->ip6rd.relay_prefix;
1215	ip6rd.prefixlen = t->ip6rd.prefixlen;
1216	ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
1217	if (copy_to_user(data, &ip6rd, sizeof(ip6rd)))
1218		return -EFAULT;
1219	return 0;
1220}
1221
1222static int
1223ipip6_tunnel_6rdctl(struct net_device *dev, struct ip_tunnel_6rd __user *data,
1224		    int cmd)
1225{
1226	struct ip_tunnel *t = netdev_priv(dev);
1227	struct ip_tunnel_6rd ip6rd;
1228	int err;
1229
1230	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1231		return -EPERM;
1232	if (copy_from_user(&ip6rd, data, sizeof(ip6rd)))
1233		return -EFAULT;
1234
1235	if (cmd != SIOCDEL6RD) {
1236		err = ipip6_tunnel_update_6rd(t, &ip6rd);
1237		if (err < 0)
1238			return err;
1239	} else
1240		ipip6_tunnel_clone_6rd(dev, dev_to_sit_net(dev));
1241	return 0;
1242}
1243
1244#endif /* CONFIG_IPV6_SIT_6RD */
1245
1246static bool ipip6_valid_ip_proto(u8 ipproto)
1247{
1248	return ipproto == IPPROTO_IPV6 ||
1249		ipproto == IPPROTO_IPIP ||
1250#if IS_ENABLED(CONFIG_MPLS)
1251		ipproto == IPPROTO_MPLS ||
1252#endif
1253		ipproto == 0;
1254}
1255
1256static int
1257__ipip6_tunnel_ioctl_validate(struct net *net, struct ip_tunnel_parm_kern *p)
1258{
1259	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1260		return -EPERM;
1261
1262	if (!ipip6_valid_ip_proto(p->iph.protocol))
1263		return -EINVAL;
1264	if (p->iph.version != 4 ||
1265	    p->iph.ihl != 5 || (p->iph.frag_off & htons(~IP_DF)))
1266		return -EINVAL;
1267
1268	if (p->iph.ttl)
1269		p->iph.frag_off |= htons(IP_DF);
1270	return 0;
1271}
1272
1273static int
1274ipip6_tunnel_get(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1275{
1276	struct ip_tunnel *t = netdev_priv(dev);
1277
1278	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev)
1279		t = ipip6_tunnel_locate(t->net, p, 0);
1280	if (!t)
1281		t = netdev_priv(dev);
1282	memcpy(p, &t->parms, sizeof(*p));
1283	return 0;
1284}
1285
1286static int
1287ipip6_tunnel_add(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1288{
1289	struct ip_tunnel *t = netdev_priv(dev);
1290	int err;
1291
1292	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1293	if (err)
1294		return err;
1295
1296	t = ipip6_tunnel_locate(t->net, p, 1);
1297	if (!t)
1298		return -ENOBUFS;
1299	return 0;
1300}
1301
1302static int
1303ipip6_tunnel_change(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1304{
1305	struct ip_tunnel *t = netdev_priv(dev);
1306	int err;
1307
1308	err = __ipip6_tunnel_ioctl_validate(t->net, p);
1309	if (err)
1310		return err;
1311
1312	t = ipip6_tunnel_locate(t->net, p, 0);
1313	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1314		if (!t)
1315			return -ENOENT;
1316	} else {
1317		if (t) {
1318			if (t->dev != dev)
1319				return -EEXIST;
1320		} else {
1321			if (((dev->flags & IFF_POINTOPOINT) && !p->iph.daddr) ||
1322			    (!(dev->flags & IFF_POINTOPOINT) && p->iph.daddr))
1323				return -EINVAL;
1324			t = netdev_priv(dev);
1325		}
1326
1327		ipip6_tunnel_update(t, p, t->fwmark);
1328	}
1329
1330	return 0;
1331}
1332
1333static int
1334ipip6_tunnel_del(struct net_device *dev, struct ip_tunnel_parm_kern *p)
1335{
1336	struct ip_tunnel *t = netdev_priv(dev);
1337
1338	if (!ns_capable(t->net->user_ns, CAP_NET_ADMIN))
1339		return -EPERM;
1340
1341	if (dev == dev_to_sit_net(dev)->fb_tunnel_dev) {
1342		t = ipip6_tunnel_locate(t->net, p, 0);
1343		if (!t)
1344			return -ENOENT;
1345		if (t == netdev_priv(dev_to_sit_net(dev)->fb_tunnel_dev))
1346			return -EPERM;
1347		dev = t->dev;
1348	}
1349	unregister_netdevice(dev);
1350	return 0;
1351}
1352
1353static int
1354ipip6_tunnel_ctl(struct net_device *dev, struct ip_tunnel_parm_kern *p,
1355		 int cmd)
1356{
1357	switch (cmd) {
1358	case SIOCGETTUNNEL:
1359		return ipip6_tunnel_get(dev, p);
1360	case SIOCADDTUNNEL:
1361		return ipip6_tunnel_add(dev, p);
1362	case SIOCCHGTUNNEL:
1363		return ipip6_tunnel_change(dev, p);
1364	case SIOCDELTUNNEL:
1365		return ipip6_tunnel_del(dev, p);
1366	default:
1367		return -EINVAL;
1368	}
1369}
1370
1371static int
1372ipip6_tunnel_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1373			    void __user *data, int cmd)
1374{
1375	switch (cmd) {
1376	case SIOCGETTUNNEL:
1377	case SIOCADDTUNNEL:
1378	case SIOCCHGTUNNEL:
1379	case SIOCDELTUNNEL:
1380		return ip_tunnel_siocdevprivate(dev, ifr, data, cmd);
1381	case SIOCGETPRL:
1382		return ipip6_tunnel_get_prl(dev, data);
1383	case SIOCADDPRL:
1384	case SIOCDELPRL:
1385	case SIOCCHGPRL:
1386		return ipip6_tunnel_prl_ctl(dev, data, cmd);
1387#ifdef CONFIG_IPV6_SIT_6RD
1388	case SIOCGET6RD:
1389		return ipip6_tunnel_get6rd(dev, data);
1390	case SIOCADD6RD:
1391	case SIOCCHG6RD:
1392	case SIOCDEL6RD:
1393		return ipip6_tunnel_6rdctl(dev, data, cmd);
1394#endif
1395	default:
1396		return -EINVAL;
1397	}
1398}
1399
1400static const struct net_device_ops ipip6_netdev_ops = {
1401	.ndo_init	= ipip6_tunnel_init,
1402	.ndo_uninit	= ipip6_tunnel_uninit,
1403	.ndo_start_xmit	= sit_tunnel_xmit,
1404	.ndo_siocdevprivate = ipip6_tunnel_siocdevprivate,
 
1405	.ndo_get_iflink = ip_tunnel_get_iflink,
1406	.ndo_tunnel_ctl = ipip6_tunnel_ctl,
1407};
1408
1409static void ipip6_dev_free(struct net_device *dev)
1410{
1411	struct ip_tunnel *tunnel = netdev_priv(dev);
1412
1413	dst_cache_destroy(&tunnel->dst_cache);
 
1414}
1415
1416#define SIT_FEATURES (NETIF_F_SG	   | \
1417		      NETIF_F_FRAGLIST	   | \
1418		      NETIF_F_HIGHDMA	   | \
1419		      NETIF_F_GSO_SOFTWARE | \
1420		      NETIF_F_HW_CSUM)
1421
1422static void ipip6_tunnel_setup(struct net_device *dev)
1423{
1424	struct ip_tunnel *tunnel = netdev_priv(dev);
1425	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
1426
1427	dev->netdev_ops		= &ipip6_netdev_ops;
1428	dev->header_ops		= &ip_tunnel_header_ops;
1429	dev->needs_free_netdev	= true;
1430	dev->priv_destructor	= ipip6_dev_free;
1431
1432	dev->type		= ARPHRD_SIT;
1433	dev->mtu		= ETH_DATA_LEN - t_hlen;
1434	dev->min_mtu		= IPV6_MIN_MTU;
1435	dev->max_mtu		= IP6_MAX_MTU - t_hlen;
1436	dev->flags		= IFF_NOARP;
1437	netif_keep_dst(dev);
1438	dev->addr_len		= 4;
1439	dev->lltx		= true;
1440	dev->features		|= SIT_FEATURES;
1441	dev->hw_features	|= SIT_FEATURES;
1442	dev->pcpu_stat_type	= NETDEV_PCPU_STAT_TSTATS;
1443
1444}
1445
1446static int ipip6_tunnel_init(struct net_device *dev)
1447{
1448	struct ip_tunnel *tunnel = netdev_priv(dev);
1449	int err;
1450
1451	tunnel->dev = dev;
1452	tunnel->net = dev_net(dev);
1453	strcpy(tunnel->parms.name, dev->name);
1454
1455	ipip6_tunnel_bind_dev(dev);
 
 
 
1456
1457	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1458	if (err)
 
 
1459		return err;
1460
1461	netdev_hold(dev, &tunnel->dev_tracker, GFP_KERNEL);
1462	netdev_lockdep_set_classes(dev);
1463	return 0;
1464}
1465
1466static void __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1467{
1468	struct ip_tunnel *tunnel = netdev_priv(dev);
1469	struct iphdr *iph = &tunnel->parms.iph;
1470	struct net *net = dev_net(dev);
1471	struct sit_net *sitn = net_generic(net, sit_net_id);
1472
1473	iph->version		= 4;
1474	iph->protocol		= IPPROTO_IPV6;
1475	iph->ihl		= 5;
1476	iph->ttl		= 64;
1477
1478	rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1479}
1480
1481static int ipip6_validate(struct nlattr *tb[], struct nlattr *data[],
1482			  struct netlink_ext_ack *extack)
1483{
1484	u8 proto;
1485
1486	if (!data || !data[IFLA_IPTUN_PROTO])
1487		return 0;
1488
1489	proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1490	if (!ipip6_valid_ip_proto(proto))
1491		return -EINVAL;
1492
1493	return 0;
1494}
1495
1496static void ipip6_netlink_parms(struct nlattr *data[],
1497				struct ip_tunnel_parm_kern *parms,
1498				__u32 *fwmark)
1499{
1500	memset(parms, 0, sizeof(*parms));
1501
1502	parms->iph.version = 4;
1503	parms->iph.protocol = IPPROTO_IPV6;
1504	parms->iph.ihl = 5;
1505	parms->iph.ttl = 64;
1506
1507	if (!data)
1508		return;
1509
1510	ip_tunnel_netlink_parms(data, parms);
1511
1512	if (data[IFLA_IPTUN_FWMARK])
1513		*fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1514}
1515
1516#ifdef CONFIG_IPV6_SIT_6RD
1517/* This function returns true when 6RD attributes are present in the nl msg */
1518static bool ipip6_netlink_6rd_parms(struct nlattr *data[],
1519				    struct ip_tunnel_6rd *ip6rd)
1520{
1521	bool ret = false;
1522	memset(ip6rd, 0, sizeof(*ip6rd));
1523
1524	if (!data)
1525		return ret;
1526
1527	if (data[IFLA_IPTUN_6RD_PREFIX]) {
1528		ret = true;
1529		ip6rd->prefix = nla_get_in6_addr(data[IFLA_IPTUN_6RD_PREFIX]);
1530	}
1531
1532	if (data[IFLA_IPTUN_6RD_RELAY_PREFIX]) {
1533		ret = true;
1534		ip6rd->relay_prefix =
1535			nla_get_be32(data[IFLA_IPTUN_6RD_RELAY_PREFIX]);
1536	}
1537
1538	if (data[IFLA_IPTUN_6RD_PREFIXLEN]) {
1539		ret = true;
1540		ip6rd->prefixlen = nla_get_u16(data[IFLA_IPTUN_6RD_PREFIXLEN]);
1541	}
1542
1543	if (data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]) {
1544		ret = true;
1545		ip6rd->relay_prefixlen =
1546			nla_get_u16(data[IFLA_IPTUN_6RD_RELAY_PREFIXLEN]);
1547	}
1548
1549	return ret;
1550}
1551#endif
1552
1553static int ipip6_newlink(struct net *src_net, struct net_device *dev,
1554			 struct nlattr *tb[], struct nlattr *data[],
1555			 struct netlink_ext_ack *extack)
1556{
1557	struct net *net = dev_net(dev);
1558	struct ip_tunnel *nt;
1559	struct ip_tunnel_encap ipencap;
1560#ifdef CONFIG_IPV6_SIT_6RD
1561	struct ip_tunnel_6rd ip6rd;
1562#endif
1563	int err;
1564
1565	nt = netdev_priv(dev);
1566
1567	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1568		err = ip_tunnel_encap_setup(nt, &ipencap);
1569		if (err < 0)
1570			return err;
1571	}
1572
1573	ipip6_netlink_parms(data, &nt->parms, &nt->fwmark);
1574
1575	if (ipip6_tunnel_locate(net, &nt->parms, 0))
1576		return -EEXIST;
1577
1578	err = ipip6_tunnel_create(dev);
1579	if (err < 0)
1580		return err;
1581
1582	if (tb[IFLA_MTU]) {
1583		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
1584
1585		if (mtu >= IPV6_MIN_MTU &&
1586		    mtu <= IP6_MAX_MTU - dev->hard_header_len)
1587			dev->mtu = mtu;
1588	}
1589
1590#ifdef CONFIG_IPV6_SIT_6RD
1591	if (ipip6_netlink_6rd_parms(data, &ip6rd)) {
1592		err = ipip6_tunnel_update_6rd(nt, &ip6rd);
1593		if (err < 0)
1594			unregister_netdevice_queue(dev, NULL);
1595	}
1596#endif
1597
1598	return err;
1599}
1600
1601static int ipip6_changelink(struct net_device *dev, struct nlattr *tb[],
1602			    struct nlattr *data[],
1603			    struct netlink_ext_ack *extack)
1604{
1605	struct ip_tunnel *t = netdev_priv(dev);
 
1606	struct ip_tunnel_encap ipencap;
1607	struct ip_tunnel_parm_kern p;
1608	struct net *net = t->net;
1609	struct sit_net *sitn = net_generic(net, sit_net_id);
1610#ifdef CONFIG_IPV6_SIT_6RD
1611	struct ip_tunnel_6rd ip6rd;
1612#endif
1613	__u32 fwmark = t->fwmark;
1614	int err;
1615
1616	if (dev == sitn->fb_tunnel_dev)
1617		return -EINVAL;
1618
1619	if (ip_tunnel_netlink_encap_parms(data, &ipencap)) {
1620		err = ip_tunnel_encap_setup(t, &ipencap);
1621		if (err < 0)
1622			return err;
1623	}
1624
1625	ipip6_netlink_parms(data, &p, &fwmark);
1626
1627	if (((dev->flags & IFF_POINTOPOINT) && !p.iph.daddr) ||
1628	    (!(dev->flags & IFF_POINTOPOINT) && p.iph.daddr))
1629		return -EINVAL;
1630
1631	t = ipip6_tunnel_locate(net, &p, 0);
1632
1633	if (t) {
1634		if (t->dev != dev)
1635			return -EEXIST;
1636	} else
1637		t = netdev_priv(dev);
1638
1639	ipip6_tunnel_update(t, &p, fwmark);
1640
1641#ifdef CONFIG_IPV6_SIT_6RD
1642	if (ipip6_netlink_6rd_parms(data, &ip6rd))
1643		return ipip6_tunnel_update_6rd(t, &ip6rd);
1644#endif
1645
1646	return 0;
1647}
1648
1649static size_t ipip6_get_size(const struct net_device *dev)
1650{
1651	return
1652		/* IFLA_IPTUN_LINK */
1653		nla_total_size(4) +
1654		/* IFLA_IPTUN_LOCAL */
1655		nla_total_size(4) +
1656		/* IFLA_IPTUN_REMOTE */
1657		nla_total_size(4) +
1658		/* IFLA_IPTUN_TTL */
1659		nla_total_size(1) +
1660		/* IFLA_IPTUN_TOS */
1661		nla_total_size(1) +
1662		/* IFLA_IPTUN_PMTUDISC */
1663		nla_total_size(1) +
1664		/* IFLA_IPTUN_FLAGS */
1665		nla_total_size(2) +
1666		/* IFLA_IPTUN_PROTO */
1667		nla_total_size(1) +
1668#ifdef CONFIG_IPV6_SIT_6RD
1669		/* IFLA_IPTUN_6RD_PREFIX */
1670		nla_total_size(sizeof(struct in6_addr)) +
1671		/* IFLA_IPTUN_6RD_RELAY_PREFIX */
1672		nla_total_size(4) +
1673		/* IFLA_IPTUN_6RD_PREFIXLEN */
1674		nla_total_size(2) +
1675		/* IFLA_IPTUN_6RD_RELAY_PREFIXLEN */
1676		nla_total_size(2) +
1677#endif
1678		/* IFLA_IPTUN_ENCAP_TYPE */
1679		nla_total_size(2) +
1680		/* IFLA_IPTUN_ENCAP_FLAGS */
1681		nla_total_size(2) +
1682		/* IFLA_IPTUN_ENCAP_SPORT */
1683		nla_total_size(2) +
1684		/* IFLA_IPTUN_ENCAP_DPORT */
1685		nla_total_size(2) +
1686		/* IFLA_IPTUN_FWMARK */
1687		nla_total_size(4) +
1688		0;
1689}
1690
1691static int ipip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1692{
1693	struct ip_tunnel *tunnel = netdev_priv(dev);
1694	struct ip_tunnel_parm_kern *parm = &tunnel->parms;
1695
1696	if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1697	    nla_put_in_addr(skb, IFLA_IPTUN_LOCAL, parm->iph.saddr) ||
1698	    nla_put_in_addr(skb, IFLA_IPTUN_REMOTE, parm->iph.daddr) ||
1699	    nla_put_u8(skb, IFLA_IPTUN_TTL, parm->iph.ttl) ||
1700	    nla_put_u8(skb, IFLA_IPTUN_TOS, parm->iph.tos) ||
1701	    nla_put_u8(skb, IFLA_IPTUN_PMTUDISC,
1702		       !!(parm->iph.frag_off & htons(IP_DF))) ||
1703	    nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->iph.protocol) ||
1704	    nla_put_be16(skb, IFLA_IPTUN_FLAGS,
1705			 ip_tunnel_flags_to_be16(parm->i_flags)) ||
1706	    nla_put_u32(skb, IFLA_IPTUN_FWMARK, tunnel->fwmark))
1707		goto nla_put_failure;
1708
1709#ifdef CONFIG_IPV6_SIT_6RD
1710	if (nla_put_in6_addr(skb, IFLA_IPTUN_6RD_PREFIX,
1711			     &tunnel->ip6rd.prefix) ||
1712	    nla_put_in_addr(skb, IFLA_IPTUN_6RD_RELAY_PREFIX,
1713			    tunnel->ip6rd.relay_prefix) ||
1714	    nla_put_u16(skb, IFLA_IPTUN_6RD_PREFIXLEN,
1715			tunnel->ip6rd.prefixlen) ||
1716	    nla_put_u16(skb, IFLA_IPTUN_6RD_RELAY_PREFIXLEN,
1717			tunnel->ip6rd.relay_prefixlen))
1718		goto nla_put_failure;
1719#endif
1720
1721	if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1722			tunnel->encap.type) ||
1723	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1724			tunnel->encap.sport) ||
1725	    nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1726			tunnel->encap.dport) ||
1727	    nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1728			tunnel->encap.flags))
1729		goto nla_put_failure;
1730
1731	return 0;
1732
1733nla_put_failure:
1734	return -EMSGSIZE;
1735}
1736
1737static const struct nla_policy ipip6_policy[IFLA_IPTUN_MAX + 1] = {
1738	[IFLA_IPTUN_LINK]		= { .type = NLA_U32 },
1739	[IFLA_IPTUN_LOCAL]		= { .type = NLA_U32 },
1740	[IFLA_IPTUN_REMOTE]		= { .type = NLA_U32 },
1741	[IFLA_IPTUN_TTL]		= { .type = NLA_U8 },
1742	[IFLA_IPTUN_TOS]		= { .type = NLA_U8 },
1743	[IFLA_IPTUN_PMTUDISC]		= { .type = NLA_U8 },
1744	[IFLA_IPTUN_FLAGS]		= { .type = NLA_U16 },
1745	[IFLA_IPTUN_PROTO]		= { .type = NLA_U8 },
1746#ifdef CONFIG_IPV6_SIT_6RD
1747	[IFLA_IPTUN_6RD_PREFIX]		= { .len = sizeof(struct in6_addr) },
1748	[IFLA_IPTUN_6RD_RELAY_PREFIX]	= { .type = NLA_U32 },
1749	[IFLA_IPTUN_6RD_PREFIXLEN]	= { .type = NLA_U16 },
1750	[IFLA_IPTUN_6RD_RELAY_PREFIXLEN] = { .type = NLA_U16 },
1751#endif
1752	[IFLA_IPTUN_ENCAP_TYPE]		= { .type = NLA_U16 },
1753	[IFLA_IPTUN_ENCAP_FLAGS]	= { .type = NLA_U16 },
1754	[IFLA_IPTUN_ENCAP_SPORT]	= { .type = NLA_U16 },
1755	[IFLA_IPTUN_ENCAP_DPORT]	= { .type = NLA_U16 },
1756	[IFLA_IPTUN_FWMARK]		= { .type = NLA_U32 },
1757};
1758
1759static void ipip6_dellink(struct net_device *dev, struct list_head *head)
1760{
1761	struct net *net = dev_net(dev);
1762	struct sit_net *sitn = net_generic(net, sit_net_id);
1763
1764	if (dev != sitn->fb_tunnel_dev)
1765		unregister_netdevice_queue(dev, head);
1766}
1767
1768static struct rtnl_link_ops sit_link_ops __read_mostly = {
1769	.kind		= "sit",
1770	.maxtype	= IFLA_IPTUN_MAX,
1771	.policy		= ipip6_policy,
1772	.priv_size	= sizeof(struct ip_tunnel),
1773	.setup		= ipip6_tunnel_setup,
1774	.validate	= ipip6_validate,
1775	.newlink	= ipip6_newlink,
1776	.changelink	= ipip6_changelink,
1777	.get_size	= ipip6_get_size,
1778	.fill_info	= ipip6_fill_info,
1779	.dellink	= ipip6_dellink,
1780	.get_link_net	= ip_tunnel_get_link_net,
1781};
1782
1783static struct xfrm_tunnel sit_handler __read_mostly = {
1784	.handler	=	ipip6_rcv,
1785	.err_handler	=	ipip6_err,
1786	.priority	=	1,
1787};
1788
1789static struct xfrm_tunnel ipip_handler __read_mostly = {
1790	.handler	=	ipip_rcv,
1791	.err_handler	=	ipip6_err,
1792	.priority	=	2,
1793};
1794
1795#if IS_ENABLED(CONFIG_MPLS)
1796static struct xfrm_tunnel mplsip_handler __read_mostly = {
1797	.handler	=	mplsip_rcv,
1798	.err_handler	=	ipip6_err,
1799	.priority	=	2,
1800};
1801#endif
1802
1803static void __net_exit sit_destroy_tunnels(struct net *net,
1804					   struct list_head *head)
1805{
1806	struct sit_net *sitn = net_generic(net, sit_net_id);
1807	struct net_device *dev, *aux;
1808	int prio;
1809
1810	for_each_netdev_safe(net, dev, aux)
1811		if (dev->rtnl_link_ops == &sit_link_ops)
1812			unregister_netdevice_queue(dev, head);
1813
1814	for (prio = 0; prio < 4; prio++) {
1815		int h;
1816		for (h = 0; h < (prio ? IP6_SIT_HASH_SIZE : 1); h++) {
1817			struct ip_tunnel *t;
1818
1819			t = rtnl_dereference(sitn->tunnels[prio][h]);
1820			while (t) {
1821				/* If dev is in the same netns, it has already
1822				 * been added to the list by the previous loop.
1823				 */
1824				if (!net_eq(dev_net(t->dev), net))
1825					unregister_netdevice_queue(t->dev,
1826								   head);
1827				t = rtnl_dereference(t->next);
1828			}
1829		}
1830	}
1831}
1832
1833static int __net_init sit_init_net(struct net *net)
1834{
1835	struct sit_net *sitn = net_generic(net, sit_net_id);
1836	struct ip_tunnel *t;
1837	int err;
1838
1839	sitn->tunnels[0] = sitn->tunnels_wc;
1840	sitn->tunnels[1] = sitn->tunnels_l;
1841	sitn->tunnels[2] = sitn->tunnels_r;
1842	sitn->tunnels[3] = sitn->tunnels_r_l;
1843
1844	if (!net_has_fallback_tunnels(net))
1845		return 0;
1846
1847	sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1848					   NET_NAME_UNKNOWN,
1849					   ipip6_tunnel_setup);
1850	if (!sitn->fb_tunnel_dev) {
1851		err = -ENOMEM;
1852		goto err_alloc_dev;
1853	}
1854	dev_net_set(sitn->fb_tunnel_dev, net);
1855	sitn->fb_tunnel_dev->rtnl_link_ops = &sit_link_ops;
1856	/* FB netdevice is special: we have one, and only one per netns.
1857	 * Allowing to move it to another netns is clearly unsafe.
1858	 */
1859	sitn->fb_tunnel_dev->netns_local = true;
1860
1861	err = register_netdev(sitn->fb_tunnel_dev);
1862	if (err)
1863		goto err_reg_dev;
1864
1865	ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1866	ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1867
1868	t = netdev_priv(sitn->fb_tunnel_dev);
1869
1870	strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1871	return 0;
1872
1873err_reg_dev:
1874	free_netdev(sitn->fb_tunnel_dev);
1875err_alloc_dev:
1876	return err;
1877}
1878
1879static void __net_exit sit_exit_batch_rtnl(struct list_head *net_list,
1880					   struct list_head *dev_to_kill)
1881{
 
1882	struct net *net;
1883
1884	ASSERT_RTNL();
1885	list_for_each_entry(net, net_list, exit_list)
1886		sit_destroy_tunnels(net, dev_to_kill);
 
 
 
1887}
1888
1889static struct pernet_operations sit_net_ops = {
1890	.init = sit_init_net,
1891	.exit_batch_rtnl = sit_exit_batch_rtnl,
1892	.id   = &sit_net_id,
1893	.size = sizeof(struct sit_net),
1894};
1895
1896static void __exit sit_cleanup(void)
1897{
1898	rtnl_link_unregister(&sit_link_ops);
1899	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1900	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1901#if IS_ENABLED(CONFIG_MPLS)
1902	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1903#endif
1904
1905	unregister_pernet_device(&sit_net_ops);
1906	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1907}
1908
1909static int __init sit_init(void)
1910{
1911	int err;
1912
1913	pr_info("IPv6, IPv4 and MPLS over IPv4 tunneling driver\n");
1914
1915	err = register_pernet_device(&sit_net_ops);
1916	if (err < 0)
1917		return err;
1918	err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1919	if (err < 0) {
1920		pr_info("%s: can't register ip6ip4\n", __func__);
1921		goto xfrm_tunnel_failed;
1922	}
1923	err = xfrm4_tunnel_register(&ipip_handler, AF_INET);
1924	if (err < 0) {
1925		pr_info("%s: can't register ip4ip4\n", __func__);
1926		goto xfrm_tunnel4_failed;
1927	}
1928#if IS_ENABLED(CONFIG_MPLS)
1929	err = xfrm4_tunnel_register(&mplsip_handler, AF_MPLS);
1930	if (err < 0) {
1931		pr_info("%s: can't register mplsip\n", __func__);
1932		goto xfrm_tunnel_mpls_failed;
1933	}
1934#endif
1935	err = rtnl_link_register(&sit_link_ops);
1936	if (err < 0)
1937		goto rtnl_link_failed;
1938
1939out:
1940	return err;
1941
1942rtnl_link_failed:
1943#if IS_ENABLED(CONFIG_MPLS)
1944	xfrm4_tunnel_deregister(&mplsip_handler, AF_MPLS);
1945xfrm_tunnel_mpls_failed:
1946#endif
1947	xfrm4_tunnel_deregister(&ipip_handler, AF_INET);
1948xfrm_tunnel4_failed:
1949	xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1950xfrm_tunnel_failed:
1951	unregister_pernet_device(&sit_net_ops);
1952	goto out;
1953}
1954
1955module_init(sit_init);
1956module_exit(sit_cleanup);
1957MODULE_DESCRIPTION("IPv6-in-IPv4 tunnel SIT driver");
1958MODULE_LICENSE("GPL");
1959MODULE_ALIAS_RTNL_LINK("sit");
1960MODULE_ALIAS_NETDEV("sit0");