Linux Audio

Check our new training course

Loading...
v4.17
 
   1/*
   2 * Copyright (c) 2013 Nicira, Inc.
   3 *
   4 * This program is free software; you can redistribute it and/or
   5 * modify it under the terms of version 2 of the GNU General Public
   6 * License as published by the Free Software Foundation.
   7 *
   8 * This program is distributed in the hope that it will be useful, but
   9 * WITHOUT ANY WARRANTY; without even the implied warranty of
  10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  11 * General Public License for more details.
  12 *
  13 * You should have received a copy of the GNU General Public License
  14 * along with this program; if not, write to the Free Software
  15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
  16 * 02110-1301, USA
  17 */
  18
  19#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  20
  21#include <linux/capability.h>
  22#include <linux/module.h>
  23#include <linux/types.h>
  24#include <linux/kernel.h>
  25#include <linux/slab.h>
  26#include <linux/uaccess.h>
  27#include <linux/skbuff.h>
  28#include <linux/netdevice.h>
  29#include <linux/in.h>
  30#include <linux/tcp.h>
  31#include <linux/udp.h>
  32#include <linux/if_arp.h>
  33#include <linux/init.h>
  34#include <linux/in6.h>
  35#include <linux/inetdevice.h>
  36#include <linux/igmp.h>
  37#include <linux/netfilter_ipv4.h>
  38#include <linux/etherdevice.h>
  39#include <linux/if_ether.h>
  40#include <linux/if_vlan.h>
  41#include <linux/rculist.h>
  42#include <linux/err.h>
  43
  44#include <net/sock.h>
  45#include <net/ip.h>
  46#include <net/icmp.h>
  47#include <net/protocol.h>
  48#include <net/ip_tunnels.h>
  49#include <net/arp.h>
  50#include <net/checksum.h>
  51#include <net/dsfield.h>
  52#include <net/inet_ecn.h>
  53#include <net/xfrm.h>
  54#include <net/net_namespace.h>
  55#include <net/netns/generic.h>
  56#include <net/rtnetlink.h>
  57#include <net/udp.h>
  58#include <net/dst_metadata.h>
  59
  60#if IS_ENABLED(CONFIG_IPV6)
  61#include <net/ipv6.h>
  62#include <net/ip6_fib.h>
  63#include <net/ip6_route.h>
  64#endif
  65
  66static unsigned int ip_tunnel_hash(__be32 key, __be32 remote)
  67{
  68	return hash_32((__force u32)key ^ (__force u32)remote,
  69			 IP_TNL_HASH_BITS);
  70}
  71
  72static bool ip_tunnel_key_match(const struct ip_tunnel_parm *p,
  73				__be16 flags, __be32 key)
  74{
  75	if (p->i_flags & TUNNEL_KEY) {
  76		if (flags & TUNNEL_KEY)
  77			return key == p->i_key;
  78		else
  79			/* key expected, none present */
  80			return false;
  81	} else
  82		return !(flags & TUNNEL_KEY);
  83}
  84
  85/* Fallback tunnel: no source, no destination, no key, no options
  86
  87   Tunnel hash table:
  88   We require exact key match i.e. if a key is present in packet
  89   it will match only tunnel with the same key; if it is not present,
  90   it will match only keyless tunnel.
  91
  92   All keysless packets, if not matched configured keyless tunnels
  93   will match fallback tunnel.
  94   Given src, dst and key, find appropriate for input tunnel.
  95*/
  96struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
  97				   int link, __be16 flags,
  98				   __be32 remote, __be32 local,
  99				   __be32 key)
 100{
 101	unsigned int hash;
 102	struct ip_tunnel *t, *cand = NULL;
 103	struct hlist_head *head;
 104
 105	hash = ip_tunnel_hash(key, remote);
 106	head = &itn->tunnels[hash];
 107
 108	hlist_for_each_entry_rcu(t, head, hash_node) {
 109		if (local != t->parms.iph.saddr ||
 110		    remote != t->parms.iph.daddr ||
 111		    !(t->dev->flags & IFF_UP))
 112			continue;
 113
 114		if (!ip_tunnel_key_match(&t->parms, flags, key))
 115			continue;
 116
 117		if (t->parms.link == link)
 118			return t;
 119		else
 120			cand = t;
 121	}
 122
 123	hlist_for_each_entry_rcu(t, head, hash_node) {
 124		if (remote != t->parms.iph.daddr ||
 125		    t->parms.iph.saddr != 0 ||
 126		    !(t->dev->flags & IFF_UP))
 127			continue;
 128
 129		if (!ip_tunnel_key_match(&t->parms, flags, key))
 130			continue;
 131
 132		if (t->parms.link == link)
 133			return t;
 134		else if (!cand)
 135			cand = t;
 136	}
 137
 138	hash = ip_tunnel_hash(key, 0);
 139	head = &itn->tunnels[hash];
 140
 141	hlist_for_each_entry_rcu(t, head, hash_node) {
 142		if ((local != t->parms.iph.saddr || t->parms.iph.daddr != 0) &&
 143		    (local != t->parms.iph.daddr || !ipv4_is_multicast(local)))
 144			continue;
 145
 146		if (!(t->dev->flags & IFF_UP))
 147			continue;
 148
 149		if (!ip_tunnel_key_match(&t->parms, flags, key))
 150			continue;
 151
 152		if (t->parms.link == link)
 153			return t;
 154		else if (!cand)
 155			cand = t;
 156	}
 157
 158	if (flags & TUNNEL_NO_KEY)
 159		goto skip_key_lookup;
 160
 161	hlist_for_each_entry_rcu(t, head, hash_node) {
 162		if (t->parms.i_key != key ||
 163		    t->parms.iph.saddr != 0 ||
 164		    t->parms.iph.daddr != 0 ||
 165		    !(t->dev->flags & IFF_UP))
 166			continue;
 167
 168		if (t->parms.link == link)
 169			return t;
 170		else if (!cand)
 171			cand = t;
 172	}
 173
 174skip_key_lookup:
 175	if (cand)
 176		return cand;
 177
 178	t = rcu_dereference(itn->collect_md_tun);
 179	if (t && t->dev->flags & IFF_UP)
 180		return t;
 181
 182	if (itn->fb_tunnel_dev && itn->fb_tunnel_dev->flags & IFF_UP)
 183		return netdev_priv(itn->fb_tunnel_dev);
 184
 185	return NULL;
 186}
 187EXPORT_SYMBOL_GPL(ip_tunnel_lookup);
 188
 189static struct hlist_head *ip_bucket(struct ip_tunnel_net *itn,
 190				    struct ip_tunnel_parm *parms)
 191{
 192	unsigned int h;
 193	__be32 remote;
 194	__be32 i_key = parms->i_key;
 195
 196	if (parms->iph.daddr && !ipv4_is_multicast(parms->iph.daddr))
 197		remote = parms->iph.daddr;
 198	else
 199		remote = 0;
 200
 201	if (!(parms->i_flags & TUNNEL_KEY) && (parms->i_flags & VTI_ISVTI))
 202		i_key = 0;
 203
 204	h = ip_tunnel_hash(i_key, remote);
 205	return &itn->tunnels[h];
 206}
 207
 208static void ip_tunnel_add(struct ip_tunnel_net *itn, struct ip_tunnel *t)
 209{
 210	struct hlist_head *head = ip_bucket(itn, &t->parms);
 211
 212	if (t->collect_md)
 213		rcu_assign_pointer(itn->collect_md_tun, t);
 214	hlist_add_head_rcu(&t->hash_node, head);
 215}
 216
 217static void ip_tunnel_del(struct ip_tunnel_net *itn, struct ip_tunnel *t)
 218{
 219	if (t->collect_md)
 220		rcu_assign_pointer(itn->collect_md_tun, NULL);
 221	hlist_del_init_rcu(&t->hash_node);
 222}
 223
 224static struct ip_tunnel *ip_tunnel_find(struct ip_tunnel_net *itn,
 225					struct ip_tunnel_parm *parms,
 226					int type)
 227{
 228	__be32 remote = parms->iph.daddr;
 229	__be32 local = parms->iph.saddr;
 230	__be32 key = parms->i_key;
 231	__be16 flags = parms->i_flags;
 232	int link = parms->link;
 233	struct ip_tunnel *t = NULL;
 234	struct hlist_head *head = ip_bucket(itn, parms);
 235
 236	hlist_for_each_entry_rcu(t, head, hash_node) {
 237		if (local == t->parms.iph.saddr &&
 238		    remote == t->parms.iph.daddr &&
 239		    link == t->parms.link &&
 240		    type == t->dev->type &&
 241		    ip_tunnel_key_match(&t->parms, flags, key))
 242			break;
 243	}
 244	return t;
 245}
 246
 247static struct net_device *__ip_tunnel_create(struct net *net,
 248					     const struct rtnl_link_ops *ops,
 249					     struct ip_tunnel_parm *parms)
 250{
 251	int err;
 252	struct ip_tunnel *tunnel;
 253	struct net_device *dev;
 254	char name[IFNAMSIZ];
 255
 256	err = -E2BIG;
 257	if (parms->name[0]) {
 258		if (!dev_valid_name(parms->name))
 259			goto failed;
 260		strlcpy(name, parms->name, IFNAMSIZ);
 261	} else {
 262		if (strlen(ops->kind) > (IFNAMSIZ - 3))
 263			goto failed;
 264		strlcpy(name, ops->kind, IFNAMSIZ);
 265		strncat(name, "%d", 2);
 266	}
 267
 268	ASSERT_RTNL();
 269	dev = alloc_netdev(ops->priv_size, name, NET_NAME_UNKNOWN, ops->setup);
 270	if (!dev) {
 271		err = -ENOMEM;
 272		goto failed;
 273	}
 274	dev_net_set(dev, net);
 275
 276	dev->rtnl_link_ops = ops;
 277
 278	tunnel = netdev_priv(dev);
 279	tunnel->parms = *parms;
 280	tunnel->net = net;
 281
 282	err = register_netdevice(dev);
 283	if (err)
 284		goto failed_free;
 285
 286	return dev;
 287
 288failed_free:
 289	free_netdev(dev);
 290failed:
 291	return ERR_PTR(err);
 292}
 293
 294static int ip_tunnel_bind_dev(struct net_device *dev)
 295{
 296	struct net_device *tdev = NULL;
 297	struct ip_tunnel *tunnel = netdev_priv(dev);
 298	const struct iphdr *iph;
 299	int hlen = LL_MAX_HEADER;
 300	int mtu = ETH_DATA_LEN;
 301	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 302
 303	iph = &tunnel->parms.iph;
 304
 305	/* Guess output device to choose reasonable mtu and needed_headroom */
 306	if (iph->daddr) {
 307		struct flowi4 fl4;
 308		struct rtable *rt;
 309
 310		ip_tunnel_init_flow(&fl4, iph->protocol, iph->daddr,
 311				    iph->saddr, tunnel->parms.o_key,
 312				    RT_TOS(iph->tos), tunnel->parms.link,
 313				    tunnel->fwmark);
 314		rt = ip_route_output_key(tunnel->net, &fl4);
 315
 316		if (!IS_ERR(rt)) {
 317			tdev = rt->dst.dev;
 318			ip_rt_put(rt);
 319		}
 320		if (dev->type != ARPHRD_ETHER)
 321			dev->flags |= IFF_POINTOPOINT;
 322
 323		dst_cache_reset(&tunnel->dst_cache);
 324	}
 325
 326	if (!tdev && tunnel->parms.link)
 327		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
 328
 329	if (tdev) {
 330		hlen = tdev->hard_header_len + tdev->needed_headroom;
 331		mtu = min(tdev->mtu, IP_MAX_MTU);
 332	}
 333
 334	dev->needed_headroom = t_hlen + hlen;
 335	mtu -= (dev->hard_header_len + t_hlen);
 336
 337	if (mtu < IPV4_MIN_MTU)
 338		mtu = IPV4_MIN_MTU;
 339
 340	return mtu;
 341}
 342
 343static struct ip_tunnel *ip_tunnel_create(struct net *net,
 344					  struct ip_tunnel_net *itn,
 345					  struct ip_tunnel_parm *parms)
 346{
 347	struct ip_tunnel *nt;
 348	struct net_device *dev;
 349	int t_hlen;
 350	int mtu;
 351	int err;
 352
 353	dev = __ip_tunnel_create(net, itn->rtnl_link_ops, parms);
 354	if (IS_ERR(dev))
 355		return ERR_CAST(dev);
 356
 357	mtu = ip_tunnel_bind_dev(dev);
 358	err = dev_set_mtu(dev, mtu);
 359	if (err)
 360		goto err_dev_set_mtu;
 361
 362	nt = netdev_priv(dev);
 363	t_hlen = nt->hlen + sizeof(struct iphdr);
 364	dev->min_mtu = ETH_MIN_MTU;
 365	dev->max_mtu = IP_MAX_MTU - dev->hard_header_len - t_hlen;
 366	ip_tunnel_add(itn, nt);
 367	return nt;
 368
 369err_dev_set_mtu:
 370	unregister_netdevice(dev);
 371	return ERR_PTR(err);
 372}
 373
 374int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
 375		  const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
 376		  bool log_ecn_error)
 377{
 378	struct pcpu_sw_netstats *tstats;
 379	const struct iphdr *iph = ip_hdr(skb);
 380	int err;
 381
 382#ifdef CONFIG_NET_IPGRE_BROADCAST
 383	if (ipv4_is_multicast(iph->daddr)) {
 384		tunnel->dev->stats.multicast++;
 385		skb->pkt_type = PACKET_BROADCAST;
 386	}
 387#endif
 388
 389	if ((!(tpi->flags&TUNNEL_CSUM) &&  (tunnel->parms.i_flags&TUNNEL_CSUM)) ||
 390	     ((tpi->flags&TUNNEL_CSUM) && !(tunnel->parms.i_flags&TUNNEL_CSUM))) {
 391		tunnel->dev->stats.rx_crc_errors++;
 392		tunnel->dev->stats.rx_errors++;
 393		goto drop;
 394	}
 395
 396	if (tunnel->parms.i_flags&TUNNEL_SEQ) {
 397		if (!(tpi->flags&TUNNEL_SEQ) ||
 398		    (tunnel->i_seqno && (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 399			tunnel->dev->stats.rx_fifo_errors++;
 400			tunnel->dev->stats.rx_errors++;
 401			goto drop;
 402		}
 403		tunnel->i_seqno = ntohl(tpi->seq) + 1;
 404	}
 405
 406	skb_reset_network_header(skb);
 407
 408	err = IP_ECN_decapsulate(iph, skb);
 409	if (unlikely(err)) {
 410		if (log_ecn_error)
 411			net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 412					&iph->saddr, iph->tos);
 413		if (err > 1) {
 414			++tunnel->dev->stats.rx_frame_errors;
 415			++tunnel->dev->stats.rx_errors;
 416			goto drop;
 417		}
 418	}
 419
 420	tstats = this_cpu_ptr(tunnel->dev->tstats);
 421	u64_stats_update_begin(&tstats->syncp);
 422	tstats->rx_packets++;
 423	tstats->rx_bytes += skb->len;
 424	u64_stats_update_end(&tstats->syncp);
 425
 426	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 427
 428	if (tunnel->dev->type == ARPHRD_ETHER) {
 429		skb->protocol = eth_type_trans(skb, tunnel->dev);
 430		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 431	} else {
 432		skb->dev = tunnel->dev;
 433	}
 434
 435	if (tun_dst)
 436		skb_dst_set(skb, (struct dst_entry *)tun_dst);
 437
 438	gro_cells_receive(&tunnel->gro_cells, skb);
 439	return 0;
 440
 441drop:
 442	if (tun_dst)
 443		dst_release((struct dst_entry *)tun_dst);
 444	kfree_skb(skb);
 445	return 0;
 446}
 447EXPORT_SYMBOL_GPL(ip_tunnel_rcv);
 448
 449int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *ops,
 450			    unsigned int num)
 451{
 452	if (num >= MAX_IPTUN_ENCAP_OPS)
 453		return -ERANGE;
 454
 455	return !cmpxchg((const struct ip_tunnel_encap_ops **)
 456			&iptun_encaps[num],
 457			NULL, ops) ? 0 : -1;
 458}
 459EXPORT_SYMBOL(ip_tunnel_encap_add_ops);
 460
 461int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *ops,
 462			    unsigned int num)
 463{
 464	int ret;
 465
 466	if (num >= MAX_IPTUN_ENCAP_OPS)
 467		return -ERANGE;
 468
 469	ret = (cmpxchg((const struct ip_tunnel_encap_ops **)
 470		       &iptun_encaps[num],
 471		       ops, NULL) == ops) ? 0 : -1;
 472
 473	synchronize_net();
 474
 475	return ret;
 476}
 477EXPORT_SYMBOL(ip_tunnel_encap_del_ops);
 478
 479int ip_tunnel_encap_setup(struct ip_tunnel *t,
 480			  struct ip_tunnel_encap *ipencap)
 481{
 482	int hlen;
 483
 484	memset(&t->encap, 0, sizeof(t->encap));
 485
 486	hlen = ip_encap_hlen(ipencap);
 487	if (hlen < 0)
 488		return hlen;
 489
 490	t->encap.type = ipencap->type;
 491	t->encap.sport = ipencap->sport;
 492	t->encap.dport = ipencap->dport;
 493	t->encap.flags = ipencap->flags;
 494
 495	t->encap_hlen = hlen;
 496	t->hlen = t->encap_hlen + t->tun_hlen;
 497
 498	return 0;
 499}
 500EXPORT_SYMBOL_GPL(ip_tunnel_encap_setup);
 501
 502static int tnl_update_pmtu(struct net_device *dev, struct sk_buff *skb,
 503			    struct rtable *rt, __be16 df,
 504			    const struct iphdr *inner_iph)
 
 505{
 506	struct ip_tunnel *tunnel = netdev_priv(dev);
 507	int pkt_size = skb->len - tunnel->hlen - dev->hard_header_len;
 508	int mtu;
 509
 
 
 
 510	if (df)
 511		mtu = dst_mtu(&rt->dst) - dev->hard_header_len
 512					- sizeof(struct iphdr) - tunnel->hlen;
 513	else
 514		mtu = skb_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
 515
 516	skb_dst_update_pmtu(skb, mtu);
 
 517
 518	if (skb->protocol == htons(ETH_P_IP)) {
 519		if (!skb_is_gso(skb) &&
 520		    (inner_iph->frag_off & htons(IP_DF)) &&
 521		    mtu < pkt_size) {
 522			memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
 523			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
 524			return -E2BIG;
 525		}
 526	}
 527#if IS_ENABLED(CONFIG_IPV6)
 528	else if (skb->protocol == htons(ETH_P_IPV6)) {
 529		struct rt6_info *rt6 = (struct rt6_info *)skb_dst(skb);
 
 
 
 
 
 530
 531		if (rt6 && mtu < dst_mtu(skb_dst(skb)) &&
 532			   mtu >= IPV6_MIN_MTU) {
 533			if ((tunnel->parms.iph.daddr &&
 534			    !ipv4_is_multicast(tunnel->parms.iph.daddr)) ||
 535			    rt6->rt6i_dst.plen == 128) {
 536				rt6->rt6i_flags |= RTF_MODIFIED;
 537				dst_metric_set(skb_dst(skb), RTAX_MTU, mtu);
 538			}
 539		}
 540
 541		if (!skb_is_gso(skb) && mtu >= IPV6_MIN_MTU &&
 542					mtu < pkt_size) {
 543			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 544			return -E2BIG;
 545		}
 546	}
 547#endif
 548	return 0;
 549}
 550
 551void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev, u8 proto)
 
 552{
 553	struct ip_tunnel *tunnel = netdev_priv(dev);
 554	u32 headroom = sizeof(struct iphdr);
 555	struct ip_tunnel_info *tun_info;
 556	const struct ip_tunnel_key *key;
 557	const struct iphdr *inner_iph;
 558	struct rtable *rt;
 559	struct flowi4 fl4;
 560	__be16 df = 0;
 561	u8 tos, ttl;
 
 562
 563	tun_info = skb_tunnel_info(skb);
 564	if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
 565		     ip_tunnel_info_af(tun_info) != AF_INET))
 566		goto tx_error;
 567	key = &tun_info->key;
 568	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 569	inner_iph = (const struct iphdr *)skb_inner_network_header(skb);
 570	tos = key->tos;
 571	if (tos == 1) {
 572		if (skb->protocol == htons(ETH_P_IP))
 573			tos = inner_iph->tos;
 574		else if (skb->protocol == htons(ETH_P_IPV6))
 575			tos = ipv6_get_dsfield((const struct ipv6hdr *)inner_iph);
 576	}
 577	ip_tunnel_init_flow(&fl4, proto, key->u.ipv4.dst, key->u.ipv4.src, 0,
 578			    RT_TOS(tos), tunnel->parms.link, tunnel->fwmark);
 
 579	if (tunnel->encap.type != TUNNEL_ENCAP_NONE)
 580		goto tx_error;
 581	rt = ip_route_output_key(tunnel->net, &fl4);
 582	if (IS_ERR(rt)) {
 583		dev->stats.tx_carrier_errors++;
 584		goto tx_error;
 
 
 
 
 
 
 
 
 
 585	}
 586	if (rt->dst.dev == dev) {
 587		ip_rt_put(rt);
 588		dev->stats.collisions++;
 589		goto tx_error;
 590	}
 
 
 
 
 
 
 
 
 
 591	tos = ip_tunnel_ecn_encap(tos, inner_iph, skb);
 592	ttl = key->ttl;
 593	if (ttl == 0) {
 594		if (skb->protocol == htons(ETH_P_IP))
 595			ttl = inner_iph->ttl;
 596		else if (skb->protocol == htons(ETH_P_IPV6))
 597			ttl = ((const struct ipv6hdr *)inner_iph)->hop_limit;
 598		else
 599			ttl = ip4_dst_hoplimit(&rt->dst);
 600	}
 601	if (key->tun_flags & TUNNEL_DONT_FRAGMENT)
 602		df = htons(IP_DF);
 603	else if (skb->protocol == htons(ETH_P_IP))
 604		df = inner_iph->frag_off & htons(IP_DF);
 
 605	headroom += LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len;
 606	if (headroom > dev->needed_headroom)
 607		dev->needed_headroom = headroom;
 608
 609	if (skb_cow_head(skb, dev->needed_headroom)) {
 610		ip_rt_put(rt);
 611		goto tx_dropped;
 612	}
 613	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, proto, tos, ttl,
 614		      df, !net_eq(tunnel->net, dev_net(dev)));
 615	return;
 616tx_error:
 617	dev->stats.tx_errors++;
 618	goto kfree;
 619tx_dropped:
 620	dev->stats.tx_dropped++;
 621kfree:
 622	kfree_skb(skb);
 623}
 624EXPORT_SYMBOL_GPL(ip_md_tunnel_xmit);
 625
 626void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
 627		    const struct iphdr *tnl_params, u8 protocol)
 628{
 629	struct ip_tunnel *tunnel = netdev_priv(dev);
 
 630	const struct iphdr *inner_iph;
 631	struct flowi4 fl4;
 632	u8     tos, ttl;
 633	__be16 df;
 634	struct rtable *rt;		/* Route to the other host */
 635	unsigned int max_headroom;	/* The extra header space needed */
 636	__be32 dst;
 
 
 
 637	bool connected;
 
 
 
 638
 639	inner_iph = (const struct iphdr *)skb_inner_network_header(skb);
 640	connected = (tunnel->parms.iph.daddr != 0);
 641
 642	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 643
 644	dst = tnl_params->daddr;
 645	if (dst == 0) {
 646		/* NBMA tunnel */
 647
 648		if (!skb_dst(skb)) {
 649			dev->stats.tx_fifo_errors++;
 650			goto tx_error;
 651		}
 652
 653		if (skb->protocol == htons(ETH_P_IP)) {
 
 
 
 
 
 
 
 
 654			rt = skb_rtable(skb);
 655			dst = rt_nexthop(rt, inner_iph->daddr);
 656		}
 657#if IS_ENABLED(CONFIG_IPV6)
 658		else if (skb->protocol == htons(ETH_P_IPV6)) {
 659			const struct in6_addr *addr6;
 660			struct neighbour *neigh;
 661			bool do_tx_error_icmp;
 662			int addr_type;
 663
 664			neigh = dst_neigh_lookup(skb_dst(skb),
 665						 &ipv6_hdr(skb)->daddr);
 666			if (!neigh)
 667				goto tx_error;
 668
 669			addr6 = (const struct in6_addr *)&neigh->primary_key;
 670			addr_type = ipv6_addr_type(addr6);
 671
 672			if (addr_type == IPV6_ADDR_ANY) {
 673				addr6 = &ipv6_hdr(skb)->daddr;
 674				addr_type = ipv6_addr_type(addr6);
 675			}
 676
 677			if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 678				do_tx_error_icmp = true;
 679			else {
 680				do_tx_error_icmp = false;
 681				dst = addr6->s6_addr32[3];
 682			}
 683			neigh_release(neigh);
 684			if (do_tx_error_icmp)
 685				goto tx_error_icmp;
 686		}
 687#endif
 688		else
 689			goto tx_error;
 690
 691		connected = false;
 
 692	}
 693
 694	tos = tnl_params->tos;
 695	if (tos & 0x1) {
 696		tos &= ~0x1;
 697		if (skb->protocol == htons(ETH_P_IP)) {
 698			tos = inner_iph->tos;
 699			connected = false;
 700		} else if (skb->protocol == htons(ETH_P_IPV6)) {
 701			tos = ipv6_get_dsfield((const struct ipv6hdr *)inner_iph);
 702			connected = false;
 703		}
 704	}
 705
 706	ip_tunnel_init_flow(&fl4, protocol, dst, tnl_params->saddr,
 707			    tunnel->parms.o_key, RT_TOS(tos), tunnel->parms.link,
 708			    tunnel->fwmark);
 709
 710	if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0)
 711		goto tx_error;
 712
 713	rt = connected ? dst_cache_get_ip4(&tunnel->dst_cache, &fl4.saddr) :
 714			 NULL;
 
 
 
 
 
 
 
 715
 716	if (!rt) {
 717		rt = ip_route_output_key(tunnel->net, &fl4);
 718
 719		if (IS_ERR(rt)) {
 720			dev->stats.tx_carrier_errors++;
 721			goto tx_error;
 722		}
 723		if (connected)
 
 
 
 724			dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst,
 725					  fl4.saddr);
 726	}
 727
 728	if (rt->dst.dev == dev) {
 729		ip_rt_put(rt);
 730		dev->stats.collisions++;
 731		goto tx_error;
 732	}
 733
 734	if (tnl_update_pmtu(dev, skb, rt, tnl_params->frag_off, inner_iph)) {
 
 735		ip_rt_put(rt);
 736		goto tx_error;
 737	}
 738
 739	if (tunnel->err_count > 0) {
 740		if (time_before(jiffies,
 741				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 742			tunnel->err_count--;
 743
 744			dst_link_failure(skb);
 745		} else
 746			tunnel->err_count = 0;
 747	}
 748
 749	tos = ip_tunnel_ecn_encap(tos, inner_iph, skb);
 750	ttl = tnl_params->ttl;
 751	if (ttl == 0) {
 752		if (skb->protocol == htons(ETH_P_IP))
 753			ttl = inner_iph->ttl;
 754#if IS_ENABLED(CONFIG_IPV6)
 755		else if (skb->protocol == htons(ETH_P_IPV6))
 756			ttl = ((const struct ipv6hdr *)inner_iph)->hop_limit;
 757#endif
 758		else
 759			ttl = ip4_dst_hoplimit(&rt->dst);
 760	}
 761
 762	df = tnl_params->frag_off;
 763	if (skb->protocol == htons(ETH_P_IP) && !tunnel->ignore_df)
 764		df |= (inner_iph->frag_off&htons(IP_DF));
 765
 766	max_headroom = LL_RESERVED_SPACE(rt->dst.dev) + sizeof(struct iphdr)
 767			+ rt->dst.header_len + ip_encap_hlen(&tunnel->encap);
 768	if (max_headroom > dev->needed_headroom)
 769		dev->needed_headroom = max_headroom;
 770
 771	if (skb_cow_head(skb, dev->needed_headroom)) {
 772		ip_rt_put(rt);
 773		dev->stats.tx_dropped++;
 774		kfree_skb(skb);
 775		return;
 776	}
 777
 778	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
 779		      df, !net_eq(tunnel->net, dev_net(dev)));
 780	return;
 781
 782#if IS_ENABLED(CONFIG_IPV6)
 783tx_error_icmp:
 784	dst_link_failure(skb);
 785#endif
 786tx_error:
 787	dev->stats.tx_errors++;
 788	kfree_skb(skb);
 789}
 790EXPORT_SYMBOL_GPL(ip_tunnel_xmit);
 791
 792static void ip_tunnel_update(struct ip_tunnel_net *itn,
 793			     struct ip_tunnel *t,
 794			     struct net_device *dev,
 795			     struct ip_tunnel_parm *p,
 796			     bool set_mtu,
 797			     __u32 fwmark)
 798{
 799	ip_tunnel_del(itn, t);
 800	t->parms.iph.saddr = p->iph.saddr;
 801	t->parms.iph.daddr = p->iph.daddr;
 802	t->parms.i_key = p->i_key;
 803	t->parms.o_key = p->o_key;
 804	if (dev->type != ARPHRD_ETHER) {
 805		memcpy(dev->dev_addr, &p->iph.saddr, 4);
 806		memcpy(dev->broadcast, &p->iph.daddr, 4);
 807	}
 808	ip_tunnel_add(itn, t);
 809
 810	t->parms.iph.ttl = p->iph.ttl;
 811	t->parms.iph.tos = p->iph.tos;
 812	t->parms.iph.frag_off = p->iph.frag_off;
 813
 814	if (t->parms.link != p->link || t->fwmark != fwmark) {
 815		int mtu;
 816
 817		t->parms.link = p->link;
 818		t->fwmark = fwmark;
 819		mtu = ip_tunnel_bind_dev(dev);
 820		if (set_mtu)
 821			dev->mtu = mtu;
 822	}
 823	dst_cache_reset(&t->dst_cache);
 824	netdev_state_change(dev);
 825}
 826
 827int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
 828{
 829	int err = 0;
 830	struct ip_tunnel *t = netdev_priv(dev);
 831	struct net *net = t->net;
 832	struct ip_tunnel_net *itn = net_generic(net, t->ip_tnl_net_id);
 833
 834	switch (cmd) {
 835	case SIOCGETTUNNEL:
 836		if (dev == itn->fb_tunnel_dev) {
 837			t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
 838			if (!t)
 839				t = netdev_priv(dev);
 840		}
 841		memcpy(p, &t->parms, sizeof(*p));
 842		break;
 843
 844	case SIOCADDTUNNEL:
 845	case SIOCCHGTUNNEL:
 846		err = -EPERM;
 847		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 848			goto done;
 849		if (p->iph.ttl)
 850			p->iph.frag_off |= htons(IP_DF);
 851		if (!(p->i_flags & VTI_ISVTI)) {
 852			if (!(p->i_flags & TUNNEL_KEY))
 853				p->i_key = 0;
 854			if (!(p->o_flags & TUNNEL_KEY))
 855				p->o_key = 0;
 856		}
 857
 858		t = ip_tunnel_find(itn, p, itn->type);
 859
 860		if (cmd == SIOCADDTUNNEL) {
 861			if (!t) {
 862				t = ip_tunnel_create(net, itn, p);
 863				err = PTR_ERR_OR_ZERO(t);
 864				break;
 865			}
 866
 867			err = -EEXIST;
 868			break;
 869		}
 870		if (dev != itn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 871			if (t) {
 872				if (t->dev != dev) {
 873					err = -EEXIST;
 874					break;
 875				}
 876			} else {
 877				unsigned int nflags = 0;
 878
 879				if (ipv4_is_multicast(p->iph.daddr))
 880					nflags = IFF_BROADCAST;
 881				else if (p->iph.daddr)
 882					nflags = IFF_POINTOPOINT;
 883
 884				if ((dev->flags^nflags)&(IFF_POINTOPOINT|IFF_BROADCAST)) {
 885					err = -EINVAL;
 886					break;
 887				}
 888
 889				t = netdev_priv(dev);
 890			}
 891		}
 892
 893		if (t) {
 894			err = 0;
 895			ip_tunnel_update(itn, t, dev, p, true, 0);
 896		} else {
 897			err = -ENOENT;
 898		}
 899		break;
 900
 901	case SIOCDELTUNNEL:
 902		err = -EPERM;
 903		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 904			goto done;
 905
 906		if (dev == itn->fb_tunnel_dev) {
 907			err = -ENOENT;
 908			t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
 909			if (!t)
 910				goto done;
 911			err = -EPERM;
 912			if (t == netdev_priv(itn->fb_tunnel_dev))
 913				goto done;
 914			dev = t->dev;
 915		}
 916		unregister_netdevice(dev);
 917		err = 0;
 918		break;
 919
 920	default:
 921		err = -EINVAL;
 922	}
 923
 924done:
 925	return err;
 926}
 927EXPORT_SYMBOL_GPL(ip_tunnel_ioctl);
 928
 929int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict)
 930{
 931	struct ip_tunnel *tunnel = netdev_priv(dev);
 932	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 933	int max_mtu = IP_MAX_MTU - dev->hard_header_len - t_hlen;
 934
 935	if (new_mtu < ETH_MIN_MTU)
 936		return -EINVAL;
 937
 938	if (new_mtu > max_mtu) {
 939		if (strict)
 940			return -EINVAL;
 941
 942		new_mtu = max_mtu;
 943	}
 944
 945	dev->mtu = new_mtu;
 946	return 0;
 947}
 948EXPORT_SYMBOL_GPL(__ip_tunnel_change_mtu);
 949
 950int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 951{
 952	return __ip_tunnel_change_mtu(dev, new_mtu, true);
 953}
 954EXPORT_SYMBOL_GPL(ip_tunnel_change_mtu);
 955
 956static void ip_tunnel_dev_free(struct net_device *dev)
 957{
 958	struct ip_tunnel *tunnel = netdev_priv(dev);
 959
 960	gro_cells_destroy(&tunnel->gro_cells);
 961	dst_cache_destroy(&tunnel->dst_cache);
 962	free_percpu(dev->tstats);
 963}
 964
 965void ip_tunnel_dellink(struct net_device *dev, struct list_head *head)
 966{
 967	struct ip_tunnel *tunnel = netdev_priv(dev);
 968	struct ip_tunnel_net *itn;
 969
 970	itn = net_generic(tunnel->net, tunnel->ip_tnl_net_id);
 971
 972	if (itn->fb_tunnel_dev != dev) {
 973		ip_tunnel_del(itn, netdev_priv(dev));
 974		unregister_netdevice_queue(dev, head);
 975	}
 976}
 977EXPORT_SYMBOL_GPL(ip_tunnel_dellink);
 978
 979struct net *ip_tunnel_get_link_net(const struct net_device *dev)
 980{
 981	struct ip_tunnel *tunnel = netdev_priv(dev);
 982
 983	return tunnel->net;
 984}
 985EXPORT_SYMBOL(ip_tunnel_get_link_net);
 986
 987int ip_tunnel_get_iflink(const struct net_device *dev)
 988{
 989	struct ip_tunnel *tunnel = netdev_priv(dev);
 990
 991	return tunnel->parms.link;
 992}
 993EXPORT_SYMBOL(ip_tunnel_get_iflink);
 994
 995int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
 996				  struct rtnl_link_ops *ops, char *devname)
 997{
 998	struct ip_tunnel_net *itn = net_generic(net, ip_tnl_net_id);
 999	struct ip_tunnel_parm parms;
1000	unsigned int i;
1001
1002	itn->rtnl_link_ops = ops;
1003	for (i = 0; i < IP_TNL_HASH_SIZE; i++)
1004		INIT_HLIST_HEAD(&itn->tunnels[i]);
1005
1006	if (!ops || !net_has_fallback_tunnels(net)) {
1007		struct ip_tunnel_net *it_init_net;
1008
1009		it_init_net = net_generic(&init_net, ip_tnl_net_id);
1010		itn->type = it_init_net->type;
1011		itn->fb_tunnel_dev = NULL;
1012		return 0;
1013	}
1014
1015	memset(&parms, 0, sizeof(parms));
1016	if (devname)
1017		strlcpy(parms.name, devname, IFNAMSIZ);
1018
1019	rtnl_lock();
1020	itn->fb_tunnel_dev = __ip_tunnel_create(net, ops, &parms);
1021	/* FB netdevice is special: we have one, and only one per netns.
1022	 * Allowing to move it to another netns is clearly unsafe.
1023	 */
1024	if (!IS_ERR(itn->fb_tunnel_dev)) {
1025		itn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1026		itn->fb_tunnel_dev->mtu = ip_tunnel_bind_dev(itn->fb_tunnel_dev);
1027		ip_tunnel_add(itn, netdev_priv(itn->fb_tunnel_dev));
1028		itn->type = itn->fb_tunnel_dev->type;
1029	}
1030	rtnl_unlock();
1031
1032	return PTR_ERR_OR_ZERO(itn->fb_tunnel_dev);
1033}
1034EXPORT_SYMBOL_GPL(ip_tunnel_init_net);
1035
1036static void ip_tunnel_destroy(struct net *net, struct ip_tunnel_net *itn,
1037			      struct list_head *head,
1038			      struct rtnl_link_ops *ops)
1039{
1040	struct net_device *dev, *aux;
1041	int h;
1042
1043	for_each_netdev_safe(net, dev, aux)
1044		if (dev->rtnl_link_ops == ops)
1045			unregister_netdevice_queue(dev, head);
1046
1047	for (h = 0; h < IP_TNL_HASH_SIZE; h++) {
1048		struct ip_tunnel *t;
1049		struct hlist_node *n;
1050		struct hlist_head *thead = &itn->tunnels[h];
1051
1052		hlist_for_each_entry_safe(t, n, thead, hash_node)
1053			/* If dev is in the same netns, it has already
1054			 * been added to the list by the previous loop.
1055			 */
1056			if (!net_eq(dev_net(t->dev), net))
1057				unregister_netdevice_queue(t->dev, head);
1058	}
1059}
1060
1061void ip_tunnel_delete_nets(struct list_head *net_list, unsigned int id,
1062			   struct rtnl_link_ops *ops)
1063{
1064	struct ip_tunnel_net *itn;
1065	struct net *net;
1066	LIST_HEAD(list);
1067
1068	rtnl_lock();
1069	list_for_each_entry(net, net_list, exit_list) {
1070		itn = net_generic(net, id);
1071		ip_tunnel_destroy(net, itn, &list, ops);
1072	}
1073	unregister_netdevice_many(&list);
1074	rtnl_unlock();
1075}
1076EXPORT_SYMBOL_GPL(ip_tunnel_delete_nets);
1077
1078int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
1079		      struct ip_tunnel_parm *p, __u32 fwmark)
1080{
1081	struct ip_tunnel *nt;
1082	struct net *net = dev_net(dev);
1083	struct ip_tunnel_net *itn;
1084	int mtu;
1085	int err;
1086
1087	nt = netdev_priv(dev);
1088	itn = net_generic(net, nt->ip_tnl_net_id);
1089
1090	if (nt->collect_md) {
1091		if (rtnl_dereference(itn->collect_md_tun))
1092			return -EEXIST;
1093	} else {
1094		if (ip_tunnel_find(itn, p, dev->type))
1095			return -EEXIST;
1096	}
1097
1098	nt->net = net;
1099	nt->parms = *p;
1100	nt->fwmark = fwmark;
1101	err = register_netdevice(dev);
1102	if (err)
1103		goto err_register_netdevice;
1104
1105	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1106		eth_hw_addr_random(dev);
1107
1108	mtu = ip_tunnel_bind_dev(dev);
1109	if (tb[IFLA_MTU]) {
1110		unsigned int max = IP_MAX_MTU - dev->hard_header_len - nt->hlen;
1111
1112		mtu = clamp(dev->mtu, (unsigned int)ETH_MIN_MTU,
1113			    (unsigned int)(max - sizeof(struct iphdr)));
1114	}
1115
1116	err = dev_set_mtu(dev, mtu);
1117	if (err)
1118		goto err_dev_set_mtu;
1119
1120	ip_tunnel_add(itn, nt);
1121	return 0;
1122
1123err_dev_set_mtu:
1124	unregister_netdevice(dev);
1125err_register_netdevice:
1126	return err;
1127}
1128EXPORT_SYMBOL_GPL(ip_tunnel_newlink);
1129
1130int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
1131			 struct ip_tunnel_parm *p, __u32 fwmark)
1132{
1133	struct ip_tunnel *t;
1134	struct ip_tunnel *tunnel = netdev_priv(dev);
1135	struct net *net = tunnel->net;
1136	struct ip_tunnel_net *itn = net_generic(net, tunnel->ip_tnl_net_id);
1137
1138	if (dev == itn->fb_tunnel_dev)
1139		return -EINVAL;
1140
1141	t = ip_tunnel_find(itn, p, dev->type);
1142
1143	if (t) {
1144		if (t->dev != dev)
1145			return -EEXIST;
1146	} else {
1147		t = tunnel;
1148
1149		if (dev->type != ARPHRD_ETHER) {
1150			unsigned int nflags = 0;
1151
1152			if (ipv4_is_multicast(p->iph.daddr))
1153				nflags = IFF_BROADCAST;
1154			else if (p->iph.daddr)
1155				nflags = IFF_POINTOPOINT;
1156
1157			if ((dev->flags ^ nflags) &
1158			    (IFF_POINTOPOINT | IFF_BROADCAST))
1159				return -EINVAL;
1160		}
1161	}
1162
1163	ip_tunnel_update(itn, t, dev, p, !tb[IFLA_MTU], fwmark);
1164	return 0;
1165}
1166EXPORT_SYMBOL_GPL(ip_tunnel_changelink);
1167
1168int ip_tunnel_init(struct net_device *dev)
1169{
1170	struct ip_tunnel *tunnel = netdev_priv(dev);
1171	struct iphdr *iph = &tunnel->parms.iph;
1172	int err;
1173
1174	dev->needs_free_netdev = true;
1175	dev->priv_destructor = ip_tunnel_dev_free;
1176	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1177	if (!dev->tstats)
1178		return -ENOMEM;
1179
1180	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1181	if (err) {
1182		free_percpu(dev->tstats);
1183		return err;
1184	}
1185
1186	err = gro_cells_init(&tunnel->gro_cells, dev);
1187	if (err) {
1188		dst_cache_destroy(&tunnel->dst_cache);
1189		free_percpu(dev->tstats);
1190		return err;
1191	}
1192
1193	tunnel->dev = dev;
1194	tunnel->net = dev_net(dev);
1195	strcpy(tunnel->parms.name, dev->name);
1196	iph->version		= 4;
1197	iph->ihl		= 5;
1198
1199	if (tunnel->collect_md) {
1200		dev->features |= NETIF_F_NETNS_LOCAL;
1201		netif_keep_dst(dev);
1202	}
1203	return 0;
1204}
1205EXPORT_SYMBOL_GPL(ip_tunnel_init);
1206
1207void ip_tunnel_uninit(struct net_device *dev)
1208{
1209	struct ip_tunnel *tunnel = netdev_priv(dev);
1210	struct net *net = tunnel->net;
1211	struct ip_tunnel_net *itn;
1212
1213	itn = net_generic(net, tunnel->ip_tnl_net_id);
1214	/* fb_tunnel_dev will be unregisted in net-exit call. */
1215	if (itn->fb_tunnel_dev != dev)
1216		ip_tunnel_del(itn, netdev_priv(dev));
1217
1218	dst_cache_reset(&tunnel->dst_cache);
1219}
1220EXPORT_SYMBOL_GPL(ip_tunnel_uninit);
1221
1222/* Do least required initialization, rest of init is done in tunnel_init call */
1223void ip_tunnel_setup(struct net_device *dev, unsigned int net_id)
1224{
1225	struct ip_tunnel *tunnel = netdev_priv(dev);
1226	tunnel->ip_tnl_net_id = net_id;
1227}
1228EXPORT_SYMBOL_GPL(ip_tunnel_setup);
1229
1230MODULE_LICENSE("GPL");
v5.4
   1// SPDX-License-Identifier: GPL-2.0-only
   2/*
   3 * Copyright (c) 2013 Nicira, Inc.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
   4 */
   5
   6#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
   7
   8#include <linux/capability.h>
   9#include <linux/module.h>
  10#include <linux/types.h>
  11#include <linux/kernel.h>
  12#include <linux/slab.h>
  13#include <linux/uaccess.h>
  14#include <linux/skbuff.h>
  15#include <linux/netdevice.h>
  16#include <linux/in.h>
  17#include <linux/tcp.h>
  18#include <linux/udp.h>
  19#include <linux/if_arp.h>
  20#include <linux/init.h>
  21#include <linux/in6.h>
  22#include <linux/inetdevice.h>
  23#include <linux/igmp.h>
  24#include <linux/netfilter_ipv4.h>
  25#include <linux/etherdevice.h>
  26#include <linux/if_ether.h>
  27#include <linux/if_vlan.h>
  28#include <linux/rculist.h>
  29#include <linux/err.h>
  30
  31#include <net/sock.h>
  32#include <net/ip.h>
  33#include <net/icmp.h>
  34#include <net/protocol.h>
  35#include <net/ip_tunnels.h>
  36#include <net/arp.h>
  37#include <net/checksum.h>
  38#include <net/dsfield.h>
  39#include <net/inet_ecn.h>
  40#include <net/xfrm.h>
  41#include <net/net_namespace.h>
  42#include <net/netns/generic.h>
  43#include <net/rtnetlink.h>
  44#include <net/udp.h>
  45#include <net/dst_metadata.h>
  46
  47#if IS_ENABLED(CONFIG_IPV6)
  48#include <net/ipv6.h>
  49#include <net/ip6_fib.h>
  50#include <net/ip6_route.h>
  51#endif
  52
  53static unsigned int ip_tunnel_hash(__be32 key, __be32 remote)
  54{
  55	return hash_32((__force u32)key ^ (__force u32)remote,
  56			 IP_TNL_HASH_BITS);
  57}
  58
  59static bool ip_tunnel_key_match(const struct ip_tunnel_parm *p,
  60				__be16 flags, __be32 key)
  61{
  62	if (p->i_flags & TUNNEL_KEY) {
  63		if (flags & TUNNEL_KEY)
  64			return key == p->i_key;
  65		else
  66			/* key expected, none present */
  67			return false;
  68	} else
  69		return !(flags & TUNNEL_KEY);
  70}
  71
  72/* Fallback tunnel: no source, no destination, no key, no options
  73
  74   Tunnel hash table:
  75   We require exact key match i.e. if a key is present in packet
  76   it will match only tunnel with the same key; if it is not present,
  77   it will match only keyless tunnel.
  78
  79   All keysless packets, if not matched configured keyless tunnels
  80   will match fallback tunnel.
  81   Given src, dst and key, find appropriate for input tunnel.
  82*/
  83struct ip_tunnel *ip_tunnel_lookup(struct ip_tunnel_net *itn,
  84				   int link, __be16 flags,
  85				   __be32 remote, __be32 local,
  86				   __be32 key)
  87{
  88	unsigned int hash;
  89	struct ip_tunnel *t, *cand = NULL;
  90	struct hlist_head *head;
  91
  92	hash = ip_tunnel_hash(key, remote);
  93	head = &itn->tunnels[hash];
  94
  95	hlist_for_each_entry_rcu(t, head, hash_node) {
  96		if (local != t->parms.iph.saddr ||
  97		    remote != t->parms.iph.daddr ||
  98		    !(t->dev->flags & IFF_UP))
  99			continue;
 100
 101		if (!ip_tunnel_key_match(&t->parms, flags, key))
 102			continue;
 103
 104		if (t->parms.link == link)
 105			return t;
 106		else
 107			cand = t;
 108	}
 109
 110	hlist_for_each_entry_rcu(t, head, hash_node) {
 111		if (remote != t->parms.iph.daddr ||
 112		    t->parms.iph.saddr != 0 ||
 113		    !(t->dev->flags & IFF_UP))
 114			continue;
 115
 116		if (!ip_tunnel_key_match(&t->parms, flags, key))
 117			continue;
 118
 119		if (t->parms.link == link)
 120			return t;
 121		else if (!cand)
 122			cand = t;
 123	}
 124
 125	hash = ip_tunnel_hash(key, 0);
 126	head = &itn->tunnels[hash];
 127
 128	hlist_for_each_entry_rcu(t, head, hash_node) {
 129		if ((local != t->parms.iph.saddr || t->parms.iph.daddr != 0) &&
 130		    (local != t->parms.iph.daddr || !ipv4_is_multicast(local)))
 131			continue;
 132
 133		if (!(t->dev->flags & IFF_UP))
 134			continue;
 135
 136		if (!ip_tunnel_key_match(&t->parms, flags, key))
 137			continue;
 138
 139		if (t->parms.link == link)
 140			return t;
 141		else if (!cand)
 142			cand = t;
 143	}
 144
 145	if (flags & TUNNEL_NO_KEY)
 146		goto skip_key_lookup;
 147
 148	hlist_for_each_entry_rcu(t, head, hash_node) {
 149		if (t->parms.i_key != key ||
 150		    t->parms.iph.saddr != 0 ||
 151		    t->parms.iph.daddr != 0 ||
 152		    !(t->dev->flags & IFF_UP))
 153			continue;
 154
 155		if (t->parms.link == link)
 156			return t;
 157		else if (!cand)
 158			cand = t;
 159	}
 160
 161skip_key_lookup:
 162	if (cand)
 163		return cand;
 164
 165	t = rcu_dereference(itn->collect_md_tun);
 166	if (t && t->dev->flags & IFF_UP)
 167		return t;
 168
 169	if (itn->fb_tunnel_dev && itn->fb_tunnel_dev->flags & IFF_UP)
 170		return netdev_priv(itn->fb_tunnel_dev);
 171
 172	return NULL;
 173}
 174EXPORT_SYMBOL_GPL(ip_tunnel_lookup);
 175
 176static struct hlist_head *ip_bucket(struct ip_tunnel_net *itn,
 177				    struct ip_tunnel_parm *parms)
 178{
 179	unsigned int h;
 180	__be32 remote;
 181	__be32 i_key = parms->i_key;
 182
 183	if (parms->iph.daddr && !ipv4_is_multicast(parms->iph.daddr))
 184		remote = parms->iph.daddr;
 185	else
 186		remote = 0;
 187
 188	if (!(parms->i_flags & TUNNEL_KEY) && (parms->i_flags & VTI_ISVTI))
 189		i_key = 0;
 190
 191	h = ip_tunnel_hash(i_key, remote);
 192	return &itn->tunnels[h];
 193}
 194
 195static void ip_tunnel_add(struct ip_tunnel_net *itn, struct ip_tunnel *t)
 196{
 197	struct hlist_head *head = ip_bucket(itn, &t->parms);
 198
 199	if (t->collect_md)
 200		rcu_assign_pointer(itn->collect_md_tun, t);
 201	hlist_add_head_rcu(&t->hash_node, head);
 202}
 203
 204static void ip_tunnel_del(struct ip_tunnel_net *itn, struct ip_tunnel *t)
 205{
 206	if (t->collect_md)
 207		rcu_assign_pointer(itn->collect_md_tun, NULL);
 208	hlist_del_init_rcu(&t->hash_node);
 209}
 210
 211static struct ip_tunnel *ip_tunnel_find(struct ip_tunnel_net *itn,
 212					struct ip_tunnel_parm *parms,
 213					int type)
 214{
 215	__be32 remote = parms->iph.daddr;
 216	__be32 local = parms->iph.saddr;
 217	__be32 key = parms->i_key;
 218	__be16 flags = parms->i_flags;
 219	int link = parms->link;
 220	struct ip_tunnel *t = NULL;
 221	struct hlist_head *head = ip_bucket(itn, parms);
 222
 223	hlist_for_each_entry_rcu(t, head, hash_node) {
 224		if (local == t->parms.iph.saddr &&
 225		    remote == t->parms.iph.daddr &&
 226		    link == t->parms.link &&
 227		    type == t->dev->type &&
 228		    ip_tunnel_key_match(&t->parms, flags, key))
 229			break;
 230	}
 231	return t;
 232}
 233
 234static struct net_device *__ip_tunnel_create(struct net *net,
 235					     const struct rtnl_link_ops *ops,
 236					     struct ip_tunnel_parm *parms)
 237{
 238	int err;
 239	struct ip_tunnel *tunnel;
 240	struct net_device *dev;
 241	char name[IFNAMSIZ];
 242
 243	err = -E2BIG;
 244	if (parms->name[0]) {
 245		if (!dev_valid_name(parms->name))
 246			goto failed;
 247		strlcpy(name, parms->name, IFNAMSIZ);
 248	} else {
 249		if (strlen(ops->kind) > (IFNAMSIZ - 3))
 250			goto failed;
 251		strcpy(name, ops->kind);
 252		strcat(name, "%d");
 253	}
 254
 255	ASSERT_RTNL();
 256	dev = alloc_netdev(ops->priv_size, name, NET_NAME_UNKNOWN, ops->setup);
 257	if (!dev) {
 258		err = -ENOMEM;
 259		goto failed;
 260	}
 261	dev_net_set(dev, net);
 262
 263	dev->rtnl_link_ops = ops;
 264
 265	tunnel = netdev_priv(dev);
 266	tunnel->parms = *parms;
 267	tunnel->net = net;
 268
 269	err = register_netdevice(dev);
 270	if (err)
 271		goto failed_free;
 272
 273	return dev;
 274
 275failed_free:
 276	free_netdev(dev);
 277failed:
 278	return ERR_PTR(err);
 279}
 280
 281static int ip_tunnel_bind_dev(struct net_device *dev)
 282{
 283	struct net_device *tdev = NULL;
 284	struct ip_tunnel *tunnel = netdev_priv(dev);
 285	const struct iphdr *iph;
 286	int hlen = LL_MAX_HEADER;
 287	int mtu = ETH_DATA_LEN;
 288	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 289
 290	iph = &tunnel->parms.iph;
 291
 292	/* Guess output device to choose reasonable mtu and needed_headroom */
 293	if (iph->daddr) {
 294		struct flowi4 fl4;
 295		struct rtable *rt;
 296
 297		ip_tunnel_init_flow(&fl4, iph->protocol, iph->daddr,
 298				    iph->saddr, tunnel->parms.o_key,
 299				    RT_TOS(iph->tos), tunnel->parms.link,
 300				    tunnel->fwmark, 0);
 301		rt = ip_route_output_key(tunnel->net, &fl4);
 302
 303		if (!IS_ERR(rt)) {
 304			tdev = rt->dst.dev;
 305			ip_rt_put(rt);
 306		}
 307		if (dev->type != ARPHRD_ETHER)
 308			dev->flags |= IFF_POINTOPOINT;
 309
 310		dst_cache_reset(&tunnel->dst_cache);
 311	}
 312
 313	if (!tdev && tunnel->parms.link)
 314		tdev = __dev_get_by_index(tunnel->net, tunnel->parms.link);
 315
 316	if (tdev) {
 317		hlen = tdev->hard_header_len + tdev->needed_headroom;
 318		mtu = min(tdev->mtu, IP_MAX_MTU);
 319	}
 320
 321	dev->needed_headroom = t_hlen + hlen;
 322	mtu -= (dev->hard_header_len + t_hlen);
 323
 324	if (mtu < IPV4_MIN_MTU)
 325		mtu = IPV4_MIN_MTU;
 326
 327	return mtu;
 328}
 329
 330static struct ip_tunnel *ip_tunnel_create(struct net *net,
 331					  struct ip_tunnel_net *itn,
 332					  struct ip_tunnel_parm *parms)
 333{
 334	struct ip_tunnel *nt;
 335	struct net_device *dev;
 336	int t_hlen;
 337	int mtu;
 338	int err;
 339
 340	dev = __ip_tunnel_create(net, itn->rtnl_link_ops, parms);
 341	if (IS_ERR(dev))
 342		return ERR_CAST(dev);
 343
 344	mtu = ip_tunnel_bind_dev(dev);
 345	err = dev_set_mtu(dev, mtu);
 346	if (err)
 347		goto err_dev_set_mtu;
 348
 349	nt = netdev_priv(dev);
 350	t_hlen = nt->hlen + sizeof(struct iphdr);
 351	dev->min_mtu = ETH_MIN_MTU;
 352	dev->max_mtu = IP_MAX_MTU - dev->hard_header_len - t_hlen;
 353	ip_tunnel_add(itn, nt);
 354	return nt;
 355
 356err_dev_set_mtu:
 357	unregister_netdevice(dev);
 358	return ERR_PTR(err);
 359}
 360
 361int ip_tunnel_rcv(struct ip_tunnel *tunnel, struct sk_buff *skb,
 362		  const struct tnl_ptk_info *tpi, struct metadata_dst *tun_dst,
 363		  bool log_ecn_error)
 364{
 365	struct pcpu_sw_netstats *tstats;
 366	const struct iphdr *iph = ip_hdr(skb);
 367	int err;
 368
 369#ifdef CONFIG_NET_IPGRE_BROADCAST
 370	if (ipv4_is_multicast(iph->daddr)) {
 371		tunnel->dev->stats.multicast++;
 372		skb->pkt_type = PACKET_BROADCAST;
 373	}
 374#endif
 375
 376	if ((!(tpi->flags&TUNNEL_CSUM) &&  (tunnel->parms.i_flags&TUNNEL_CSUM)) ||
 377	     ((tpi->flags&TUNNEL_CSUM) && !(tunnel->parms.i_flags&TUNNEL_CSUM))) {
 378		tunnel->dev->stats.rx_crc_errors++;
 379		tunnel->dev->stats.rx_errors++;
 380		goto drop;
 381	}
 382
 383	if (tunnel->parms.i_flags&TUNNEL_SEQ) {
 384		if (!(tpi->flags&TUNNEL_SEQ) ||
 385		    (tunnel->i_seqno && (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
 386			tunnel->dev->stats.rx_fifo_errors++;
 387			tunnel->dev->stats.rx_errors++;
 388			goto drop;
 389		}
 390		tunnel->i_seqno = ntohl(tpi->seq) + 1;
 391	}
 392
 393	skb_reset_network_header(skb);
 394
 395	err = IP_ECN_decapsulate(iph, skb);
 396	if (unlikely(err)) {
 397		if (log_ecn_error)
 398			net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
 399					&iph->saddr, iph->tos);
 400		if (err > 1) {
 401			++tunnel->dev->stats.rx_frame_errors;
 402			++tunnel->dev->stats.rx_errors;
 403			goto drop;
 404		}
 405	}
 406
 407	tstats = this_cpu_ptr(tunnel->dev->tstats);
 408	u64_stats_update_begin(&tstats->syncp);
 409	tstats->rx_packets++;
 410	tstats->rx_bytes += skb->len;
 411	u64_stats_update_end(&tstats->syncp);
 412
 413	skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
 414
 415	if (tunnel->dev->type == ARPHRD_ETHER) {
 416		skb->protocol = eth_type_trans(skb, tunnel->dev);
 417		skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
 418	} else {
 419		skb->dev = tunnel->dev;
 420	}
 421
 422	if (tun_dst)
 423		skb_dst_set(skb, (struct dst_entry *)tun_dst);
 424
 425	gro_cells_receive(&tunnel->gro_cells, skb);
 426	return 0;
 427
 428drop:
 429	if (tun_dst)
 430		dst_release((struct dst_entry *)tun_dst);
 431	kfree_skb(skb);
 432	return 0;
 433}
 434EXPORT_SYMBOL_GPL(ip_tunnel_rcv);
 435
 436int ip_tunnel_encap_add_ops(const struct ip_tunnel_encap_ops *ops,
 437			    unsigned int num)
 438{
 439	if (num >= MAX_IPTUN_ENCAP_OPS)
 440		return -ERANGE;
 441
 442	return !cmpxchg((const struct ip_tunnel_encap_ops **)
 443			&iptun_encaps[num],
 444			NULL, ops) ? 0 : -1;
 445}
 446EXPORT_SYMBOL(ip_tunnel_encap_add_ops);
 447
 448int ip_tunnel_encap_del_ops(const struct ip_tunnel_encap_ops *ops,
 449			    unsigned int num)
 450{
 451	int ret;
 452
 453	if (num >= MAX_IPTUN_ENCAP_OPS)
 454		return -ERANGE;
 455
 456	ret = (cmpxchg((const struct ip_tunnel_encap_ops **)
 457		       &iptun_encaps[num],
 458		       ops, NULL) == ops) ? 0 : -1;
 459
 460	synchronize_net();
 461
 462	return ret;
 463}
 464EXPORT_SYMBOL(ip_tunnel_encap_del_ops);
 465
 466int ip_tunnel_encap_setup(struct ip_tunnel *t,
 467			  struct ip_tunnel_encap *ipencap)
 468{
 469	int hlen;
 470
 471	memset(&t->encap, 0, sizeof(t->encap));
 472
 473	hlen = ip_encap_hlen(ipencap);
 474	if (hlen < 0)
 475		return hlen;
 476
 477	t->encap.type = ipencap->type;
 478	t->encap.sport = ipencap->sport;
 479	t->encap.dport = ipencap->dport;
 480	t->encap.flags = ipencap->flags;
 481
 482	t->encap_hlen = hlen;
 483	t->hlen = t->encap_hlen + t->tun_hlen;
 484
 485	return 0;
 486}
 487EXPORT_SYMBOL_GPL(ip_tunnel_encap_setup);
 488
 489static int tnl_update_pmtu(struct net_device *dev, struct sk_buff *skb,
 490			    struct rtable *rt, __be16 df,
 491			    const struct iphdr *inner_iph,
 492			    int tunnel_hlen, __be32 dst, bool md)
 493{
 494	struct ip_tunnel *tunnel = netdev_priv(dev);
 495	int pkt_size;
 496	int mtu;
 497
 498	tunnel_hlen = md ? tunnel_hlen : tunnel->hlen;
 499	pkt_size = skb->len - tunnel_hlen - dev->hard_header_len;
 500
 501	if (df)
 502		mtu = dst_mtu(&rt->dst) - dev->hard_header_len
 503					- sizeof(struct iphdr) - tunnel_hlen;
 504	else
 505		mtu = skb_valid_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
 506
 507	if (skb_valid_dst(skb))
 508		skb_dst_update_pmtu(skb, mtu);
 509
 510	if (skb->protocol == htons(ETH_P_IP)) {
 511		if (!skb_is_gso(skb) &&
 512		    (inner_iph->frag_off & htons(IP_DF)) &&
 513		    mtu < pkt_size) {
 514			memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
 515			icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
 516			return -E2BIG;
 517		}
 518	}
 519#if IS_ENABLED(CONFIG_IPV6)
 520	else if (skb->protocol == htons(ETH_P_IPV6)) {
 521		struct rt6_info *rt6;
 522		__be32 daddr;
 523
 524		rt6 = skb_valid_dst(skb) ? (struct rt6_info *)skb_dst(skb) :
 525					   NULL;
 526		daddr = md ? dst : tunnel->parms.iph.daddr;
 527
 528		if (rt6 && mtu < dst_mtu(skb_dst(skb)) &&
 529			   mtu >= IPV6_MIN_MTU) {
 530			if ((daddr && !ipv4_is_multicast(daddr)) ||
 
 531			    rt6->rt6i_dst.plen == 128) {
 532				rt6->rt6i_flags |= RTF_MODIFIED;
 533				dst_metric_set(skb_dst(skb), RTAX_MTU, mtu);
 534			}
 535		}
 536
 537		if (!skb_is_gso(skb) && mtu >= IPV6_MIN_MTU &&
 538					mtu < pkt_size) {
 539			icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
 540			return -E2BIG;
 541		}
 542	}
 543#endif
 544	return 0;
 545}
 546
 547void ip_md_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
 548		       u8 proto, int tunnel_hlen)
 549{
 550	struct ip_tunnel *tunnel = netdev_priv(dev);
 551	u32 headroom = sizeof(struct iphdr);
 552	struct ip_tunnel_info *tun_info;
 553	const struct ip_tunnel_key *key;
 554	const struct iphdr *inner_iph;
 555	struct rtable *rt = NULL;
 556	struct flowi4 fl4;
 557	__be16 df = 0;
 558	u8 tos, ttl;
 559	bool use_cache;
 560
 561	tun_info = skb_tunnel_info(skb);
 562	if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
 563		     ip_tunnel_info_af(tun_info) != AF_INET))
 564		goto tx_error;
 565	key = &tun_info->key;
 566	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 567	inner_iph = (const struct iphdr *)skb_inner_network_header(skb);
 568	tos = key->tos;
 569	if (tos == 1) {
 570		if (skb->protocol == htons(ETH_P_IP))
 571			tos = inner_iph->tos;
 572		else if (skb->protocol == htons(ETH_P_IPV6))
 573			tos = ipv6_get_dsfield((const struct ipv6hdr *)inner_iph);
 574	}
 575	ip_tunnel_init_flow(&fl4, proto, key->u.ipv4.dst, key->u.ipv4.src,
 576			    tunnel_id_to_key32(key->tun_id), RT_TOS(tos),
 577			    0, skb->mark, skb_get_hash(skb));
 578	if (tunnel->encap.type != TUNNEL_ENCAP_NONE)
 579		goto tx_error;
 580
 581	use_cache = ip_tunnel_dst_cache_usable(skb, tun_info);
 582	if (use_cache)
 583		rt = dst_cache_get_ip4(&tun_info->dst_cache, &fl4.saddr);
 584	if (!rt) {
 585		rt = ip_route_output_key(tunnel->net, &fl4);
 586		if (IS_ERR(rt)) {
 587			dev->stats.tx_carrier_errors++;
 588			goto tx_error;
 589		}
 590		if (use_cache)
 591			dst_cache_set_ip4(&tun_info->dst_cache, &rt->dst,
 592					  fl4.saddr);
 593	}
 594	if (rt->dst.dev == dev) {
 595		ip_rt_put(rt);
 596		dev->stats.collisions++;
 597		goto tx_error;
 598	}
 599
 600	if (key->tun_flags & TUNNEL_DONT_FRAGMENT)
 601		df = htons(IP_DF);
 602	if (tnl_update_pmtu(dev, skb, rt, df, inner_iph, tunnel_hlen,
 603			    key->u.ipv4.dst, true)) {
 604		ip_rt_put(rt);
 605		goto tx_error;
 606	}
 607
 608	tos = ip_tunnel_ecn_encap(tos, inner_iph, skb);
 609	ttl = key->ttl;
 610	if (ttl == 0) {
 611		if (skb->protocol == htons(ETH_P_IP))
 612			ttl = inner_iph->ttl;
 613		else if (skb->protocol == htons(ETH_P_IPV6))
 614			ttl = ((const struct ipv6hdr *)inner_iph)->hop_limit;
 615		else
 616			ttl = ip4_dst_hoplimit(&rt->dst);
 617	}
 618
 619	if (!df && skb->protocol == htons(ETH_P_IP))
 
 620		df = inner_iph->frag_off & htons(IP_DF);
 621
 622	headroom += LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len;
 623	if (headroom > dev->needed_headroom)
 624		dev->needed_headroom = headroom;
 625
 626	if (skb_cow_head(skb, dev->needed_headroom)) {
 627		ip_rt_put(rt);
 628		goto tx_dropped;
 629	}
 630	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, proto, tos, ttl,
 631		      df, !net_eq(tunnel->net, dev_net(dev)));
 632	return;
 633tx_error:
 634	dev->stats.tx_errors++;
 635	goto kfree;
 636tx_dropped:
 637	dev->stats.tx_dropped++;
 638kfree:
 639	kfree_skb(skb);
 640}
 641EXPORT_SYMBOL_GPL(ip_md_tunnel_xmit);
 642
 643void ip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev,
 644		    const struct iphdr *tnl_params, u8 protocol)
 645{
 646	struct ip_tunnel *tunnel = netdev_priv(dev);
 647	struct ip_tunnel_info *tun_info = NULL;
 648	const struct iphdr *inner_iph;
 
 
 
 
 649	unsigned int max_headroom;	/* The extra header space needed */
 650	struct rtable *rt = NULL;		/* Route to the other host */
 651	bool use_cache = false;
 652	struct flowi4 fl4;
 653	bool md = false;
 654	bool connected;
 655	u8 tos, ttl;
 656	__be32 dst;
 657	__be16 df;
 658
 659	inner_iph = (const struct iphdr *)skb_inner_network_header(skb);
 660	connected = (tunnel->parms.iph.daddr != 0);
 661
 662	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 663
 664	dst = tnl_params->daddr;
 665	if (dst == 0) {
 666		/* NBMA tunnel */
 667
 668		if (!skb_dst(skb)) {
 669			dev->stats.tx_fifo_errors++;
 670			goto tx_error;
 671		}
 672
 673		tun_info = skb_tunnel_info(skb);
 674		if (tun_info && (tun_info->mode & IP_TUNNEL_INFO_TX) &&
 675		    ip_tunnel_info_af(tun_info) == AF_INET &&
 676		    tun_info->key.u.ipv4.dst) {
 677			dst = tun_info->key.u.ipv4.dst;
 678			md = true;
 679			connected = true;
 680		}
 681		else if (skb->protocol == htons(ETH_P_IP)) {
 682			rt = skb_rtable(skb);
 683			dst = rt_nexthop(rt, inner_iph->daddr);
 684		}
 685#if IS_ENABLED(CONFIG_IPV6)
 686		else if (skb->protocol == htons(ETH_P_IPV6)) {
 687			const struct in6_addr *addr6;
 688			struct neighbour *neigh;
 689			bool do_tx_error_icmp;
 690			int addr_type;
 691
 692			neigh = dst_neigh_lookup(skb_dst(skb),
 693						 &ipv6_hdr(skb)->daddr);
 694			if (!neigh)
 695				goto tx_error;
 696
 697			addr6 = (const struct in6_addr *)&neigh->primary_key;
 698			addr_type = ipv6_addr_type(addr6);
 699
 700			if (addr_type == IPV6_ADDR_ANY) {
 701				addr6 = &ipv6_hdr(skb)->daddr;
 702				addr_type = ipv6_addr_type(addr6);
 703			}
 704
 705			if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 706				do_tx_error_icmp = true;
 707			else {
 708				do_tx_error_icmp = false;
 709				dst = addr6->s6_addr32[3];
 710			}
 711			neigh_release(neigh);
 712			if (do_tx_error_icmp)
 713				goto tx_error_icmp;
 714		}
 715#endif
 716		else
 717			goto tx_error;
 718
 719		if (!md)
 720			connected = false;
 721	}
 722
 723	tos = tnl_params->tos;
 724	if (tos & 0x1) {
 725		tos &= ~0x1;
 726		if (skb->protocol == htons(ETH_P_IP)) {
 727			tos = inner_iph->tos;
 728			connected = false;
 729		} else if (skb->protocol == htons(ETH_P_IPV6)) {
 730			tos = ipv6_get_dsfield((const struct ipv6hdr *)inner_iph);
 731			connected = false;
 732		}
 733	}
 734
 735	ip_tunnel_init_flow(&fl4, protocol, dst, tnl_params->saddr,
 736			    tunnel->parms.o_key, RT_TOS(tos), tunnel->parms.link,
 737			    tunnel->fwmark, skb_get_hash(skb));
 738
 739	if (ip_tunnel_encap(skb, tunnel, &protocol, &fl4) < 0)
 740		goto tx_error;
 741
 742	if (connected && md) {
 743		use_cache = ip_tunnel_dst_cache_usable(skb, tun_info);
 744		if (use_cache)
 745			rt = dst_cache_get_ip4(&tun_info->dst_cache,
 746					       &fl4.saddr);
 747	} else {
 748		rt = connected ? dst_cache_get_ip4(&tunnel->dst_cache,
 749						&fl4.saddr) : NULL;
 750	}
 751
 752	if (!rt) {
 753		rt = ip_route_output_key(tunnel->net, &fl4);
 754
 755		if (IS_ERR(rt)) {
 756			dev->stats.tx_carrier_errors++;
 757			goto tx_error;
 758		}
 759		if (use_cache)
 760			dst_cache_set_ip4(&tun_info->dst_cache, &rt->dst,
 761					  fl4.saddr);
 762		else if (!md && connected)
 763			dst_cache_set_ip4(&tunnel->dst_cache, &rt->dst,
 764					  fl4.saddr);
 765	}
 766
 767	if (rt->dst.dev == dev) {
 768		ip_rt_put(rt);
 769		dev->stats.collisions++;
 770		goto tx_error;
 771	}
 772
 773	if (tnl_update_pmtu(dev, skb, rt, tnl_params->frag_off, inner_iph,
 774			    0, 0, false)) {
 775		ip_rt_put(rt);
 776		goto tx_error;
 777	}
 778
 779	if (tunnel->err_count > 0) {
 780		if (time_before(jiffies,
 781				tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
 782			tunnel->err_count--;
 783
 784			dst_link_failure(skb);
 785		} else
 786			tunnel->err_count = 0;
 787	}
 788
 789	tos = ip_tunnel_ecn_encap(tos, inner_iph, skb);
 790	ttl = tnl_params->ttl;
 791	if (ttl == 0) {
 792		if (skb->protocol == htons(ETH_P_IP))
 793			ttl = inner_iph->ttl;
 794#if IS_ENABLED(CONFIG_IPV6)
 795		else if (skb->protocol == htons(ETH_P_IPV6))
 796			ttl = ((const struct ipv6hdr *)inner_iph)->hop_limit;
 797#endif
 798		else
 799			ttl = ip4_dst_hoplimit(&rt->dst);
 800	}
 801
 802	df = tnl_params->frag_off;
 803	if (skb->protocol == htons(ETH_P_IP) && !tunnel->ignore_df)
 804		df |= (inner_iph->frag_off&htons(IP_DF));
 805
 806	max_headroom = LL_RESERVED_SPACE(rt->dst.dev) + sizeof(struct iphdr)
 807			+ rt->dst.header_len + ip_encap_hlen(&tunnel->encap);
 808	if (max_headroom > dev->needed_headroom)
 809		dev->needed_headroom = max_headroom;
 810
 811	if (skb_cow_head(skb, dev->needed_headroom)) {
 812		ip_rt_put(rt);
 813		dev->stats.tx_dropped++;
 814		kfree_skb(skb);
 815		return;
 816	}
 817
 818	iptunnel_xmit(NULL, rt, skb, fl4.saddr, fl4.daddr, protocol, tos, ttl,
 819		      df, !net_eq(tunnel->net, dev_net(dev)));
 820	return;
 821
 822#if IS_ENABLED(CONFIG_IPV6)
 823tx_error_icmp:
 824	dst_link_failure(skb);
 825#endif
 826tx_error:
 827	dev->stats.tx_errors++;
 828	kfree_skb(skb);
 829}
 830EXPORT_SYMBOL_GPL(ip_tunnel_xmit);
 831
 832static void ip_tunnel_update(struct ip_tunnel_net *itn,
 833			     struct ip_tunnel *t,
 834			     struct net_device *dev,
 835			     struct ip_tunnel_parm *p,
 836			     bool set_mtu,
 837			     __u32 fwmark)
 838{
 839	ip_tunnel_del(itn, t);
 840	t->parms.iph.saddr = p->iph.saddr;
 841	t->parms.iph.daddr = p->iph.daddr;
 842	t->parms.i_key = p->i_key;
 843	t->parms.o_key = p->o_key;
 844	if (dev->type != ARPHRD_ETHER) {
 845		memcpy(dev->dev_addr, &p->iph.saddr, 4);
 846		memcpy(dev->broadcast, &p->iph.daddr, 4);
 847	}
 848	ip_tunnel_add(itn, t);
 849
 850	t->parms.iph.ttl = p->iph.ttl;
 851	t->parms.iph.tos = p->iph.tos;
 852	t->parms.iph.frag_off = p->iph.frag_off;
 853
 854	if (t->parms.link != p->link || t->fwmark != fwmark) {
 855		int mtu;
 856
 857		t->parms.link = p->link;
 858		t->fwmark = fwmark;
 859		mtu = ip_tunnel_bind_dev(dev);
 860		if (set_mtu)
 861			dev->mtu = mtu;
 862	}
 863	dst_cache_reset(&t->dst_cache);
 864	netdev_state_change(dev);
 865}
 866
 867int ip_tunnel_ioctl(struct net_device *dev, struct ip_tunnel_parm *p, int cmd)
 868{
 869	int err = 0;
 870	struct ip_tunnel *t = netdev_priv(dev);
 871	struct net *net = t->net;
 872	struct ip_tunnel_net *itn = net_generic(net, t->ip_tnl_net_id);
 873
 874	switch (cmd) {
 875	case SIOCGETTUNNEL:
 876		if (dev == itn->fb_tunnel_dev) {
 877			t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
 878			if (!t)
 879				t = netdev_priv(dev);
 880		}
 881		memcpy(p, &t->parms, sizeof(*p));
 882		break;
 883
 884	case SIOCADDTUNNEL:
 885	case SIOCCHGTUNNEL:
 886		err = -EPERM;
 887		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 888			goto done;
 889		if (p->iph.ttl)
 890			p->iph.frag_off |= htons(IP_DF);
 891		if (!(p->i_flags & VTI_ISVTI)) {
 892			if (!(p->i_flags & TUNNEL_KEY))
 893				p->i_key = 0;
 894			if (!(p->o_flags & TUNNEL_KEY))
 895				p->o_key = 0;
 896		}
 897
 898		t = ip_tunnel_find(itn, p, itn->type);
 899
 900		if (cmd == SIOCADDTUNNEL) {
 901			if (!t) {
 902				t = ip_tunnel_create(net, itn, p);
 903				err = PTR_ERR_OR_ZERO(t);
 904				break;
 905			}
 906
 907			err = -EEXIST;
 908			break;
 909		}
 910		if (dev != itn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 911			if (t) {
 912				if (t->dev != dev) {
 913					err = -EEXIST;
 914					break;
 915				}
 916			} else {
 917				unsigned int nflags = 0;
 918
 919				if (ipv4_is_multicast(p->iph.daddr))
 920					nflags = IFF_BROADCAST;
 921				else if (p->iph.daddr)
 922					nflags = IFF_POINTOPOINT;
 923
 924				if ((dev->flags^nflags)&(IFF_POINTOPOINT|IFF_BROADCAST)) {
 925					err = -EINVAL;
 926					break;
 927				}
 928
 929				t = netdev_priv(dev);
 930			}
 931		}
 932
 933		if (t) {
 934			err = 0;
 935			ip_tunnel_update(itn, t, dev, p, true, 0);
 936		} else {
 937			err = -ENOENT;
 938		}
 939		break;
 940
 941	case SIOCDELTUNNEL:
 942		err = -EPERM;
 943		if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
 944			goto done;
 945
 946		if (dev == itn->fb_tunnel_dev) {
 947			err = -ENOENT;
 948			t = ip_tunnel_find(itn, p, itn->fb_tunnel_dev->type);
 949			if (!t)
 950				goto done;
 951			err = -EPERM;
 952			if (t == netdev_priv(itn->fb_tunnel_dev))
 953				goto done;
 954			dev = t->dev;
 955		}
 956		unregister_netdevice(dev);
 957		err = 0;
 958		break;
 959
 960	default:
 961		err = -EINVAL;
 962	}
 963
 964done:
 965	return err;
 966}
 967EXPORT_SYMBOL_GPL(ip_tunnel_ioctl);
 968
 969int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict)
 970{
 971	struct ip_tunnel *tunnel = netdev_priv(dev);
 972	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
 973	int max_mtu = IP_MAX_MTU - dev->hard_header_len - t_hlen;
 974
 975	if (new_mtu < ETH_MIN_MTU)
 976		return -EINVAL;
 977
 978	if (new_mtu > max_mtu) {
 979		if (strict)
 980			return -EINVAL;
 981
 982		new_mtu = max_mtu;
 983	}
 984
 985	dev->mtu = new_mtu;
 986	return 0;
 987}
 988EXPORT_SYMBOL_GPL(__ip_tunnel_change_mtu);
 989
 990int ip_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 991{
 992	return __ip_tunnel_change_mtu(dev, new_mtu, true);
 993}
 994EXPORT_SYMBOL_GPL(ip_tunnel_change_mtu);
 995
 996static void ip_tunnel_dev_free(struct net_device *dev)
 997{
 998	struct ip_tunnel *tunnel = netdev_priv(dev);
 999
1000	gro_cells_destroy(&tunnel->gro_cells);
1001	dst_cache_destroy(&tunnel->dst_cache);
1002	free_percpu(dev->tstats);
1003}
1004
1005void ip_tunnel_dellink(struct net_device *dev, struct list_head *head)
1006{
1007	struct ip_tunnel *tunnel = netdev_priv(dev);
1008	struct ip_tunnel_net *itn;
1009
1010	itn = net_generic(tunnel->net, tunnel->ip_tnl_net_id);
1011
1012	if (itn->fb_tunnel_dev != dev) {
1013		ip_tunnel_del(itn, netdev_priv(dev));
1014		unregister_netdevice_queue(dev, head);
1015	}
1016}
1017EXPORT_SYMBOL_GPL(ip_tunnel_dellink);
1018
1019struct net *ip_tunnel_get_link_net(const struct net_device *dev)
1020{
1021	struct ip_tunnel *tunnel = netdev_priv(dev);
1022
1023	return tunnel->net;
1024}
1025EXPORT_SYMBOL(ip_tunnel_get_link_net);
1026
1027int ip_tunnel_get_iflink(const struct net_device *dev)
1028{
1029	struct ip_tunnel *tunnel = netdev_priv(dev);
1030
1031	return tunnel->parms.link;
1032}
1033EXPORT_SYMBOL(ip_tunnel_get_iflink);
1034
1035int ip_tunnel_init_net(struct net *net, unsigned int ip_tnl_net_id,
1036				  struct rtnl_link_ops *ops, char *devname)
1037{
1038	struct ip_tunnel_net *itn = net_generic(net, ip_tnl_net_id);
1039	struct ip_tunnel_parm parms;
1040	unsigned int i;
1041
1042	itn->rtnl_link_ops = ops;
1043	for (i = 0; i < IP_TNL_HASH_SIZE; i++)
1044		INIT_HLIST_HEAD(&itn->tunnels[i]);
1045
1046	if (!ops || !net_has_fallback_tunnels(net)) {
1047		struct ip_tunnel_net *it_init_net;
1048
1049		it_init_net = net_generic(&init_net, ip_tnl_net_id);
1050		itn->type = it_init_net->type;
1051		itn->fb_tunnel_dev = NULL;
1052		return 0;
1053	}
1054
1055	memset(&parms, 0, sizeof(parms));
1056	if (devname)
1057		strlcpy(parms.name, devname, IFNAMSIZ);
1058
1059	rtnl_lock();
1060	itn->fb_tunnel_dev = __ip_tunnel_create(net, ops, &parms);
1061	/* FB netdevice is special: we have one, and only one per netns.
1062	 * Allowing to move it to another netns is clearly unsafe.
1063	 */
1064	if (!IS_ERR(itn->fb_tunnel_dev)) {
1065		itn->fb_tunnel_dev->features |= NETIF_F_NETNS_LOCAL;
1066		itn->fb_tunnel_dev->mtu = ip_tunnel_bind_dev(itn->fb_tunnel_dev);
1067		ip_tunnel_add(itn, netdev_priv(itn->fb_tunnel_dev));
1068		itn->type = itn->fb_tunnel_dev->type;
1069	}
1070	rtnl_unlock();
1071
1072	return PTR_ERR_OR_ZERO(itn->fb_tunnel_dev);
1073}
1074EXPORT_SYMBOL_GPL(ip_tunnel_init_net);
1075
1076static void ip_tunnel_destroy(struct net *net, struct ip_tunnel_net *itn,
1077			      struct list_head *head,
1078			      struct rtnl_link_ops *ops)
1079{
1080	struct net_device *dev, *aux;
1081	int h;
1082
1083	for_each_netdev_safe(net, dev, aux)
1084		if (dev->rtnl_link_ops == ops)
1085			unregister_netdevice_queue(dev, head);
1086
1087	for (h = 0; h < IP_TNL_HASH_SIZE; h++) {
1088		struct ip_tunnel *t;
1089		struct hlist_node *n;
1090		struct hlist_head *thead = &itn->tunnels[h];
1091
1092		hlist_for_each_entry_safe(t, n, thead, hash_node)
1093			/* If dev is in the same netns, it has already
1094			 * been added to the list by the previous loop.
1095			 */
1096			if (!net_eq(dev_net(t->dev), net))
1097				unregister_netdevice_queue(t->dev, head);
1098	}
1099}
1100
1101void ip_tunnel_delete_nets(struct list_head *net_list, unsigned int id,
1102			   struct rtnl_link_ops *ops)
1103{
1104	struct ip_tunnel_net *itn;
1105	struct net *net;
1106	LIST_HEAD(list);
1107
1108	rtnl_lock();
1109	list_for_each_entry(net, net_list, exit_list) {
1110		itn = net_generic(net, id);
1111		ip_tunnel_destroy(net, itn, &list, ops);
1112	}
1113	unregister_netdevice_many(&list);
1114	rtnl_unlock();
1115}
1116EXPORT_SYMBOL_GPL(ip_tunnel_delete_nets);
1117
1118int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
1119		      struct ip_tunnel_parm *p, __u32 fwmark)
1120{
1121	struct ip_tunnel *nt;
1122	struct net *net = dev_net(dev);
1123	struct ip_tunnel_net *itn;
1124	int mtu;
1125	int err;
1126
1127	nt = netdev_priv(dev);
1128	itn = net_generic(net, nt->ip_tnl_net_id);
1129
1130	if (nt->collect_md) {
1131		if (rtnl_dereference(itn->collect_md_tun))
1132			return -EEXIST;
1133	} else {
1134		if (ip_tunnel_find(itn, p, dev->type))
1135			return -EEXIST;
1136	}
1137
1138	nt->net = net;
1139	nt->parms = *p;
1140	nt->fwmark = fwmark;
1141	err = register_netdevice(dev);
1142	if (err)
1143		goto err_register_netdevice;
1144
1145	if (dev->type == ARPHRD_ETHER && !tb[IFLA_ADDRESS])
1146		eth_hw_addr_random(dev);
1147
1148	mtu = ip_tunnel_bind_dev(dev);
1149	if (tb[IFLA_MTU]) {
1150		unsigned int max = IP_MAX_MTU - dev->hard_header_len - nt->hlen;
1151
1152		mtu = clamp(dev->mtu, (unsigned int)ETH_MIN_MTU,
1153			    (unsigned int)(max - sizeof(struct iphdr)));
1154	}
1155
1156	err = dev_set_mtu(dev, mtu);
1157	if (err)
1158		goto err_dev_set_mtu;
1159
1160	ip_tunnel_add(itn, nt);
1161	return 0;
1162
1163err_dev_set_mtu:
1164	unregister_netdevice(dev);
1165err_register_netdevice:
1166	return err;
1167}
1168EXPORT_SYMBOL_GPL(ip_tunnel_newlink);
1169
1170int ip_tunnel_changelink(struct net_device *dev, struct nlattr *tb[],
1171			 struct ip_tunnel_parm *p, __u32 fwmark)
1172{
1173	struct ip_tunnel *t;
1174	struct ip_tunnel *tunnel = netdev_priv(dev);
1175	struct net *net = tunnel->net;
1176	struct ip_tunnel_net *itn = net_generic(net, tunnel->ip_tnl_net_id);
1177
1178	if (dev == itn->fb_tunnel_dev)
1179		return -EINVAL;
1180
1181	t = ip_tunnel_find(itn, p, dev->type);
1182
1183	if (t) {
1184		if (t->dev != dev)
1185			return -EEXIST;
1186	} else {
1187		t = tunnel;
1188
1189		if (dev->type != ARPHRD_ETHER) {
1190			unsigned int nflags = 0;
1191
1192			if (ipv4_is_multicast(p->iph.daddr))
1193				nflags = IFF_BROADCAST;
1194			else if (p->iph.daddr)
1195				nflags = IFF_POINTOPOINT;
1196
1197			if ((dev->flags ^ nflags) &
1198			    (IFF_POINTOPOINT | IFF_BROADCAST))
1199				return -EINVAL;
1200		}
1201	}
1202
1203	ip_tunnel_update(itn, t, dev, p, !tb[IFLA_MTU], fwmark);
1204	return 0;
1205}
1206EXPORT_SYMBOL_GPL(ip_tunnel_changelink);
1207
1208int ip_tunnel_init(struct net_device *dev)
1209{
1210	struct ip_tunnel *tunnel = netdev_priv(dev);
1211	struct iphdr *iph = &tunnel->parms.iph;
1212	int err;
1213
1214	dev->needs_free_netdev = true;
1215	dev->priv_destructor = ip_tunnel_dev_free;
1216	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1217	if (!dev->tstats)
1218		return -ENOMEM;
1219
1220	err = dst_cache_init(&tunnel->dst_cache, GFP_KERNEL);
1221	if (err) {
1222		free_percpu(dev->tstats);
1223		return err;
1224	}
1225
1226	err = gro_cells_init(&tunnel->gro_cells, dev);
1227	if (err) {
1228		dst_cache_destroy(&tunnel->dst_cache);
1229		free_percpu(dev->tstats);
1230		return err;
1231	}
1232
1233	tunnel->dev = dev;
1234	tunnel->net = dev_net(dev);
1235	strcpy(tunnel->parms.name, dev->name);
1236	iph->version		= 4;
1237	iph->ihl		= 5;
1238
1239	if (tunnel->collect_md) {
1240		dev->features |= NETIF_F_NETNS_LOCAL;
1241		netif_keep_dst(dev);
1242	}
1243	return 0;
1244}
1245EXPORT_SYMBOL_GPL(ip_tunnel_init);
1246
1247void ip_tunnel_uninit(struct net_device *dev)
1248{
1249	struct ip_tunnel *tunnel = netdev_priv(dev);
1250	struct net *net = tunnel->net;
1251	struct ip_tunnel_net *itn;
1252
1253	itn = net_generic(net, tunnel->ip_tnl_net_id);
1254	/* fb_tunnel_dev will be unregisted in net-exit call. */
1255	if (itn->fb_tunnel_dev != dev)
1256		ip_tunnel_del(itn, netdev_priv(dev));
1257
1258	dst_cache_reset(&tunnel->dst_cache);
1259}
1260EXPORT_SYMBOL_GPL(ip_tunnel_uninit);
1261
1262/* Do least required initialization, rest of init is done in tunnel_init call */
1263void ip_tunnel_setup(struct net_device *dev, unsigned int net_id)
1264{
1265	struct ip_tunnel *tunnel = netdev_priv(dev);
1266	tunnel->ip_tnl_net_id = net_id;
1267}
1268EXPORT_SYMBOL_GPL(ip_tunnel_setup);
1269
1270MODULE_LICENSE("GPL");