Linux Audio

Check our new training course

Loading...
v4.17
 
   1/*
   2 * VXLAN: Virtual eXtensible Local Area Network
   3 *
   4 * Copyright (c) 2012-2013 Vyatta Inc.
   5 *
   6 * This program is free software; you can redistribute it and/or modify
   7 * it under the terms of the GNU General Public License version 2 as
   8 * published by the Free Software Foundation.
   9 */
  10
  11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  12
  13#include <linux/kernel.h>
  14#include <linux/module.h>
  15#include <linux/errno.h>
  16#include <linux/slab.h>
  17#include <linux/udp.h>
  18#include <linux/igmp.h>
  19#include <linux/if_ether.h>
  20#include <linux/ethtool.h>
  21#include <net/arp.h>
  22#include <net/ndisc.h>
 
  23#include <net/ip.h>
  24#include <net/icmp.h>
  25#include <net/rtnetlink.h>
  26#include <net/inet_ecn.h>
  27#include <net/net_namespace.h>
  28#include <net/netns/generic.h>
  29#include <net/tun_proto.h>
  30#include <net/vxlan.h>
 
  31
  32#if IS_ENABLED(CONFIG_IPV6)
  33#include <net/ip6_tunnel.h>
  34#include <net/ip6_checksum.h>
  35#endif
  36
  37#define VXLAN_VERSION	"0.1"
  38
  39#define PORT_HASH_BITS	8
  40#define PORT_HASH_SIZE  (1<<PORT_HASH_BITS)
  41#define FDB_AGE_DEFAULT 300 /* 5 min */
  42#define FDB_AGE_INTERVAL (10 * HZ)	/* rescan interval */
  43
  44/* UDP port for VXLAN traffic.
  45 * The IANA assigned port is 4789, but the Linux default is 8472
  46 * for compatibility with early adopters.
  47 */
  48static unsigned short vxlan_port __read_mostly = 8472;
  49module_param_named(udp_port, vxlan_port, ushort, 0444);
  50MODULE_PARM_DESC(udp_port, "Destination UDP port");
  51
  52static bool log_ecn_error = true;
  53module_param(log_ecn_error, bool, 0644);
  54MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  55
  56static unsigned int vxlan_net_id;
  57static struct rtnl_link_ops vxlan_link_ops;
  58
  59static const u8 all_zeros_mac[ETH_ALEN + 2];
  60
  61static int vxlan_sock_add(struct vxlan_dev *vxlan);
  62
  63static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
  64
  65/* per-network namespace private data for this module */
  66struct vxlan_net {
  67	struct list_head  vxlan_list;
  68	struct hlist_head sock_list[PORT_HASH_SIZE];
  69	spinlock_t	  sock_lock;
  70};
  71
  72/* Forwarding table entry */
  73struct vxlan_fdb {
  74	struct hlist_node hlist;	/* linked list of entries */
  75	struct rcu_head	  rcu;
  76	unsigned long	  updated;	/* jiffies */
  77	unsigned long	  used;
  78	struct list_head  remotes;
  79	u8		  eth_addr[ETH_ALEN];
  80	u16		  state;	/* see ndm_state */
  81	__be32		  vni;
  82	u8		  flags;	/* see ndm_flags */
 
 
 
  83};
  84
 
 
  85/* salt for hash table */
  86static u32 vxlan_salt __read_mostly;
  87
  88static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
  89{
  90	return vs->flags & VXLAN_F_COLLECT_METADATA ||
  91	       ip_tunnel_collect_metadata();
  92}
  93
  94#if IS_ENABLED(CONFIG_IPV6)
  95static inline
  96bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
  97{
  98	if (a->sa.sa_family != b->sa.sa_family)
  99		return false;
 100	if (a->sa.sa_family == AF_INET6)
 101		return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
 102	else
 103		return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
 104}
 105
 106static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
 107{
 108	if (ipa->sa.sa_family == AF_INET6)
 109		return ipv6_addr_any(&ipa->sin6.sin6_addr);
 110	else
 111		return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
 112}
 113
 114static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
 115{
 116	if (ipa->sa.sa_family == AF_INET6)
 117		return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
 118	else
 119		return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
 120}
 121
 122static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
 123{
 124	if (nla_len(nla) >= sizeof(struct in6_addr)) {
 125		ip->sin6.sin6_addr = nla_get_in6_addr(nla);
 126		ip->sa.sa_family = AF_INET6;
 127		return 0;
 128	} else if (nla_len(nla) >= sizeof(__be32)) {
 129		ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
 130		ip->sa.sa_family = AF_INET;
 131		return 0;
 132	} else {
 133		return -EAFNOSUPPORT;
 134	}
 135}
 136
 137static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
 138			      const union vxlan_addr *ip)
 139{
 140	if (ip->sa.sa_family == AF_INET6)
 141		return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
 142	else
 143		return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
 144}
 145
 146#else /* !CONFIG_IPV6 */
 147
 148static inline
 149bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
 150{
 151	return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
 152}
 153
 154static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
 155{
 156	return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
 157}
 158
 159static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
 160{
 161	return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
 162}
 163
 164static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
 165{
 166	if (nla_len(nla) >= sizeof(struct in6_addr)) {
 167		return -EAFNOSUPPORT;
 168	} else if (nla_len(nla) >= sizeof(__be32)) {
 169		ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
 170		ip->sa.sa_family = AF_INET;
 171		return 0;
 172	} else {
 173		return -EAFNOSUPPORT;
 174	}
 175}
 176
 177static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
 178			      const union vxlan_addr *ip)
 179{
 180	return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
 181}
 182#endif
 183
 184/* Virtual Network hash table head */
 185static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
 186{
 187	return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
 188}
 189
 190/* Socket hash table head */
 191static inline struct hlist_head *vs_head(struct net *net, __be16 port)
 192{
 193	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
 194
 195	return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
 196}
 197
 198/* First remote destination for a forwarding entry.
 199 * Guaranteed to be non-NULL because remotes are never deleted.
 200 */
 201static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
 202{
 
 
 203	return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
 204}
 205
 206static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
 207{
 
 
 208	return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
 209}
 210
 211/* Find VXLAN socket based on network namespace, address family and UDP port
 212 * and enabled unshareable flags.
 213 */
 214static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
 215					  __be16 port, u32 flags)
 216{
 217	struct vxlan_sock *vs;
 218
 219	flags &= VXLAN_F_RCV_FLAGS;
 220
 221	hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
 222		if (inet_sk(vs->sock->sk)->inet_sport == port &&
 223		    vxlan_get_sk_family(vs) == family &&
 224		    vs->flags == flags)
 
 225			return vs;
 226	}
 227	return NULL;
 228}
 229
 230static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
 231					   __be32 vni)
 232{
 233	struct vxlan_dev_node *node;
 234
 235	/* For flow based devices, map all packets to VNI 0 */
 236	if (vs->flags & VXLAN_F_COLLECT_METADATA)
 237		vni = 0;
 238
 239	hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
 240		if (node->vxlan->default_dst.remote_vni != vni)
 241			continue;
 242
 243		if (IS_ENABLED(CONFIG_IPV6)) {
 244			const struct vxlan_config *cfg = &node->vxlan->cfg;
 245
 246			if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
 247			    cfg->remote_ifindex != ifindex)
 248				continue;
 249		}
 250
 251		return node->vxlan;
 252	}
 253
 254	return NULL;
 255}
 256
 257/* Look up VNI in a per net namespace table */
 258static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
 259					__be32 vni, sa_family_t family,
 260					__be16 port, u32 flags)
 261{
 262	struct vxlan_sock *vs;
 263
 264	vs = vxlan_find_sock(net, family, port, flags);
 265	if (!vs)
 266		return NULL;
 267
 268	return vxlan_vs_find_vni(vs, ifindex, vni);
 269}
 270
 271/* Fill in neighbour message in skbuff. */
 272static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
 273			  const struct vxlan_fdb *fdb,
 274			  u32 portid, u32 seq, int type, unsigned int flags,
 275			  const struct vxlan_rdst *rdst)
 276{
 277	unsigned long now = jiffies;
 278	struct nda_cacheinfo ci;
 
 279	struct nlmsghdr *nlh;
 
 280	struct ndmsg *ndm;
 281	bool send_ip, send_eth;
 
 282
 283	nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
 284	if (nlh == NULL)
 285		return -EMSGSIZE;
 286
 287	ndm = nlmsg_data(nlh);
 288	memset(ndm, 0, sizeof(*ndm));
 289
 290	send_eth = send_ip = true;
 291
 
 
 
 
 
 
 
 
 292	if (type == RTM_GETNEIGH) {
 293		send_ip = !vxlan_addr_any(&rdst->remote_ip);
 
 
 
 
 
 294		send_eth = !is_zero_ether_addr(fdb->eth_addr);
 295		ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
 296	} else
 297		ndm->ndm_family	= AF_BRIDGE;
 298	ndm->ndm_state = fdb->state;
 299	ndm->ndm_ifindex = vxlan->dev->ifindex;
 300	ndm->ndm_flags = fdb->flags;
 
 
 301	ndm->ndm_type = RTN_UNICAST;
 302
 303	if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
 304	    nla_put_s32(skb, NDA_LINK_NETNSID,
 305			peernet2id(dev_net(vxlan->dev), vxlan->net)))
 306		goto nla_put_failure;
 307
 308	if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
 309		goto nla_put_failure;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 310
 311	if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
 312		goto nla_put_failure;
 313
 314	if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
 315	    nla_put_be16(skb, NDA_PORT, rdst->remote_port))
 316		goto nla_put_failure;
 317	if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
 318	    nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
 319		goto nla_put_failure;
 320	if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
 321	    nla_put_u32(skb, NDA_SRC_VNI,
 322			be32_to_cpu(fdb->vni)))
 323		goto nla_put_failure;
 324	if (rdst->remote_ifindex &&
 325	    nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
 326		goto nla_put_failure;
 327
 328	ci.ndm_used	 = jiffies_to_clock_t(now - fdb->used);
 329	ci.ndm_confirmed = 0;
 330	ci.ndm_updated	 = jiffies_to_clock_t(now - fdb->updated);
 331	ci.ndm_refcnt	 = 0;
 332
 333	if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
 334		goto nla_put_failure;
 335
 336	nlmsg_end(skb, nlh);
 337	return 0;
 338
 339nla_put_failure:
 340	nlmsg_cancel(skb, nlh);
 341	return -EMSGSIZE;
 342}
 343
 344static inline size_t vxlan_nlmsg_size(void)
 345{
 346	return NLMSG_ALIGN(sizeof(struct ndmsg))
 347		+ nla_total_size(ETH_ALEN) /* NDA_LLADDR */
 348		+ nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
 349		+ nla_total_size(sizeof(__be16)) /* NDA_PORT */
 350		+ nla_total_size(sizeof(__be32)) /* NDA_VNI */
 351		+ nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
 352		+ nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
 353		+ nla_total_size(sizeof(struct nda_cacheinfo));
 354}
 355
 356static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
 357			     struct vxlan_rdst *rd, int type)
 358{
 359	struct net *net = dev_net(vxlan->dev);
 360	struct sk_buff *skb;
 361	int err = -ENOBUFS;
 362
 363	skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
 364	if (skb == NULL)
 365		goto errout;
 366
 367	err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
 368	if (err < 0) {
 369		/* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
 370		WARN_ON(err == -EMSGSIZE);
 371		kfree_skb(skb);
 372		goto errout;
 373	}
 374
 375	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
 376	return;
 377errout:
 378	if (err < 0)
 379		rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
 380}
 381
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 382static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
 383{
 384	struct vxlan_dev *vxlan = netdev_priv(dev);
 385	struct vxlan_fdb f = {
 386		.state = NUD_STALE,
 387	};
 388	struct vxlan_rdst remote = {
 389		.remote_ip = *ipa, /* goes to NDA_DST */
 390		.remote_vni = cpu_to_be32(VXLAN_N_VID),
 391	};
 392
 393	vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
 394}
 395
 396static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
 397{
 398	struct vxlan_fdb f = {
 399		.state = NUD_STALE,
 400	};
 401	struct vxlan_rdst remote = { };
 402
 403	memcpy(f.eth_addr, eth_addr, ETH_ALEN);
 404
 405	vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
 406}
 407
 408/* Hash Ethernet address */
 409static u32 eth_hash(const unsigned char *addr)
 410{
 411	u64 value = get_unaligned((u64 *)addr);
 412
 413	/* only want 6 bytes */
 414#ifdef __BIG_ENDIAN
 415	value >>= 16;
 416#else
 417	value <<= 16;
 418#endif
 419	return hash_64(value, FDB_HASH_BITS);
 420}
 421
 422static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
 423{
 424	/* use 1 byte of OUI and 3 bytes of NIC */
 425	u32 key = get_unaligned((u32 *)(addr + 2));
 426
 427	return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
 428}
 429
 
 
 
 
 
 
 
 
 430/* Hash chain to use given mac address */
 431static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
 432						const u8 *mac, __be32 vni)
 433{
 434	if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
 435		return &vxlan->fdb_head[eth_vni_hash(mac, vni)];
 436	else
 437		return &vxlan->fdb_head[eth_hash(mac)];
 438}
 439
 440/* Look up Ethernet address in forwarding table */
 441static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
 442					  const u8 *mac, __be32 vni)
 443{
 444	struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
 445	struct vxlan_fdb *f;
 446
 447	hlist_for_each_entry_rcu(f, head, hlist) {
 448		if (ether_addr_equal(mac, f->eth_addr)) {
 449			if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
 450				if (vni == f->vni)
 451					return f;
 452			} else {
 453				return f;
 454			}
 455		}
 456	}
 457
 458	return NULL;
 459}
 460
 461static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
 462					const u8 *mac, __be32 vni)
 463{
 464	struct vxlan_fdb *f;
 465
 466	f = __vxlan_find_mac(vxlan, mac, vni);
 467	if (f)
 468		f->used = jiffies;
 469
 470	return f;
 471}
 472
 473/* caller should hold vxlan->hash_lock */
 474static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
 475					      union vxlan_addr *ip, __be16 port,
 476					      __be32 vni, __u32 ifindex)
 477{
 478	struct vxlan_rdst *rd;
 479
 480	list_for_each_entry(rd, &f->remotes, list) {
 481		if (vxlan_addr_equal(&rd->remote_ip, ip) &&
 482		    rd->remote_port == port &&
 483		    rd->remote_vni == vni &&
 484		    rd->remote_ifindex == ifindex)
 485			return rd;
 486	}
 487
 488	return NULL;
 489}
 490
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 491/* Replace destination of unicast mac */
 492static int vxlan_fdb_replace(struct vxlan_fdb *f,
 493			     union vxlan_addr *ip, __be16 port, __be32 vni,
 494			     __u32 ifindex)
 495{
 496	struct vxlan_rdst *rd;
 497
 498	rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
 499	if (rd)
 500		return 0;
 501
 502	rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
 503	if (!rd)
 504		return 0;
 505
 
 506	dst_cache_reset(&rd->dst_cache);
 507	rd->remote_ip = *ip;
 508	rd->remote_port = port;
 509	rd->remote_vni = vni;
 510	rd->remote_ifindex = ifindex;
 
 511	return 1;
 512}
 513
 514/* Add/update destinations for multicast */
 515static int vxlan_fdb_append(struct vxlan_fdb *f,
 516			    union vxlan_addr *ip, __be16 port, __be32 vni,
 517			    __u32 ifindex, struct vxlan_rdst **rdp)
 518{
 519	struct vxlan_rdst *rd;
 520
 521	rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
 522	if (rd)
 523		return 0;
 524
 525	rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
 526	if (rd == NULL)
 527		return -ENOBUFS;
 528
 529	if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
 530		kfree(rd);
 531		return -ENOBUFS;
 532	}
 533
 534	rd->remote_ip = *ip;
 535	rd->remote_port = port;
 
 536	rd->remote_vni = vni;
 537	rd->remote_ifindex = ifindex;
 538
 539	list_add_tail_rcu(&rd->list, &f->remotes);
 540
 541	*rdp = rd;
 542	return 1;
 543}
 544
 545static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
 546					  unsigned int off,
 547					  struct vxlanhdr *vh, size_t hdrlen,
 548					  __be32 vni_field,
 549					  struct gro_remcsum *grc,
 550					  bool nopartial)
 551{
 552	size_t start, offset;
 553
 554	if (skb->remcsum_offload)
 555		return vh;
 556
 557	if (!NAPI_GRO_CB(skb)->csum_valid)
 558		return NULL;
 559
 560	start = vxlan_rco_start(vni_field);
 561	offset = start + vxlan_rco_offset(vni_field);
 562
 563	vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
 564				     start, offset, grc, nopartial);
 565
 566	skb->remcsum_offload = 1;
 567
 568	return vh;
 569}
 570
 571static struct sk_buff **vxlan_gro_receive(struct sock *sk,
 572					  struct sk_buff **head,
 573					  struct sk_buff *skb)
 574{
 575	struct sk_buff *p, **pp = NULL;
 
 576	struct vxlanhdr *vh, *vh2;
 577	unsigned int hlen, off_vx;
 578	int flush = 1;
 579	struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
 580	__be32 flags;
 581	struct gro_remcsum grc;
 582
 583	skb_gro_remcsum_init(&grc);
 584
 585	off_vx = skb_gro_offset(skb);
 586	hlen = off_vx + sizeof(*vh);
 587	vh   = skb_gro_header_fast(skb, off_vx);
 588	if (skb_gro_header_hard(skb, hlen)) {
 589		vh = skb_gro_header_slow(skb, hlen, off_vx);
 590		if (unlikely(!vh))
 591			goto out;
 592	}
 593
 594	skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
 595
 596	flags = vh->vx_flags;
 597
 598	if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
 599		vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
 600				       vh->vx_vni, &grc,
 601				       !!(vs->flags &
 602					  VXLAN_F_REMCSUM_NOPARTIAL));
 603
 604		if (!vh)
 605			goto out;
 606	}
 607
 608	skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
 609
 610	for (p = *head; p; p = p->next) {
 611		if (!NAPI_GRO_CB(p)->same_flow)
 612			continue;
 613
 614		vh2 = (struct vxlanhdr *)(p->data + off_vx);
 615		if (vh->vx_flags != vh2->vx_flags ||
 616		    vh->vx_vni != vh2->vx_vni) {
 617			NAPI_GRO_CB(p)->same_flow = 0;
 618			continue;
 619		}
 620	}
 621
 622	pp = call_gro_receive(eth_gro_receive, head, skb);
 623	flush = 0;
 624
 625out:
 626	skb_gro_remcsum_cleanup(skb, &grc);
 627	skb->remcsum_offload = 0;
 628	NAPI_GRO_CB(skb)->flush |= flush;
 629
 630	return pp;
 631}
 632
 633static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
 634{
 635	/* Sets 'skb->inner_mac_header' since we are always called with
 636	 * 'skb->encapsulation' set.
 637	 */
 638	return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
 639}
 640
 641/* Add new entry to forwarding table -- assumes lock held */
 642static int vxlan_fdb_create(struct vxlan_dev *vxlan,
 643			    const u8 *mac, union vxlan_addr *ip,
 644			    __u16 state, __u16 flags,
 645			    __be16 port, __be32 src_vni, __be32 vni,
 646			    __u32 ifindex, __u8 ndm_flags)
 647{
 648	struct vxlan_rdst *rd = NULL;
 649	struct vxlan_fdb *f;
 650	int notify = 0;
 651	int rc;
 652
 653	f = __vxlan_find_mac(vxlan, mac, src_vni);
 654	if (f) {
 655		if (flags & NLM_F_EXCL) {
 656			netdev_dbg(vxlan->dev,
 657				   "lost race to create %pM\n", mac);
 658			return -EEXIST;
 659		}
 660		if (f->state != state) {
 661			f->state = state;
 662			f->updated = jiffies;
 663			notify = 1;
 664		}
 665		if (f->flags != ndm_flags) {
 666			f->flags = ndm_flags;
 667			f->updated = jiffies;
 668			notify = 1;
 669		}
 670		if ((flags & NLM_F_REPLACE)) {
 671			/* Only change unicasts */
 672			if (!(is_multicast_ether_addr(f->eth_addr) ||
 673			     is_zero_ether_addr(f->eth_addr))) {
 674				notify |= vxlan_fdb_replace(f, ip, port, vni,
 675							   ifindex);
 676			} else
 677				return -EOPNOTSUPP;
 678		}
 679		if ((flags & NLM_F_APPEND) &&
 680		    (is_multicast_ether_addr(f->eth_addr) ||
 681		     is_zero_ether_addr(f->eth_addr))) {
 682			rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
 683
 684			if (rc < 0)
 685				return rc;
 686			notify |= rc;
 687		}
 688	} else {
 689		if (!(flags & NLM_F_CREATE))
 690			return -ENOENT;
 691
 692		if (vxlan->cfg.addrmax &&
 693		    vxlan->addrcnt >= vxlan->cfg.addrmax)
 694			return -ENOSPC;
 695
 696		/* Disallow replace to add a multicast entry */
 697		if ((flags & NLM_F_REPLACE) &&
 698		    (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
 699			return -EOPNOTSUPP;
 700
 701		netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
 702		f = kmalloc(sizeof(*f), GFP_ATOMIC);
 703		if (!f)
 704			return -ENOMEM;
 705
 706		notify = 1;
 707		f->state = state;
 708		f->flags = ndm_flags;
 709		f->updated = f->used = jiffies;
 710		f->vni = src_vni;
 711		INIT_LIST_HEAD(&f->remotes);
 712		memcpy(f->eth_addr, mac, ETH_ALEN);
 713
 714		rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
 715		if (rc < 0) {
 716			kfree(f);
 717			return rc;
 718		}
 719
 720		++vxlan->addrcnt;
 721		hlist_add_head_rcu(&f->hlist,
 722				   vxlan_fdb_head(vxlan, mac, src_vni));
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 723	}
 724
 725	if (notify) {
 726		if (rd == NULL)
 727			rd = first_remote_rtnl(f);
 728		vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
 729	}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 730
 731	return 0;
 
 
 
 
 732}
 733
 734static void vxlan_fdb_free(struct rcu_head *head)
 735{
 736	struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
 737	struct vxlan_rdst *rd, *nd;
 
 
 
 
 
 
 
 
 738
 739	list_for_each_entry_safe(rd, nd, &f->remotes, list) {
 740		dst_cache_destroy(&rd->dst_cache);
 741		kfree(rd);
 742	}
 743	kfree(f);
 744}
 745
 746static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
 
 
 
 
 
 
 
 
 747{
 748	netdev_dbg(vxlan->dev,
 749		    "delete %pM\n", f->eth_addr);
 
 750
 751	--vxlan->addrcnt;
 752	vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
 
 
 
 
 
 
 
 
 753
 754	hlist_del_rcu(&f->hlist);
 
 755	call_rcu(&f->rcu, vxlan_fdb_free);
 756}
 757
 758static void vxlan_dst_free(struct rcu_head *head)
 759{
 760	struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
 761
 762	dst_cache_destroy(&rd->dst_cache);
 763	kfree(rd);
 764}
 765
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 766static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
 767				  struct vxlan_rdst *rd)
 768{
 769	list_del_rcu(&rd->list);
 770	vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
 771	call_rcu(&rd->rcu, vxlan_dst_free);
 772}
 773
 774static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
 775			   union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
 776			   __be32 *vni, u32 *ifindex)
 777{
 778	struct net *net = dev_net(vxlan->dev);
 779	int err;
 780
 
 
 
 
 781	if (tb[NDA_DST]) {
 782		err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
 783		if (err)
 784			return err;
 785	} else {
 786		union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
 
 787		if (remote->sa.sa_family == AF_INET) {
 788			ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
 789			ip->sa.sa_family = AF_INET;
 790#if IS_ENABLED(CONFIG_IPV6)
 791		} else {
 792			ip->sin6.sin6_addr = in6addr_any;
 793			ip->sa.sa_family = AF_INET6;
 794#endif
 795		}
 796	}
 797
 798	if (tb[NDA_PORT]) {
 799		if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
 800			return -EINVAL;
 801		*port = nla_get_be16(tb[NDA_PORT]);
 802	} else {
 803		*port = vxlan->cfg.dst_port;
 804	}
 805
 806	if (tb[NDA_VNI]) {
 807		if (nla_len(tb[NDA_VNI]) != sizeof(u32))
 808			return -EINVAL;
 809		*vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
 810	} else {
 811		*vni = vxlan->default_dst.remote_vni;
 812	}
 813
 814	if (tb[NDA_SRC_VNI]) {
 815		if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
 816			return -EINVAL;
 817		*src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
 818	} else {
 819		*src_vni = vxlan->default_dst.remote_vni;
 820	}
 821
 822	if (tb[NDA_IFINDEX]) {
 823		struct net_device *tdev;
 824
 825		if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
 826			return -EINVAL;
 827		*ifindex = nla_get_u32(tb[NDA_IFINDEX]);
 828		tdev = __dev_get_by_index(net, *ifindex);
 829		if (!tdev)
 830			return -EADDRNOTAVAIL;
 831	} else {
 832		*ifindex = 0;
 833	}
 834
 
 
 
 
 
 835	return 0;
 836}
 837
 838/* Add static entry (via netlink) */
 839static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
 840			 struct net_device *dev,
 841			 const unsigned char *addr, u16 vid, u16 flags)
 
 842{
 843	struct vxlan_dev *vxlan = netdev_priv(dev);
 844	/* struct net *net = dev_net(vxlan->dev); */
 845	union vxlan_addr ip;
 846	__be16 port;
 847	__be32 src_vni, vni;
 848	u32 ifindex;
 
 849	int err;
 850
 851	if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
 852		pr_info("RTM_NEWNEIGH with invalid state %#x\n",
 853			ndm->ndm_state);
 854		return -EINVAL;
 855	}
 856
 857	if (tb[NDA_DST] == NULL)
 858		return -EINVAL;
 859
 860	err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
 
 861	if (err)
 862		return err;
 863
 864	if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
 865		return -EAFNOSUPPORT;
 866
 867	spin_lock_bh(&vxlan->hash_lock);
 868	err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
 869			       port, src_vni, vni, ifindex, ndm->ndm_flags);
 870	spin_unlock_bh(&vxlan->hash_lock);
 
 
 
 871
 872	return err;
 873}
 874
 875static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
 876			      const unsigned char *addr, union vxlan_addr ip,
 877			      __be16 port, __be32 src_vni, __be32 vni,
 878			      u32 ifindex, u16 vid)
 879{
 880	struct vxlan_fdb *f;
 881	struct vxlan_rdst *rd = NULL;
 
 882	int err = -ENOENT;
 883
 884	f = vxlan_find_mac(vxlan, addr, src_vni);
 885	if (!f)
 886		return err;
 887
 888	if (!vxlan_addr_any(&ip)) {
 889		rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
 890		if (!rd)
 891			goto out;
 892	}
 893
 894	/* remove a destination if it's not the only one on the list,
 895	 * otherwise destroy the fdb entry
 896	 */
 897	if (rd && !list_is_singular(&f->remotes)) {
 898		vxlan_fdb_dst_destroy(vxlan, f, rd);
 899		goto out;
 900	}
 901
 902	vxlan_fdb_destroy(vxlan, f);
 903
 904out:
 905	return 0;
 906}
 907
 908/* Delete entry (via netlink) */
 909static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
 910			    struct net_device *dev,
 911			    const unsigned char *addr, u16 vid)
 912{
 913	struct vxlan_dev *vxlan = netdev_priv(dev);
 914	union vxlan_addr ip;
 915	__be32 src_vni, vni;
 
 
 916	__be16 port;
 917	u32 ifindex;
 918	int err;
 919
 920	err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
 
 921	if (err)
 922		return err;
 923
 924	spin_lock_bh(&vxlan->hash_lock);
 
 925	err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
 926				 vid);
 927	spin_unlock_bh(&vxlan->hash_lock);
 928
 929	return err;
 930}
 931
 932/* Dump forwarding table */
 933static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
 934			  struct net_device *dev,
 935			  struct net_device *filter_dev, int *idx)
 936{
 937	struct vxlan_dev *vxlan = netdev_priv(dev);
 938	unsigned int h;
 939	int err = 0;
 940
 941	for (h = 0; h < FDB_HASH_SIZE; ++h) {
 942		struct vxlan_fdb *f;
 943
 
 944		hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
 945			struct vxlan_rdst *rd;
 946
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 947			list_for_each_entry_rcu(rd, &f->remotes, list) {
 948				if (*idx < cb->args[2])
 949					goto skip;
 950
 951				err = vxlan_fdb_info(skb, vxlan, f,
 952						     NETLINK_CB(cb->skb).portid,
 953						     cb->nlh->nlmsg_seq,
 954						     RTM_NEWNEIGH,
 955						     NLM_F_MULTI, rd);
 956				if (err < 0)
 
 957					goto out;
 
 958skip:
 959				*idx += 1;
 960			}
 961		}
 
 962	}
 963out:
 964	return err;
 965}
 966
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 967/* Watch incoming packets to learn mapping between Ethernet address
 968 * and Tunnel endpoint.
 969 * Return true if packet is bogus and should be dropped.
 970 */
 971static bool vxlan_snoop(struct net_device *dev,
 972			union vxlan_addr *src_ip, const u8 *src_mac,
 973			u32 src_ifindex, __be32 vni)
 974{
 975	struct vxlan_dev *vxlan = netdev_priv(dev);
 976	struct vxlan_fdb *f;
 977	u32 ifindex = 0;
 978
 979#if IS_ENABLED(CONFIG_IPV6)
 980	if (src_ip->sa.sa_family == AF_INET6 &&
 981	    (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
 982		ifindex = src_ifindex;
 983#endif
 984
 985	f = vxlan_find_mac(vxlan, src_mac, vni);
 986	if (likely(f)) {
 987		struct vxlan_rdst *rdst = first_remote_rcu(f);
 988
 989		if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
 990			   rdst->remote_ifindex == ifindex))
 991			return false;
 992
 993		/* Don't migrate static entries, drop packets */
 994		if (f->state & (NUD_PERMANENT | NUD_NOARP))
 995			return true;
 996
 
 
 
 
 997		if (net_ratelimit())
 998			netdev_info(dev,
 999				    "%pM migrated from %pIS to %pIS\n",
1000				    src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1001
1002		rdst->remote_ip = *src_ip;
1003		f->updated = jiffies;
1004		vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
1005	} else {
 
 
1006		/* learned new entry */
1007		spin_lock(&vxlan->hash_lock);
1008
1009		/* close off race between vxlan_flush and incoming packets */
1010		if (netif_running(dev))
1011			vxlan_fdb_create(vxlan, src_mac, src_ip,
1012					 NUD_REACHABLE,
1013					 NLM_F_EXCL|NLM_F_CREATE,
1014					 vxlan->cfg.dst_port,
1015					 vni,
1016					 vxlan->default_dst.remote_vni,
1017					 ifindex, NTF_SELF);
1018		spin_unlock(&vxlan->hash_lock);
1019	}
1020
1021	return false;
1022}
1023
1024/* See if multicast group is already in use by other ID */
1025static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1026{
1027	struct vxlan_dev *vxlan;
1028	struct vxlan_sock *sock4;
1029#if IS_ENABLED(CONFIG_IPV6)
1030	struct vxlan_sock *sock6;
1031#endif
1032	unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1033
1034	sock4 = rtnl_dereference(dev->vn4_sock);
1035
1036	/* The vxlan_sock is only used by dev, leaving group has
1037	 * no effect on other vxlan devices.
1038	 */
1039	if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
1040		return false;
1041#if IS_ENABLED(CONFIG_IPV6)
1042	sock6 = rtnl_dereference(dev->vn6_sock);
1043	if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
1044		return false;
1045#endif
1046
1047	list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1048		if (!netif_running(vxlan->dev) || vxlan == dev)
1049			continue;
1050
1051		if (family == AF_INET &&
1052		    rtnl_dereference(vxlan->vn4_sock) != sock4)
1053			continue;
1054#if IS_ENABLED(CONFIG_IPV6)
1055		if (family == AF_INET6 &&
1056		    rtnl_dereference(vxlan->vn6_sock) != sock6)
1057			continue;
1058#endif
1059
1060		if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1061				      &dev->default_dst.remote_ip))
1062			continue;
1063
1064		if (vxlan->default_dst.remote_ifindex !=
1065		    dev->default_dst.remote_ifindex)
1066			continue;
1067
1068		return true;
1069	}
1070
1071	return false;
1072}
1073
1074static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1075{
1076	struct vxlan_net *vn;
1077
1078	if (!vs)
1079		return false;
1080	if (!refcount_dec_and_test(&vs->refcnt))
1081		return false;
1082
1083	vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1084	spin_lock(&vn->sock_lock);
1085	hlist_del_rcu(&vs->hlist);
1086	udp_tunnel_notify_del_rx_port(vs->sock,
1087				      (vs->flags & VXLAN_F_GPE) ?
1088				      UDP_TUNNEL_TYPE_VXLAN_GPE :
1089				      UDP_TUNNEL_TYPE_VXLAN);
1090	spin_unlock(&vn->sock_lock);
1091
1092	return true;
1093}
1094
1095static void vxlan_sock_release(struct vxlan_dev *vxlan)
1096{
1097	struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1098#if IS_ENABLED(CONFIG_IPV6)
1099	struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1100
1101	RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1102#endif
1103
1104	RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1105	synchronize_net();
1106
1107	vxlan_vs_del_dev(vxlan);
1108
1109	if (__vxlan_sock_release_prep(sock4)) {
1110		udp_tunnel_sock_release(sock4->sock);
1111		kfree(sock4);
1112	}
1113
1114#if IS_ENABLED(CONFIG_IPV6)
1115	if (__vxlan_sock_release_prep(sock6)) {
1116		udp_tunnel_sock_release(sock6->sock);
1117		kfree(sock6);
1118	}
1119#endif
1120}
1121
1122/* Update multicast group membership when first VNI on
1123 * multicast address is brought up
1124 */
1125static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1126{
1127	struct sock *sk;
1128	union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1129	int ifindex = vxlan->default_dst.remote_ifindex;
1130	int ret = -EINVAL;
1131
1132	if (ip->sa.sa_family == AF_INET) {
1133		struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1134		struct ip_mreqn mreq = {
1135			.imr_multiaddr.s_addr	= ip->sin.sin_addr.s_addr,
1136			.imr_ifindex		= ifindex,
1137		};
1138
1139		sk = sock4->sock->sk;
1140		lock_sock(sk);
1141		ret = ip_mc_join_group(sk, &mreq);
1142		release_sock(sk);
1143#if IS_ENABLED(CONFIG_IPV6)
1144	} else {
1145		struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1146
1147		sk = sock6->sock->sk;
1148		lock_sock(sk);
1149		ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1150						   &ip->sin6.sin6_addr);
1151		release_sock(sk);
1152#endif
1153	}
1154
1155	return ret;
1156}
1157
1158/* Inverse of vxlan_igmp_join when last VNI is brought down */
1159static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1160{
1161	struct sock *sk;
1162	union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1163	int ifindex = vxlan->default_dst.remote_ifindex;
1164	int ret = -EINVAL;
1165
1166	if (ip->sa.sa_family == AF_INET) {
1167		struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1168		struct ip_mreqn mreq = {
1169			.imr_multiaddr.s_addr	= ip->sin.sin_addr.s_addr,
1170			.imr_ifindex		= ifindex,
1171		};
1172
1173		sk = sock4->sock->sk;
1174		lock_sock(sk);
1175		ret = ip_mc_leave_group(sk, &mreq);
1176		release_sock(sk);
1177#if IS_ENABLED(CONFIG_IPV6)
1178	} else {
1179		struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1180
1181		sk = sock6->sock->sk;
1182		lock_sock(sk);
1183		ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1184						   &ip->sin6.sin6_addr);
1185		release_sock(sk);
1186#endif
1187	}
1188
1189	return ret;
1190}
1191
1192static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1193			  struct sk_buff *skb, u32 vxflags)
1194{
1195	size_t start, offset;
1196
1197	if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1198		goto out;
1199
1200	start = vxlan_rco_start(unparsed->vx_vni);
1201	offset = start + vxlan_rco_offset(unparsed->vx_vni);
1202
1203	if (!pskb_may_pull(skb, offset + sizeof(u16)))
1204		return false;
1205
1206	skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1207			    !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1208out:
1209	unparsed->vx_flags &= ~VXLAN_HF_RCO;
1210	unparsed->vx_vni &= VXLAN_VNI_MASK;
1211	return true;
1212}
1213
1214static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1215				struct sk_buff *skb, u32 vxflags,
1216				struct vxlan_metadata *md)
1217{
1218	struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1219	struct metadata_dst *tun_dst;
1220
1221	if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1222		goto out;
1223
1224	md->gbp = ntohs(gbp->policy_id);
1225
1226	tun_dst = (struct metadata_dst *)skb_dst(skb);
1227	if (tun_dst) {
1228		tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1229		tun_dst->u.tun_info.options_len = sizeof(*md);
1230	}
1231	if (gbp->dont_learn)
1232		md->gbp |= VXLAN_GBP_DONT_LEARN;
1233
1234	if (gbp->policy_applied)
1235		md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1236
1237	/* In flow-based mode, GBP is carried in dst_metadata */
1238	if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1239		skb->mark = md->gbp;
1240out:
1241	unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1242}
1243
1244static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1245				__be16 *protocol,
1246				struct sk_buff *skb, u32 vxflags)
1247{
1248	struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1249
1250	/* Need to have Next Protocol set for interfaces in GPE mode. */
1251	if (!gpe->np_applied)
1252		return false;
1253	/* "The initial version is 0. If a receiver does not support the
1254	 * version indicated it MUST drop the packet.
1255	 */
1256	if (gpe->version != 0)
1257		return false;
1258	/* "When the O bit is set to 1, the packet is an OAM packet and OAM
1259	 * processing MUST occur." However, we don't implement OAM
1260	 * processing, thus drop the packet.
1261	 */
1262	if (gpe->oam_flag)
1263		return false;
1264
1265	*protocol = tun_p_to_eth_p(gpe->next_protocol);
1266	if (!*protocol)
1267		return false;
1268
1269	unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1270	return true;
1271}
1272
1273static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1274			  struct vxlan_sock *vs,
1275			  struct sk_buff *skb, __be32 vni)
1276{
1277	union vxlan_addr saddr;
1278	u32 ifindex = skb->dev->ifindex;
1279
1280	skb_reset_mac_header(skb);
1281	skb->protocol = eth_type_trans(skb, vxlan->dev);
1282	skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1283
1284	/* Ignore packet loops (and multicast echo) */
1285	if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1286		return false;
1287
1288	/* Get address from the outer IP header */
1289	if (vxlan_get_sk_family(vs) == AF_INET) {
1290		saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1291		saddr.sa.sa_family = AF_INET;
1292#if IS_ENABLED(CONFIG_IPV6)
1293	} else {
1294		saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1295		saddr.sa.sa_family = AF_INET6;
1296#endif
1297	}
1298
1299	if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1300	    vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1301		return false;
1302
1303	return true;
1304}
1305
1306static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1307				  struct sk_buff *skb)
1308{
1309	int err = 0;
1310
1311	if (vxlan_get_sk_family(vs) == AF_INET)
1312		err = IP_ECN_decapsulate(oiph, skb);
1313#if IS_ENABLED(CONFIG_IPV6)
1314	else
1315		err = IP6_ECN_decapsulate(oiph, skb);
1316#endif
1317
1318	if (unlikely(err) && log_ecn_error) {
1319		if (vxlan_get_sk_family(vs) == AF_INET)
1320			net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1321					     &((struct iphdr *)oiph)->saddr,
1322					     ((struct iphdr *)oiph)->tos);
1323		else
1324			net_info_ratelimited("non-ECT from %pI6\n",
1325					     &((struct ipv6hdr *)oiph)->saddr);
1326	}
1327	return err <= 1;
1328}
1329
1330/* Callback from net/ipv4/udp.c to receive packets */
1331static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1332{
1333	struct pcpu_sw_netstats *stats;
1334	struct vxlan_dev *vxlan;
1335	struct vxlan_sock *vs;
1336	struct vxlanhdr unparsed;
1337	struct vxlan_metadata _md;
1338	struct vxlan_metadata *md = &_md;
1339	__be16 protocol = htons(ETH_P_TEB);
1340	bool raw_proto = false;
1341	void *oiph;
1342	__be32 vni = 0;
1343
1344	/* Need UDP and VXLAN header to be present */
1345	if (!pskb_may_pull(skb, VXLAN_HLEN))
1346		goto drop;
1347
1348	unparsed = *vxlan_hdr(skb);
1349	/* VNI flag always required to be set */
1350	if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1351		netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1352			   ntohl(vxlan_hdr(skb)->vx_flags),
1353			   ntohl(vxlan_hdr(skb)->vx_vni));
1354		/* Return non vxlan pkt */
1355		goto drop;
1356	}
1357	unparsed.vx_flags &= ~VXLAN_HF_VNI;
1358	unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1359
1360	vs = rcu_dereference_sk_user_data(sk);
1361	if (!vs)
1362		goto drop;
1363
1364	vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1365
1366	vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1367	if (!vxlan)
1368		goto drop;
1369
1370	/* For backwards compatibility, only allow reserved fields to be
1371	 * used by VXLAN extensions if explicitly requested.
1372	 */
1373	if (vs->flags & VXLAN_F_GPE) {
1374		if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1375			goto drop;
1376		raw_proto = true;
1377	}
1378
1379	if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1380				   !net_eq(vxlan->net, dev_net(vxlan->dev))))
1381			goto drop;
1382
1383	if (vxlan_collect_metadata(vs)) {
1384		struct metadata_dst *tun_dst;
1385
1386		tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1387					 key32_to_tunnel_id(vni), sizeof(*md));
1388
1389		if (!tun_dst)
1390			goto drop;
1391
1392		md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1393
1394		skb_dst_set(skb, (struct dst_entry *)tun_dst);
1395	} else {
1396		memset(md, 0, sizeof(*md));
1397	}
1398
1399	if (vs->flags & VXLAN_F_REMCSUM_RX)
1400		if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1401			goto drop;
1402	if (vs->flags & VXLAN_F_GBP)
1403		vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1404	/* Note that GBP and GPE can never be active together. This is
1405	 * ensured in vxlan_dev_configure.
1406	 */
1407
1408	if (unparsed.vx_flags || unparsed.vx_vni) {
1409		/* If there are any unprocessed flags remaining treat
1410		 * this as a malformed packet. This behavior diverges from
1411		 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1412		 * in reserved fields are to be ignored. The approach here
1413		 * maintains compatibility with previous stack code, and also
1414		 * is more robust and provides a little more security in
1415		 * adding extensions to VXLAN.
1416		 */
1417		goto drop;
1418	}
1419
1420	if (!raw_proto) {
1421		if (!vxlan_set_mac(vxlan, vs, skb, vni))
1422			goto drop;
1423	} else {
1424		skb_reset_mac_header(skb);
1425		skb->dev = vxlan->dev;
1426		skb->pkt_type = PACKET_HOST;
1427	}
1428
1429	oiph = skb_network_header(skb);
1430	skb_reset_network_header(skb);
1431
1432	if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1433		++vxlan->dev->stats.rx_frame_errors;
1434		++vxlan->dev->stats.rx_errors;
1435		goto drop;
1436	}
1437
 
 
 
 
 
 
 
 
1438	stats = this_cpu_ptr(vxlan->dev->tstats);
1439	u64_stats_update_begin(&stats->syncp);
1440	stats->rx_packets++;
1441	stats->rx_bytes += skb->len;
1442	u64_stats_update_end(&stats->syncp);
1443
1444	gro_cells_receive(&vxlan->gro_cells, skb);
 
 
 
1445	return 0;
1446
1447drop:
1448	/* Consume bad packet */
1449	kfree_skb(skb);
1450	return 0;
1451}
1452
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1453static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1454{
1455	struct vxlan_dev *vxlan = netdev_priv(dev);
1456	struct arphdr *parp;
1457	u8 *arpptr, *sha;
1458	__be32 sip, tip;
1459	struct neighbour *n;
1460
1461	if (dev->flags & IFF_NOARP)
1462		goto out;
1463
1464	if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1465		dev->stats.tx_dropped++;
1466		goto out;
1467	}
1468	parp = arp_hdr(skb);
1469
1470	if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1471	     parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1472	    parp->ar_pro != htons(ETH_P_IP) ||
1473	    parp->ar_op != htons(ARPOP_REQUEST) ||
1474	    parp->ar_hln != dev->addr_len ||
1475	    parp->ar_pln != 4)
1476		goto out;
1477	arpptr = (u8 *)parp + sizeof(struct arphdr);
1478	sha = arpptr;
1479	arpptr += dev->addr_len;	/* sha */
1480	memcpy(&sip, arpptr, sizeof(sip));
1481	arpptr += sizeof(sip);
1482	arpptr += dev->addr_len;	/* tha */
1483	memcpy(&tip, arpptr, sizeof(tip));
1484
1485	if (ipv4_is_loopback(tip) ||
1486	    ipv4_is_multicast(tip))
1487		goto out;
1488
1489	n = neigh_lookup(&arp_tbl, &tip, dev);
1490
1491	if (n) {
1492		struct vxlan_fdb *f;
1493		struct sk_buff	*reply;
1494
1495		if (!(n->nud_state & NUD_CONNECTED)) {
1496			neigh_release(n);
1497			goto out;
1498		}
1499
1500		f = vxlan_find_mac(vxlan, n->ha, vni);
1501		if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1502			/* bridge-local neighbor */
1503			neigh_release(n);
1504			goto out;
1505		}
1506
1507		reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1508				n->ha, sha);
1509
1510		neigh_release(n);
1511
1512		if (reply == NULL)
1513			goto out;
1514
1515		skb_reset_mac_header(reply);
1516		__skb_pull(reply, skb_network_offset(reply));
1517		reply->ip_summed = CHECKSUM_UNNECESSARY;
1518		reply->pkt_type = PACKET_HOST;
1519
1520		if (netif_rx_ni(reply) == NET_RX_DROP)
1521			dev->stats.rx_dropped++;
1522	} else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1523		union vxlan_addr ipa = {
1524			.sin.sin_addr.s_addr = tip,
1525			.sin.sin_family = AF_INET,
1526		};
1527
1528		vxlan_ip_miss(dev, &ipa);
1529	}
1530out:
1531	consume_skb(skb);
1532	return NETDEV_TX_OK;
1533}
1534
1535#if IS_ENABLED(CONFIG_IPV6)
1536static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1537	struct neighbour *n, bool isrouter)
1538{
1539	struct net_device *dev = request->dev;
1540	struct sk_buff *reply;
1541	struct nd_msg *ns, *na;
1542	struct ipv6hdr *pip6;
1543	u8 *daddr;
1544	int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1545	int ns_olen;
1546	int i, len;
1547
1548	if (dev == NULL || !pskb_may_pull(request, request->len))
1549		return NULL;
1550
1551	len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1552		sizeof(*na) + na_olen + dev->needed_tailroom;
1553	reply = alloc_skb(len, GFP_ATOMIC);
1554	if (reply == NULL)
1555		return NULL;
1556
1557	reply->protocol = htons(ETH_P_IPV6);
1558	reply->dev = dev;
1559	skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1560	skb_push(reply, sizeof(struct ethhdr));
1561	skb_reset_mac_header(reply);
1562
1563	ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
1564
1565	daddr = eth_hdr(request)->h_source;
1566	ns_olen = request->len - skb_network_offset(request) -
1567		sizeof(struct ipv6hdr) - sizeof(*ns);
1568	for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
 
 
 
 
1569		if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1570			daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1571			break;
1572		}
1573	}
1574
1575	/* Ethernet header */
1576	ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1577	ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1578	eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1579	reply->protocol = htons(ETH_P_IPV6);
1580
1581	skb_pull(reply, sizeof(struct ethhdr));
1582	skb_reset_network_header(reply);
1583	skb_put(reply, sizeof(struct ipv6hdr));
1584
1585	/* IPv6 header */
1586
1587	pip6 = ipv6_hdr(reply);
1588	memset(pip6, 0, sizeof(struct ipv6hdr));
1589	pip6->version = 6;
1590	pip6->priority = ipv6_hdr(request)->priority;
1591	pip6->nexthdr = IPPROTO_ICMPV6;
1592	pip6->hop_limit = 255;
1593	pip6->daddr = ipv6_hdr(request)->saddr;
1594	pip6->saddr = *(struct in6_addr *)n->primary_key;
1595
1596	skb_pull(reply, sizeof(struct ipv6hdr));
1597	skb_reset_transport_header(reply);
1598
1599	/* Neighbor Advertisement */
1600	na = skb_put_zero(reply, sizeof(*na) + na_olen);
1601	na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1602	na->icmph.icmp6_router = isrouter;
1603	na->icmph.icmp6_override = 1;
1604	na->icmph.icmp6_solicited = 1;
1605	na->target = ns->target;
1606	ether_addr_copy(&na->opt[2], n->ha);
1607	na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1608	na->opt[1] = na_olen >> 3;
1609
1610	na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1611		&pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1612		csum_partial(na, sizeof(*na)+na_olen, 0));
1613
1614	pip6->payload_len = htons(sizeof(*na)+na_olen);
1615
1616	skb_push(reply, sizeof(struct ipv6hdr));
1617
1618	reply->ip_summed = CHECKSUM_UNNECESSARY;
1619
1620	return reply;
1621}
1622
1623static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1624{
1625	struct vxlan_dev *vxlan = netdev_priv(dev);
1626	const struct in6_addr *daddr;
1627	const struct ipv6hdr *iphdr;
1628	struct inet6_dev *in6_dev;
1629	struct neighbour *n;
1630	struct nd_msg *msg;
1631
1632	in6_dev = __in6_dev_get(dev);
1633	if (!in6_dev)
1634		goto out;
1635
1636	iphdr = ipv6_hdr(skb);
1637	daddr = &iphdr->daddr;
1638	msg = (struct nd_msg *)(iphdr + 1);
1639
1640	if (ipv6_addr_loopback(daddr) ||
1641	    ipv6_addr_is_multicast(&msg->target))
1642		goto out;
1643
1644	n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
1645
1646	if (n) {
1647		struct vxlan_fdb *f;
1648		struct sk_buff *reply;
1649
1650		if (!(n->nud_state & NUD_CONNECTED)) {
1651			neigh_release(n);
1652			goto out;
1653		}
1654
1655		f = vxlan_find_mac(vxlan, n->ha, vni);
1656		if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1657			/* bridge-local neighbor */
1658			neigh_release(n);
1659			goto out;
1660		}
1661
1662		reply = vxlan_na_create(skb, n,
1663					!!(f ? f->flags & NTF_ROUTER : 0));
1664
1665		neigh_release(n);
1666
1667		if (reply == NULL)
1668			goto out;
1669
1670		if (netif_rx_ni(reply) == NET_RX_DROP)
1671			dev->stats.rx_dropped++;
1672
1673	} else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1674		union vxlan_addr ipa = {
1675			.sin6.sin6_addr = msg->target,
1676			.sin6.sin6_family = AF_INET6,
1677		};
1678
1679		vxlan_ip_miss(dev, &ipa);
1680	}
1681
1682out:
1683	consume_skb(skb);
1684	return NETDEV_TX_OK;
1685}
1686#endif
1687
1688static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1689{
1690	struct vxlan_dev *vxlan = netdev_priv(dev);
1691	struct neighbour *n;
1692
1693	if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1694		return false;
1695
1696	n = NULL;
1697	switch (ntohs(eth_hdr(skb)->h_proto)) {
1698	case ETH_P_IP:
1699	{
1700		struct iphdr *pip;
1701
1702		if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1703			return false;
1704		pip = ip_hdr(skb);
1705		n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1706		if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
1707			union vxlan_addr ipa = {
1708				.sin.sin_addr.s_addr = pip->daddr,
1709				.sin.sin_family = AF_INET,
1710			};
1711
1712			vxlan_ip_miss(dev, &ipa);
1713			return false;
1714		}
1715
1716		break;
1717	}
1718#if IS_ENABLED(CONFIG_IPV6)
1719	case ETH_P_IPV6:
1720	{
1721		struct ipv6hdr *pip6;
1722
1723		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1724			return false;
1725		pip6 = ipv6_hdr(skb);
1726		n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1727		if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
1728			union vxlan_addr ipa = {
1729				.sin6.sin6_addr = pip6->daddr,
1730				.sin6.sin6_family = AF_INET6,
1731			};
1732
1733			vxlan_ip_miss(dev, &ipa);
1734			return false;
1735		}
1736
1737		break;
1738	}
1739#endif
1740	default:
1741		return false;
1742	}
1743
1744	if (n) {
1745		bool diff;
1746
1747		diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1748		if (diff) {
1749			memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1750				dev->addr_len);
1751			memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1752		}
1753		neigh_release(n);
1754		return diff;
1755	}
1756
1757	return false;
1758}
1759
1760static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
1761				struct vxlan_metadata *md)
1762{
1763	struct vxlanhdr_gbp *gbp;
1764
1765	if (!md->gbp)
1766		return;
1767
1768	gbp = (struct vxlanhdr_gbp *)vxh;
1769	vxh->vx_flags |= VXLAN_HF_GBP;
1770
1771	if (md->gbp & VXLAN_GBP_DONT_LEARN)
1772		gbp->dont_learn = 1;
1773
1774	if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1775		gbp->policy_applied = 1;
1776
1777	gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1778}
1779
1780static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1781			       __be16 protocol)
1782{
1783	struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1784
1785	gpe->np_applied = 1;
1786	gpe->next_protocol = tun_p_from_eth_p(protocol);
1787	if (!gpe->next_protocol)
1788		return -EPFNOSUPPORT;
1789	return 0;
1790}
1791
1792static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1793			   int iphdr_len, __be32 vni,
1794			   struct vxlan_metadata *md, u32 vxflags,
1795			   bool udp_sum)
1796{
1797	struct vxlanhdr *vxh;
1798	int min_headroom;
1799	int err;
1800	int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1801	__be16 inner_protocol = htons(ETH_P_TEB);
1802
1803	if ((vxflags & VXLAN_F_REMCSUM_TX) &&
1804	    skb->ip_summed == CHECKSUM_PARTIAL) {
1805		int csum_start = skb_checksum_start_offset(skb);
1806
1807		if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1808		    !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1809		    (skb->csum_offset == offsetof(struct udphdr, check) ||
1810		     skb->csum_offset == offsetof(struct tcphdr, check)))
1811			type |= SKB_GSO_TUNNEL_REMCSUM;
1812	}
1813
1814	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1815			+ VXLAN_HLEN + iphdr_len;
1816
1817	/* Need space for new headers (invalidates iph ptr) */
1818	err = skb_cow_head(skb, min_headroom);
1819	if (unlikely(err))
1820		return err;
1821
1822	err = iptunnel_handle_offloads(skb, type);
1823	if (err)
1824		return err;
1825
1826	vxh = __skb_push(skb, sizeof(*vxh));
1827	vxh->vx_flags = VXLAN_HF_VNI;
1828	vxh->vx_vni = vxlan_vni_field(vni);
1829
1830	if (type & SKB_GSO_TUNNEL_REMCSUM) {
1831		unsigned int start;
1832
1833		start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1834		vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1835		vxh->vx_flags |= VXLAN_HF_RCO;
1836
1837		if (!skb_is_gso(skb)) {
1838			skb->ip_summed = CHECKSUM_NONE;
1839			skb->encapsulation = 0;
1840		}
1841	}
1842
1843	if (vxflags & VXLAN_F_GBP)
1844		vxlan_build_gbp_hdr(vxh, vxflags, md);
1845	if (vxflags & VXLAN_F_GPE) {
1846		err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1847		if (err < 0)
1848			return err;
1849		inner_protocol = skb->protocol;
1850	}
1851
1852	skb_set_inner_protocol(skb, inner_protocol);
1853	return 0;
1854}
1855
1856static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
1857				      struct vxlan_sock *sock4,
1858				      struct sk_buff *skb, int oif, u8 tos,
1859				      __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
1860				      struct dst_cache *dst_cache,
1861				      const struct ip_tunnel_info *info)
1862{
1863	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1864	struct rtable *rt = NULL;
1865	struct flowi4 fl4;
1866
1867	if (!sock4)
1868		return ERR_PTR(-EIO);
1869
1870	if (tos && !info)
1871		use_cache = false;
1872	if (use_cache) {
1873		rt = dst_cache_get_ip4(dst_cache, saddr);
1874		if (rt)
1875			return rt;
1876	}
1877
1878	memset(&fl4, 0, sizeof(fl4));
1879	fl4.flowi4_oif = oif;
1880	fl4.flowi4_tos = RT_TOS(tos);
1881	fl4.flowi4_mark = skb->mark;
1882	fl4.flowi4_proto = IPPROTO_UDP;
1883	fl4.daddr = daddr;
1884	fl4.saddr = *saddr;
1885	fl4.fl4_dport = dport;
1886	fl4.fl4_sport = sport;
1887
1888	rt = ip_route_output_key(vxlan->net, &fl4);
1889	if (likely(!IS_ERR(rt))) {
1890		if (rt->dst.dev == dev) {
1891			netdev_dbg(dev, "circular route to %pI4\n", &daddr);
1892			ip_rt_put(rt);
1893			return ERR_PTR(-ELOOP);
1894		}
1895
1896		*saddr = fl4.saddr;
1897		if (use_cache)
1898			dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
1899	} else {
1900		netdev_dbg(dev, "no route to %pI4\n", &daddr);
1901		return ERR_PTR(-ENETUNREACH);
1902	}
1903	return rt;
1904}
1905
1906#if IS_ENABLED(CONFIG_IPV6)
1907static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1908					  struct net_device *dev,
1909					  struct vxlan_sock *sock6,
1910					  struct sk_buff *skb, int oif, u8 tos,
1911					  __be32 label,
1912					  const struct in6_addr *daddr,
1913					  struct in6_addr *saddr,
1914					  __be16 dport, __be16 sport,
1915					  struct dst_cache *dst_cache,
1916					  const struct ip_tunnel_info *info)
1917{
1918	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1919	struct dst_entry *ndst;
1920	struct flowi6 fl6;
1921	int err;
1922
1923	if (!sock6)
1924		return ERR_PTR(-EIO);
1925
1926	if (tos && !info)
1927		use_cache = false;
1928	if (use_cache) {
1929		ndst = dst_cache_get_ip6(dst_cache, saddr);
1930		if (ndst)
1931			return ndst;
1932	}
1933
1934	memset(&fl6, 0, sizeof(fl6));
1935	fl6.flowi6_oif = oif;
1936	fl6.daddr = *daddr;
1937	fl6.saddr = *saddr;
1938	fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
1939	fl6.flowi6_mark = skb->mark;
1940	fl6.flowi6_proto = IPPROTO_UDP;
1941	fl6.fl6_dport = dport;
1942	fl6.fl6_sport = sport;
1943
1944	err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1945					 sock6->sock->sk,
1946					 &ndst, &fl6);
1947	if (unlikely(err < 0)) {
1948		netdev_dbg(dev, "no route to %pI6\n", daddr);
1949		return ERR_PTR(-ENETUNREACH);
1950	}
1951
1952	if (unlikely(ndst->dev == dev)) {
1953		netdev_dbg(dev, "circular route to %pI6\n", daddr);
1954		dst_release(ndst);
1955		return ERR_PTR(-ELOOP);
1956	}
1957
1958	*saddr = fl6.saddr;
1959	if (use_cache)
1960		dst_cache_set_ip6(dst_cache, ndst, saddr);
1961	return ndst;
1962}
1963#endif
1964
1965/* Bypass encapsulation if the destination is local */
1966static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1967			       struct vxlan_dev *dst_vxlan, __be32 vni)
 
1968{
1969	struct pcpu_sw_netstats *tx_stats, *rx_stats;
1970	union vxlan_addr loopback;
1971	union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
1972	struct net_device *dev = skb->dev;
1973	int len = skb->len;
1974
1975	tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1976	rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1977	skb->pkt_type = PACKET_HOST;
1978	skb->encapsulation = 0;
1979	skb->dev = dst_vxlan->dev;
1980	__skb_pull(skb, skb_network_offset(skb));
1981
1982	if (remote_ip->sa.sa_family == AF_INET) {
1983		loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1984		loopback.sa.sa_family =  AF_INET;
1985#if IS_ENABLED(CONFIG_IPV6)
1986	} else {
1987		loopback.sin6.sin6_addr = in6addr_loopback;
1988		loopback.sa.sa_family =  AF_INET6;
1989#endif
1990	}
1991
1992	if (dst_vxlan->cfg.flags & VXLAN_F_LEARN)
1993		vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source, 0,
1994			    vni);
 
 
 
 
 
 
1995
1996	u64_stats_update_begin(&tx_stats->syncp);
1997	tx_stats->tx_packets++;
1998	tx_stats->tx_bytes += len;
1999	u64_stats_update_end(&tx_stats->syncp);
2000
2001	if (netif_rx(skb) == NET_RX_SUCCESS) {
2002		u64_stats_update_begin(&rx_stats->syncp);
2003		rx_stats->rx_packets++;
2004		rx_stats->rx_bytes += len;
2005		u64_stats_update_end(&rx_stats->syncp);
2006	} else {
 
2007		dev->stats.rx_dropped++;
2008	}
 
2009}
2010
2011static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2012				 struct vxlan_dev *vxlan,
2013				 union vxlan_addr *daddr,
2014				 __be16 dst_port, int dst_ifindex, __be32 vni,
2015				 struct dst_entry *dst,
2016				 u32 rt_flags)
2017{
2018#if IS_ENABLED(CONFIG_IPV6)
2019	/* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2020	 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2021	 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2022	 */
2023	BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2024#endif
2025	/* Bypass encapsulation if the destination is local */
2026	if (rt_flags & RTCF_LOCAL &&
2027	    !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2028		struct vxlan_dev *dst_vxlan;
2029
2030		dst_release(dst);
2031		dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2032					   daddr->sa.sa_family, dst_port,
2033					   vxlan->cfg.flags);
2034		if (!dst_vxlan) {
2035			dev->stats.tx_errors++;
2036			kfree_skb(skb);
2037
2038			return -ENOENT;
2039		}
2040		vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
2041		return 1;
2042	}
2043
2044	return 0;
2045}
2046
2047static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2048			   __be32 default_vni, struct vxlan_rdst *rdst,
2049			   bool did_rsc)
2050{
2051	struct dst_cache *dst_cache;
2052	struct ip_tunnel_info *info;
2053	struct vxlan_dev *vxlan = netdev_priv(dev);
2054	const struct iphdr *old_iph = ip_hdr(skb);
2055	union vxlan_addr *dst;
2056	union vxlan_addr remote_ip, local_ip;
2057	struct vxlan_metadata _md;
2058	struct vxlan_metadata *md = &_md;
2059	__be16 src_port = 0, dst_port;
2060	struct dst_entry *ndst = NULL;
2061	__be32 vni, label;
2062	__u8 tos, ttl;
2063	int ifindex;
2064	int err;
2065	u32 flags = vxlan->cfg.flags;
2066	bool udp_sum = false;
2067	bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2068
2069	info = skb_tunnel_info(skb);
2070
2071	if (rdst) {
2072		dst = &rdst->remote_ip;
2073		if (vxlan_addr_any(dst)) {
2074			if (did_rsc) {
2075				/* short-circuited back to local bridge */
2076				vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
 
2077				return;
2078			}
2079			goto drop;
2080		}
2081
2082		dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2083		vni = (rdst->remote_vni) ? : default_vni;
2084		ifindex = rdst->remote_ifindex;
2085		local_ip = vxlan->cfg.saddr;
2086		dst_cache = &rdst->dst_cache;
2087		md->gbp = skb->mark;
2088		ttl = vxlan->cfg.ttl;
2089		if (!ttl && vxlan_addr_multicast(dst))
2090			ttl = 1;
 
 
 
 
2091
2092		tos = vxlan->cfg.tos;
2093		if (tos == 1)
2094			tos = ip_tunnel_get_dsfield(old_iph, skb);
2095
2096		if (dst->sa.sa_family == AF_INET)
2097			udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2098		else
2099			udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2100		label = vxlan->cfg.label;
2101	} else {
2102		if (!info) {
2103			WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2104				  dev->name);
2105			goto drop;
2106		}
2107		remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2108		if (remote_ip.sa.sa_family == AF_INET) {
2109			remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2110			local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2111		} else {
2112			remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2113			local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2114		}
2115		dst = &remote_ip;
2116		dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2117		vni = tunnel_id_to_key32(info->key.tun_id);
2118		ifindex = 0;
2119		dst_cache = &info->dst_cache;
2120		if (info->options_len)
 
 
2121			md = ip_tunnel_info_opts(info);
 
2122		ttl = info->key.ttl;
2123		tos = info->key.tos;
2124		label = info->key.label;
2125		udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2126	}
2127	src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2128				     vxlan->cfg.port_max, true);
2129
2130	rcu_read_lock();
2131	if (dst->sa.sa_family == AF_INET) {
2132		struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2133		struct rtable *rt;
2134		__be16 df = 0;
2135
 
 
 
2136		rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2137				     dst->sin.sin_addr.s_addr,
2138				     &local_ip.sin.sin_addr.s_addr,
2139				     dst_port, src_port,
2140				     dst_cache, info);
2141		if (IS_ERR(rt)) {
2142			err = PTR_ERR(rt);
2143			goto tx_error;
2144		}
2145
2146		/* Bypass encapsulation if the destination is local */
2147		if (!info) {
 
2148			err = encap_bypass_if_local(skb, dev, vxlan, dst,
2149						    dst_port, ifindex, vni,
2150						    &rt->dst, rt->rt_flags);
2151			if (err)
2152				goto out_unlock;
 
 
 
 
 
 
 
 
 
 
 
2153		} else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2154			df = htons(IP_DF);
2155		}
2156
2157		ndst = &rt->dst;
2158		if (skb_dst(skb)) {
2159			int mtu = dst_mtu(ndst) - VXLAN_HEADROOM;
2160
2161			skb_dst_update_pmtu(skb, mtu);
 
 
 
 
 
 
 
 
 
 
 
 
2162		}
2163
2164		tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2165		ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2166		err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2167				      vni, md, flags, udp_sum);
2168		if (err < 0)
2169			goto tx_error;
2170
2171		udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2172				    dst->sin.sin_addr.s_addr, tos, ttl, df,
2173				    src_port, dst_port, xnet, !udp_sum);
2174#if IS_ENABLED(CONFIG_IPV6)
2175	} else {
2176		struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2177
 
 
 
2178		ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2179					label, &dst->sin6.sin6_addr,
2180					&local_ip.sin6.sin6_addr,
2181					dst_port, src_port,
2182					dst_cache, info);
2183		if (IS_ERR(ndst)) {
2184			err = PTR_ERR(ndst);
2185			ndst = NULL;
2186			goto tx_error;
2187		}
2188
2189		if (!info) {
2190			u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2191
2192			err = encap_bypass_if_local(skb, dev, vxlan, dst,
2193						    dst_port, ifindex, vni,
2194						    ndst, rt6i_flags);
2195			if (err)
2196				goto out_unlock;
2197		}
2198
2199		if (skb_dst(skb)) {
2200			int mtu = dst_mtu(ndst) - VXLAN6_HEADROOM;
 
 
 
 
 
 
 
 
 
 
 
2201
2202			skb_dst_update_pmtu(skb, mtu);
 
 
2203		}
2204
2205		tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2206		ttl = ttl ? : ip6_dst_hoplimit(ndst);
2207		skb_scrub_packet(skb, xnet);
2208		err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2209				      vni, md, flags, udp_sum);
2210		if (err < 0)
2211			goto tx_error;
2212
2213		udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2214				     &local_ip.sin6.sin6_addr,
2215				     &dst->sin6.sin6_addr, tos, ttl,
2216				     label, src_port, dst_port, !udp_sum);
2217#endif
2218	}
2219out_unlock:
2220	rcu_read_unlock();
2221	return;
2222
2223drop:
2224	dev->stats.tx_dropped++;
2225	dev_kfree_skb(skb);
2226	return;
2227
2228tx_error:
2229	rcu_read_unlock();
2230	if (err == -ELOOP)
2231		dev->stats.collisions++;
2232	else if (err == -ENETUNREACH)
2233		dev->stats.tx_carrier_errors++;
2234	dst_release(ndst);
2235	dev->stats.tx_errors++;
2236	kfree_skb(skb);
2237}
2238
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2239/* Transmit local packets over Vxlan
2240 *
2241 * Outer IP header inherits ECN and DF from inner header.
2242 * Outer UDP destination is the VXLAN assigned port.
2243 *           source port is based on hash of flow
2244 */
2245static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2246{
2247	struct vxlan_dev *vxlan = netdev_priv(dev);
2248	struct vxlan_rdst *rdst, *fdst = NULL;
2249	const struct ip_tunnel_info *info;
2250	bool did_rsc = false;
2251	struct vxlan_fdb *f;
2252	struct ethhdr *eth;
2253	__be32 vni = 0;
2254
2255	info = skb_tunnel_info(skb);
2256
2257	skb_reset_mac_header(skb);
2258
2259	if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2260		if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2261		    info->mode & IP_TUNNEL_INFO_TX) {
2262			vni = tunnel_id_to_key32(info->key.tun_id);
2263		} else {
2264			if (info && info->mode & IP_TUNNEL_INFO_TX)
2265				vxlan_xmit_one(skb, dev, vni, NULL, false);
2266			else
2267				kfree_skb(skb);
2268			return NETDEV_TX_OK;
2269		}
2270	}
2271
2272	if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2273		eth = eth_hdr(skb);
2274		if (ntohs(eth->h_proto) == ETH_P_ARP)
2275			return arp_reduce(dev, skb, vni);
2276#if IS_ENABLED(CONFIG_IPV6)
2277		else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2278			 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2279					    sizeof(struct nd_msg)) &&
2280			 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2281			struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2282
2283			if (m->icmph.icmp6_code == 0 &&
2284			    m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2285				return neigh_reduce(dev, skb, vni);
2286		}
2287#endif
2288	}
2289
2290	eth = eth_hdr(skb);
2291	f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2292	did_rsc = false;
2293
2294	if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2295	    (ntohs(eth->h_proto) == ETH_P_IP ||
2296	     ntohs(eth->h_proto) == ETH_P_IPV6)) {
2297		did_rsc = route_shortcircuit(dev, skb);
2298		if (did_rsc)
2299			f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2300	}
2301
2302	if (f == NULL) {
2303		f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2304		if (f == NULL) {
2305			if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2306			    !is_multicast_ether_addr(eth->h_dest))
2307				vxlan_fdb_miss(vxlan, eth->h_dest);
2308
2309			dev->stats.tx_dropped++;
2310			kfree_skb(skb);
2311			return NETDEV_TX_OK;
2312		}
2313	}
2314
2315	list_for_each_entry_rcu(rdst, &f->remotes, list) {
2316		struct sk_buff *skb1;
 
 
 
 
2317
2318		if (!fdst) {
2319			fdst = rdst;
2320			continue;
 
 
 
 
2321		}
2322		skb1 = skb_clone(skb, GFP_ATOMIC);
2323		if (skb1)
2324			vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
 
2325	}
2326
2327	if (fdst)
2328		vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2329	else
2330		kfree_skb(skb);
2331	return NETDEV_TX_OK;
2332}
2333
2334/* Walk the forwarding table and purge stale entries */
2335static void vxlan_cleanup(struct timer_list *t)
2336{
2337	struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2338	unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2339	unsigned int h;
2340
2341	if (!netif_running(vxlan->dev))
2342		return;
2343
2344	for (h = 0; h < FDB_HASH_SIZE; ++h) {
2345		struct hlist_node *p, *n;
2346
2347		spin_lock_bh(&vxlan->hash_lock);
2348		hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2349			struct vxlan_fdb *f
2350				= container_of(p, struct vxlan_fdb, hlist);
2351			unsigned long timeout;
2352
2353			if (f->state & (NUD_PERMANENT | NUD_NOARP))
2354				continue;
2355
2356			if (f->flags & NTF_EXT_LEARNED)
2357				continue;
2358
2359			timeout = f->used + vxlan->cfg.age_interval * HZ;
2360			if (time_before_eq(timeout, jiffies)) {
2361				netdev_dbg(vxlan->dev,
2362					   "garbage collect %pM\n",
2363					   f->eth_addr);
2364				f->state = NUD_STALE;
2365				vxlan_fdb_destroy(vxlan, f);
2366			} else if (time_before(timeout, next_timer))
2367				next_timer = timeout;
2368		}
2369		spin_unlock_bh(&vxlan->hash_lock);
2370	}
2371
2372	mod_timer(&vxlan->age_timer, next_timer);
2373}
2374
2375static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2376{
2377	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2378
2379	spin_lock(&vn->sock_lock);
2380	hlist_del_init_rcu(&vxlan->hlist4.hlist);
2381#if IS_ENABLED(CONFIG_IPV6)
2382	hlist_del_init_rcu(&vxlan->hlist6.hlist);
2383#endif
2384	spin_unlock(&vn->sock_lock);
2385}
2386
2387static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2388			     struct vxlan_dev_node *node)
2389{
2390	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2391	__be32 vni = vxlan->default_dst.remote_vni;
2392
2393	node->vxlan = vxlan;
2394	spin_lock(&vn->sock_lock);
2395	hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2396	spin_unlock(&vn->sock_lock);
2397}
2398
2399/* Setup stats when device is created */
2400static int vxlan_init(struct net_device *dev)
2401{
 
 
 
2402	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2403	if (!dev->tstats)
2404		return -ENOMEM;
2405
 
 
 
 
 
 
2406	return 0;
2407}
2408
2409static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
2410{
2411	struct vxlan_fdb *f;
 
2412
2413	spin_lock_bh(&vxlan->hash_lock);
2414	f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
2415	if (f)
2416		vxlan_fdb_destroy(vxlan, f);
2417	spin_unlock_bh(&vxlan->hash_lock);
2418}
2419
2420static void vxlan_uninit(struct net_device *dev)
2421{
2422	struct vxlan_dev *vxlan = netdev_priv(dev);
2423
 
 
2424	vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
2425
2426	free_percpu(dev->tstats);
2427}
2428
2429/* Start ageing timer and join group when device is brought up */
2430static int vxlan_open(struct net_device *dev)
2431{
2432	struct vxlan_dev *vxlan = netdev_priv(dev);
2433	int ret;
2434
2435	ret = vxlan_sock_add(vxlan);
2436	if (ret < 0)
2437		return ret;
2438
2439	if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
2440		ret = vxlan_igmp_join(vxlan);
2441		if (ret == -EADDRINUSE)
2442			ret = 0;
2443		if (ret) {
2444			vxlan_sock_release(vxlan);
2445			return ret;
2446		}
2447	}
2448
2449	if (vxlan->cfg.age_interval)
2450		mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2451
2452	return ret;
2453}
2454
2455/* Purge the forwarding table */
2456static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
2457{
2458	unsigned int h;
2459
2460	spin_lock_bh(&vxlan->hash_lock);
2461	for (h = 0; h < FDB_HASH_SIZE; ++h) {
2462		struct hlist_node *p, *n;
 
 
2463		hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2464			struct vxlan_fdb *f
2465				= container_of(p, struct vxlan_fdb, hlist);
2466			if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2467				continue;
2468			/* the all_zeros_mac entry is deleted at vxlan_uninit */
2469			if (!is_zero_ether_addr(f->eth_addr))
2470				vxlan_fdb_destroy(vxlan, f);
 
 
2471		}
 
2472	}
2473	spin_unlock_bh(&vxlan->hash_lock);
2474}
2475
2476/* Cleanup timer and forwarding table on shutdown */
2477static int vxlan_stop(struct net_device *dev)
2478{
2479	struct vxlan_dev *vxlan = netdev_priv(dev);
2480	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2481	int ret = 0;
2482
2483	if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2484	    !vxlan_group_used(vn, vxlan))
2485		ret = vxlan_igmp_leave(vxlan);
2486
2487	del_timer_sync(&vxlan->age_timer);
2488
2489	vxlan_flush(vxlan, false);
2490	vxlan_sock_release(vxlan);
2491
2492	return ret;
2493}
2494
2495/* Stub, nothing needs to be done. */
2496static void vxlan_set_multicast_list(struct net_device *dev)
2497{
2498}
2499
2500static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2501{
2502	struct vxlan_dev *vxlan = netdev_priv(dev);
2503	struct vxlan_rdst *dst = &vxlan->default_dst;
2504	struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2505							 dst->remote_ifindex);
2506	bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
2507
2508	/* This check is different than dev->max_mtu, because it looks at
2509	 * the lowerdev->mtu, rather than the static dev->max_mtu
2510	 */
2511	if (lowerdev) {
2512		int max_mtu = lowerdev->mtu -
2513			      (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2514		if (new_mtu > max_mtu)
2515			return -EINVAL;
2516	}
2517
2518	dev->mtu = new_mtu;
2519	return 0;
2520}
2521
2522static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2523{
2524	struct vxlan_dev *vxlan = netdev_priv(dev);
2525	struct ip_tunnel_info *info = skb_tunnel_info(skb);
2526	__be16 sport, dport;
2527
2528	sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2529				  vxlan->cfg.port_max, true);
2530	dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2531
2532	if (ip_tunnel_info_af(info) == AF_INET) {
2533		struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2534		struct rtable *rt;
2535
2536		rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
2537				     info->key.u.ipv4.dst,
2538				     &info->key.u.ipv4.src, dport, sport,
2539				     &info->dst_cache, info);
2540		if (IS_ERR(rt))
2541			return PTR_ERR(rt);
2542		ip_rt_put(rt);
2543	} else {
2544#if IS_ENABLED(CONFIG_IPV6)
2545		struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2546		struct dst_entry *ndst;
2547
2548		ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
2549					info->key.label, &info->key.u.ipv6.dst,
2550					&info->key.u.ipv6.src, dport, sport,
2551					&info->dst_cache, info);
2552		if (IS_ERR(ndst))
2553			return PTR_ERR(ndst);
2554		dst_release(ndst);
2555#else /* !CONFIG_IPV6 */
2556		return -EPFNOSUPPORT;
2557#endif
2558	}
2559	info->key.tp_src = sport;
2560	info->key.tp_dst = dport;
2561	return 0;
2562}
2563
2564static const struct net_device_ops vxlan_netdev_ether_ops = {
2565	.ndo_init		= vxlan_init,
2566	.ndo_uninit		= vxlan_uninit,
2567	.ndo_open		= vxlan_open,
2568	.ndo_stop		= vxlan_stop,
2569	.ndo_start_xmit		= vxlan_xmit,
2570	.ndo_get_stats64	= ip_tunnel_get_stats64,
2571	.ndo_set_rx_mode	= vxlan_set_multicast_list,
2572	.ndo_change_mtu		= vxlan_change_mtu,
2573	.ndo_validate_addr	= eth_validate_addr,
2574	.ndo_set_mac_address	= eth_mac_addr,
2575	.ndo_fdb_add		= vxlan_fdb_add,
2576	.ndo_fdb_del		= vxlan_fdb_delete,
2577	.ndo_fdb_dump		= vxlan_fdb_dump,
 
2578	.ndo_fill_metadata_dst	= vxlan_fill_metadata_dst,
 
2579};
2580
2581static const struct net_device_ops vxlan_netdev_raw_ops = {
2582	.ndo_init		= vxlan_init,
2583	.ndo_uninit		= vxlan_uninit,
2584	.ndo_open		= vxlan_open,
2585	.ndo_stop		= vxlan_stop,
2586	.ndo_start_xmit		= vxlan_xmit,
2587	.ndo_get_stats64	= ip_tunnel_get_stats64,
2588	.ndo_change_mtu		= vxlan_change_mtu,
2589	.ndo_fill_metadata_dst	= vxlan_fill_metadata_dst,
2590};
2591
2592/* Info for udev, that this is a virtual tunnel endpoint */
2593static struct device_type vxlan_type = {
2594	.name = "vxlan",
2595};
2596
2597/* Calls the ndo_udp_tunnel_add of the caller in order to
2598 * supply the listening VXLAN udp ports. Callers are expected
2599 * to implement the ndo_udp_tunnel_add.
2600 */
2601static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
2602{
2603	struct vxlan_sock *vs;
2604	struct net *net = dev_net(dev);
2605	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2606	unsigned int i;
2607
2608	spin_lock(&vn->sock_lock);
2609	for (i = 0; i < PORT_HASH_SIZE; ++i) {
2610		hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2611			unsigned short type;
2612
2613			if (vs->flags & VXLAN_F_GPE)
2614				type = UDP_TUNNEL_TYPE_VXLAN_GPE;
2615			else
2616				type = UDP_TUNNEL_TYPE_VXLAN;
2617
2618			if (push)
2619				udp_tunnel_push_rx_port(dev, vs->sock, type);
2620			else
2621				udp_tunnel_drop_rx_port(dev, vs->sock, type);
2622		}
2623	}
2624	spin_unlock(&vn->sock_lock);
2625}
2626
2627/* Initialize the device structure. */
2628static void vxlan_setup(struct net_device *dev)
2629{
2630	struct vxlan_dev *vxlan = netdev_priv(dev);
2631	unsigned int h;
2632
2633	eth_hw_addr_random(dev);
2634	ether_setup(dev);
2635
2636	dev->needs_free_netdev = true;
2637	SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2638
2639	dev->features	|= NETIF_F_LLTX;
2640	dev->features	|= NETIF_F_SG | NETIF_F_HW_CSUM;
2641	dev->features   |= NETIF_F_RXCSUM;
2642	dev->features   |= NETIF_F_GSO_SOFTWARE;
2643
2644	dev->vlan_features = dev->features;
2645	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2646	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2647	netif_keep_dst(dev);
2648	dev->priv_flags |= IFF_NO_QUEUE;
2649
2650	/* MTU range: 68 - 65535 */
2651	dev->min_mtu = ETH_MIN_MTU;
2652	dev->max_mtu = ETH_MAX_MTU;
2653
2654	INIT_LIST_HEAD(&vxlan->next);
2655	spin_lock_init(&vxlan->hash_lock);
2656
2657	timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
2658
2659	vxlan->dev = dev;
2660
2661	gro_cells_init(&vxlan->gro_cells, dev);
2662
2663	for (h = 0; h < FDB_HASH_SIZE; ++h)
2664		INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
 
2665}
2666
2667static void vxlan_ether_setup(struct net_device *dev)
2668{
2669	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2670	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2671	dev->netdev_ops = &vxlan_netdev_ether_ops;
2672}
2673
2674static void vxlan_raw_setup(struct net_device *dev)
2675{
2676	dev->header_ops = NULL;
2677	dev->type = ARPHRD_NONE;
2678	dev->hard_header_len = 0;
2679	dev->addr_len = 0;
2680	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2681	dev->netdev_ops = &vxlan_netdev_raw_ops;
2682}
2683
2684static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2685	[IFLA_VXLAN_ID]		= { .type = NLA_U32 },
2686	[IFLA_VXLAN_GROUP]	= { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2687	[IFLA_VXLAN_GROUP6]	= { .len = sizeof(struct in6_addr) },
2688	[IFLA_VXLAN_LINK]	= { .type = NLA_U32 },
2689	[IFLA_VXLAN_LOCAL]	= { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2690	[IFLA_VXLAN_LOCAL6]	= { .len = sizeof(struct in6_addr) },
2691	[IFLA_VXLAN_TOS]	= { .type = NLA_U8 },
2692	[IFLA_VXLAN_TTL]	= { .type = NLA_U8 },
2693	[IFLA_VXLAN_LABEL]	= { .type = NLA_U32 },
2694	[IFLA_VXLAN_LEARNING]	= { .type = NLA_U8 },
2695	[IFLA_VXLAN_AGEING]	= { .type = NLA_U32 },
2696	[IFLA_VXLAN_LIMIT]	= { .type = NLA_U32 },
2697	[IFLA_VXLAN_PORT_RANGE] = { .len  = sizeof(struct ifla_vxlan_port_range) },
2698	[IFLA_VXLAN_PROXY]	= { .type = NLA_U8 },
2699	[IFLA_VXLAN_RSC]	= { .type = NLA_U8 },
2700	[IFLA_VXLAN_L2MISS]	= { .type = NLA_U8 },
2701	[IFLA_VXLAN_L3MISS]	= { .type = NLA_U8 },
2702	[IFLA_VXLAN_COLLECT_METADATA]	= { .type = NLA_U8 },
2703	[IFLA_VXLAN_PORT]	= { .type = NLA_U16 },
2704	[IFLA_VXLAN_UDP_CSUM]	= { .type = NLA_U8 },
2705	[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]	= { .type = NLA_U8 },
2706	[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]	= { .type = NLA_U8 },
2707	[IFLA_VXLAN_REMCSUM_TX]	= { .type = NLA_U8 },
2708	[IFLA_VXLAN_REMCSUM_RX]	= { .type = NLA_U8 },
2709	[IFLA_VXLAN_GBP]	= { .type = NLA_FLAG, },
2710	[IFLA_VXLAN_GPE]	= { .type = NLA_FLAG, },
2711	[IFLA_VXLAN_REMCSUM_NOPARTIAL]	= { .type = NLA_FLAG },
 
 
2712};
2713
2714static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
2715			  struct netlink_ext_ack *extack)
2716{
2717	if (tb[IFLA_ADDRESS]) {
2718		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2719			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2720					    "Provided link layer address is not Ethernet");
2721			return -EINVAL;
2722		}
2723
2724		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2725			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2726					    "Provided Ethernet address is not unicast");
2727			return -EADDRNOTAVAIL;
2728		}
2729	}
2730
2731	if (tb[IFLA_MTU]) {
2732		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
2733
2734		if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
2735			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
2736					    "MTU must be between 68 and 65535");
2737			return -EINVAL;
2738		}
2739	}
2740
2741	if (!data) {
2742		NL_SET_ERR_MSG(extack,
2743			       "Required attributes not provided to perform the operation");
2744		return -EINVAL;
2745	}
2746
2747	if (data[IFLA_VXLAN_ID]) {
2748		u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2749
2750		if (id >= VXLAN_N_VID) {
2751			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID],
2752					    "VXLAN ID must be lower than 16777216");
2753			return -ERANGE;
2754		}
2755	}
2756
2757	if (data[IFLA_VXLAN_PORT_RANGE]) {
2758		const struct ifla_vxlan_port_range *p
2759			= nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2760
2761		if (ntohs(p->high) < ntohs(p->low)) {
2762			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
2763					    "Invalid source port range");
2764			return -EINVAL;
2765		}
2766	}
2767
 
 
 
 
 
 
 
 
 
 
2768	return 0;
2769}
2770
2771static void vxlan_get_drvinfo(struct net_device *netdev,
2772			      struct ethtool_drvinfo *drvinfo)
2773{
2774	strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2775	strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2776}
2777
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2778static const struct ethtool_ops vxlan_ethtool_ops = {
2779	.get_drvinfo	= vxlan_get_drvinfo,
2780	.get_link	= ethtool_op_get_link,
 
2781};
2782
2783static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2784					__be16 port, u32 flags)
2785{
2786	struct socket *sock;
2787	struct udp_port_cfg udp_conf;
2788	int err;
2789
2790	memset(&udp_conf, 0, sizeof(udp_conf));
2791
2792	if (ipv6) {
2793		udp_conf.family = AF_INET6;
2794		udp_conf.use_udp6_rx_checksums =
2795		    !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
2796		udp_conf.ipv6_v6only = 1;
2797	} else {
2798		udp_conf.family = AF_INET;
2799	}
2800
2801	udp_conf.local_udp_port = port;
 
2802
2803	/* Open UDP socket */
2804	err = udp_sock_create(net, &udp_conf, &sock);
2805	if (err < 0)
2806		return ERR_PTR(err);
2807
2808	return sock;
2809}
2810
2811/* Create new listen socket if needed */
2812static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2813					      __be16 port, u32 flags)
 
2814{
2815	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2816	struct vxlan_sock *vs;
2817	struct socket *sock;
2818	unsigned int h;
2819	struct udp_tunnel_sock_cfg tunnel_cfg;
2820
2821	vs = kzalloc(sizeof(*vs), GFP_KERNEL);
2822	if (!vs)
2823		return ERR_PTR(-ENOMEM);
2824
2825	for (h = 0; h < VNI_HASH_SIZE; ++h)
2826		INIT_HLIST_HEAD(&vs->vni_list[h]);
2827
2828	sock = vxlan_create_sock(net, ipv6, port, flags);
2829	if (IS_ERR(sock)) {
2830		kfree(vs);
2831		return ERR_CAST(sock);
2832	}
2833
2834	vs->sock = sock;
2835	refcount_set(&vs->refcnt, 1);
2836	vs->flags = (flags & VXLAN_F_RCV_FLAGS);
2837
2838	spin_lock(&vn->sock_lock);
2839	hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2840	udp_tunnel_notify_add_rx_port(sock,
2841				      (vs->flags & VXLAN_F_GPE) ?
2842				      UDP_TUNNEL_TYPE_VXLAN_GPE :
2843				      UDP_TUNNEL_TYPE_VXLAN);
2844	spin_unlock(&vn->sock_lock);
2845
2846	/* Mark socket as an encapsulation socket. */
2847	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
2848	tunnel_cfg.sk_user_data = vs;
2849	tunnel_cfg.encap_type = 1;
2850	tunnel_cfg.encap_rcv = vxlan_rcv;
 
2851	tunnel_cfg.encap_destroy = NULL;
2852	tunnel_cfg.gro_receive = vxlan_gro_receive;
2853	tunnel_cfg.gro_complete = vxlan_gro_complete;
2854
2855	setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
2856
2857	return vs;
2858}
2859
2860static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
2861{
2862	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2863	struct vxlan_sock *vs = NULL;
2864	struct vxlan_dev_node *node;
 
 
 
 
 
2865
2866	if (!vxlan->cfg.no_share) {
2867		spin_lock(&vn->sock_lock);
2868		vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2869				     vxlan->cfg.dst_port, vxlan->cfg.flags);
 
2870		if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
2871			spin_unlock(&vn->sock_lock);
2872			return -EBUSY;
2873		}
2874		spin_unlock(&vn->sock_lock);
2875	}
2876	if (!vs)
2877		vs = vxlan_socket_create(vxlan->net, ipv6,
2878					 vxlan->cfg.dst_port, vxlan->cfg.flags);
 
2879	if (IS_ERR(vs))
2880		return PTR_ERR(vs);
2881#if IS_ENABLED(CONFIG_IPV6)
2882	if (ipv6) {
2883		rcu_assign_pointer(vxlan->vn6_sock, vs);
2884		node = &vxlan->hlist6;
2885	} else
2886#endif
2887	{
2888		rcu_assign_pointer(vxlan->vn4_sock, vs);
2889		node = &vxlan->hlist4;
2890	}
2891	vxlan_vs_add_dev(vs, vxlan, node);
2892	return 0;
2893}
2894
2895static int vxlan_sock_add(struct vxlan_dev *vxlan)
2896{
2897	bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
2898	bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
2899	bool ipv4 = !ipv6 || metadata;
2900	int ret = 0;
2901
2902	RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
2903#if IS_ENABLED(CONFIG_IPV6)
2904	RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
2905	if (ipv6) {
2906		ret = __vxlan_sock_add(vxlan, true);
2907		if (ret < 0 && ret != -EAFNOSUPPORT)
2908			ipv4 = false;
2909	}
2910#endif
2911	if (ipv4)
2912		ret = __vxlan_sock_add(vxlan, false);
2913	if (ret < 0)
2914		vxlan_sock_release(vxlan);
2915	return ret;
2916}
2917
2918static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
2919				 struct net_device **lower,
2920				 struct vxlan_dev *old,
2921				 struct netlink_ext_ack *extack)
2922{
2923	struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
2924	struct vxlan_dev *tmp;
2925	bool use_ipv6 = false;
2926
2927	if (conf->flags & VXLAN_F_GPE) {
2928		/* For now, allow GPE only together with
2929		 * COLLECT_METADATA. This can be relaxed later; in such
2930		 * case, the other side of the PtP link will have to be
2931		 * provided.
2932		 */
2933		if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
2934		    !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
2935			NL_SET_ERR_MSG(extack,
2936				       "VXLAN GPE does not support this combination of attributes");
2937			return -EINVAL;
2938		}
2939	}
2940
2941	if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
2942		/* Unless IPv6 is explicitly requested, assume IPv4 */
2943		conf->remote_ip.sa.sa_family = AF_INET;
2944		conf->saddr.sa.sa_family = AF_INET;
2945	} else if (!conf->remote_ip.sa.sa_family) {
2946		conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
2947	} else if (!conf->saddr.sa.sa_family) {
2948		conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
2949	}
2950
2951	if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
2952		NL_SET_ERR_MSG(extack,
2953			       "Local and remote address must be from the same family");
2954		return -EINVAL;
2955	}
2956
2957	if (vxlan_addr_multicast(&conf->saddr)) {
2958		NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
2959		return -EINVAL;
2960	}
2961
2962	if (conf->saddr.sa.sa_family == AF_INET6) {
2963		if (!IS_ENABLED(CONFIG_IPV6)) {
2964			NL_SET_ERR_MSG(extack,
2965				       "IPv6 support not enabled in the kernel");
2966			return -EPFNOSUPPORT;
2967		}
2968		use_ipv6 = true;
2969		conf->flags |= VXLAN_F_IPV6;
2970
2971		if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
2972			int local_type =
2973				ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
2974			int remote_type =
2975				ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
2976
2977			if (local_type & IPV6_ADDR_LINKLOCAL) {
2978				if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
2979				    (remote_type != IPV6_ADDR_ANY)) {
2980					NL_SET_ERR_MSG(extack,
2981						       "Invalid combination of local and remote address scopes");
2982					return -EINVAL;
2983				}
2984
2985				conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
2986			} else {
2987				if (remote_type ==
2988				    (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
2989					NL_SET_ERR_MSG(extack,
2990						       "Invalid combination of local and remote address scopes");
2991					return -EINVAL;
2992				}
2993
2994				conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
2995			}
2996		}
2997	}
2998
2999	if (conf->label && !use_ipv6) {
3000		NL_SET_ERR_MSG(extack,
3001			       "Label attribute only applies to IPv6 VXLAN devices");
3002		return -EINVAL;
3003	}
3004
3005	if (conf->remote_ifindex) {
3006		struct net_device *lowerdev;
3007
3008		lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3009		if (!lowerdev) {
3010			NL_SET_ERR_MSG(extack,
3011				       "Invalid local interface, device not found");
3012			return -ENODEV;
3013		}
3014
3015#if IS_ENABLED(CONFIG_IPV6)
3016		if (use_ipv6) {
3017			struct inet6_dev *idev = __in6_dev_get(lowerdev);
3018			if (idev && idev->cnf.disable_ipv6) {
3019				NL_SET_ERR_MSG(extack,
3020					       "IPv6 support disabled by administrator");
3021				return -EPERM;
3022			}
3023		}
3024#endif
3025
3026		*lower = lowerdev;
3027	} else {
3028		if (vxlan_addr_multicast(&conf->remote_ip)) {
3029			NL_SET_ERR_MSG(extack,
3030				       "Local interface required for multicast remote destination");
3031
3032			return -EINVAL;
3033		}
3034
3035#if IS_ENABLED(CONFIG_IPV6)
3036		if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3037			NL_SET_ERR_MSG(extack,
3038				       "Local interface required for link-local local/remote addresses");
3039			return -EINVAL;
3040		}
3041#endif
3042
3043		*lower = NULL;
3044	}
3045
3046	if (!conf->dst_port) {
3047		if (conf->flags & VXLAN_F_GPE)
3048			conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3049		else
3050			conf->dst_port = htons(vxlan_port);
3051	}
3052
3053	if (!conf->age_interval)
3054		conf->age_interval = FDB_AGE_DEFAULT;
3055
3056	list_for_each_entry(tmp, &vn->vxlan_list, next) {
3057		if (tmp == old)
3058			continue;
3059
3060		if (tmp->cfg.vni != conf->vni)
3061			continue;
3062		if (tmp->cfg.dst_port != conf->dst_port)
3063			continue;
3064		if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3065		    (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3066			continue;
3067
3068		if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3069		    tmp->cfg.remote_ifindex != conf->remote_ifindex)
3070			continue;
3071
3072		NL_SET_ERR_MSG(extack,
3073			       "A VXLAN device with the specified VNI already exists");
3074		return -EEXIST;
3075	}
3076
3077	return 0;
3078}
3079
3080static void vxlan_config_apply(struct net_device *dev,
3081			       struct vxlan_config *conf,
3082			       struct net_device *lowerdev,
3083			       struct net *src_net,
3084			       bool changelink)
3085{
3086	struct vxlan_dev *vxlan = netdev_priv(dev);
3087	struct vxlan_rdst *dst = &vxlan->default_dst;
3088	unsigned short needed_headroom = ETH_HLEN;
3089	bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3090	int max_mtu = ETH_MAX_MTU;
3091
3092	if (!changelink) {
3093		if (conf->flags & VXLAN_F_GPE)
3094			vxlan_raw_setup(dev);
3095		else
3096			vxlan_ether_setup(dev);
3097
3098		if (conf->mtu)
3099			dev->mtu = conf->mtu;
3100
3101		vxlan->net = src_net;
3102	}
3103
3104	dst->remote_vni = conf->vni;
3105
3106	memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3107
3108	if (lowerdev) {
3109		dst->remote_ifindex = conf->remote_ifindex;
3110
3111		dev->gso_max_size = lowerdev->gso_max_size;
3112		dev->gso_max_segs = lowerdev->gso_max_segs;
3113
3114		needed_headroom = lowerdev->hard_header_len;
3115
3116		max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3117					   VXLAN_HEADROOM);
3118		if (max_mtu < ETH_MIN_MTU)
3119			max_mtu = ETH_MIN_MTU;
3120
3121		if (!changelink && !conf->mtu)
3122			dev->mtu = max_mtu;
3123	}
3124
3125	if (dev->mtu > max_mtu)
3126		dev->mtu = max_mtu;
3127
3128	if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3129		needed_headroom += VXLAN6_HEADROOM;
3130	else
3131		needed_headroom += VXLAN_HEADROOM;
3132	dev->needed_headroom = needed_headroom;
3133
3134	memcpy(&vxlan->cfg, conf, sizeof(*conf));
3135}
3136
3137static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3138			       struct vxlan_config *conf, bool changelink,
3139			       struct netlink_ext_ack *extack)
3140{
3141	struct vxlan_dev *vxlan = netdev_priv(dev);
3142	struct net_device *lowerdev;
3143	int ret;
3144
3145	ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3146	if (ret)
3147		return ret;
3148
3149	vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3150
3151	return 0;
3152}
3153
3154static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3155			      struct vxlan_config *conf,
3156			      struct netlink_ext_ack *extack)
3157{
3158	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3159	struct vxlan_dev *vxlan = netdev_priv(dev);
 
 
 
 
3160	int err;
3161
 
3162	err = vxlan_dev_configure(net, dev, conf, false, extack);
3163	if (err)
3164		return err;
3165
3166	dev->ethtool_ops = &vxlan_ethtool_ops;
3167
3168	/* create an fdb entry for a valid default destination */
3169	if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
3170		err = vxlan_fdb_create(vxlan, all_zeros_mac,
3171				       &vxlan->default_dst.remote_ip,
3172				       NUD_REACHABLE | NUD_PERMANENT,
3173				       NLM_F_EXCL | NLM_F_CREATE,
3174				       vxlan->cfg.dst_port,
3175				       vxlan->default_dst.remote_vni,
3176				       vxlan->default_dst.remote_vni,
3177				       vxlan->default_dst.remote_ifindex,
3178				       NTF_SELF);
3179		if (err)
3180			return err;
3181	}
3182
3183	err = register_netdevice(dev);
3184	if (err) {
3185		vxlan_fdb_delete_default(vxlan, vxlan->default_dst.remote_vni);
3186		return err;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3187	}
3188
3189	list_add(&vxlan->next, &vn->vxlan_list);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3190	return 0;
3191}
3192
3193static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3194			 struct net_device *dev, struct vxlan_config *conf,
3195			 bool changelink)
3196{
3197	struct vxlan_dev *vxlan = netdev_priv(dev);
 
3198
3199	memset(conf, 0, sizeof(*conf));
3200
3201	/* if changelink operation, start with old existing cfg */
3202	if (changelink)
3203		memcpy(conf, &vxlan->cfg, sizeof(*conf));
3204
3205	if (data[IFLA_VXLAN_ID]) {
3206		__be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3207
3208		if (changelink && (vni != conf->vni))
 
3209			return -EOPNOTSUPP;
 
3210		conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3211	}
3212
3213	if (data[IFLA_VXLAN_GROUP]) {
3214		if (changelink && (conf->remote_ip.sa.sa_family != AF_INET))
 
3215			return -EOPNOTSUPP;
 
3216
3217		conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
3218		conf->remote_ip.sa.sa_family = AF_INET;
3219	} else if (data[IFLA_VXLAN_GROUP6]) {
3220		if (!IS_ENABLED(CONFIG_IPV6))
 
3221			return -EPFNOSUPPORT;
 
3222
3223		if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6))
 
3224			return -EOPNOTSUPP;
 
3225
3226		conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3227		conf->remote_ip.sa.sa_family = AF_INET6;
3228	}
3229
3230	if (data[IFLA_VXLAN_LOCAL]) {
3231		if (changelink && (conf->saddr.sa.sa_family != AF_INET))
 
3232			return -EOPNOTSUPP;
 
3233
3234		conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3235		conf->saddr.sa.sa_family = AF_INET;
3236	} else if (data[IFLA_VXLAN_LOCAL6]) {
3237		if (!IS_ENABLED(CONFIG_IPV6))
 
3238			return -EPFNOSUPPORT;
 
3239
3240		if (changelink && (conf->saddr.sa.sa_family != AF_INET6))
 
3241			return -EOPNOTSUPP;
 
3242
3243		/* TODO: respect scope id */
3244		conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3245		conf->saddr.sa.sa_family = AF_INET6;
3246	}
3247
3248	if (data[IFLA_VXLAN_LINK])
3249		conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
3250
3251	if (data[IFLA_VXLAN_TOS])
3252		conf->tos  = nla_get_u8(data[IFLA_VXLAN_TOS]);
3253
3254	if (data[IFLA_VXLAN_TTL])
3255		conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
3256
 
 
 
 
 
 
 
 
 
3257	if (data[IFLA_VXLAN_LABEL])
3258		conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
3259			     IPV6_FLOWLABEL_MASK;
3260
3261	if (data[IFLA_VXLAN_LEARNING]) {
3262		if (nla_get_u8(data[IFLA_VXLAN_LEARNING]))
3263			conf->flags |= VXLAN_F_LEARN;
3264		else
3265			conf->flags &= ~VXLAN_F_LEARN;
 
3266	} else if (!changelink) {
3267		/* default to learn on a new device */
3268		conf->flags |= VXLAN_F_LEARN;
3269	}
3270
3271	if (data[IFLA_VXLAN_AGEING]) {
3272		if (changelink)
3273			return -EOPNOTSUPP;
3274		conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3275	}
3276
3277	if (data[IFLA_VXLAN_PROXY]) {
3278		if (changelink)
3279			return -EOPNOTSUPP;
3280		if (nla_get_u8(data[IFLA_VXLAN_PROXY]))
3281			conf->flags |= VXLAN_F_PROXY;
 
3282	}
3283
3284	if (data[IFLA_VXLAN_RSC]) {
3285		if (changelink)
3286			return -EOPNOTSUPP;
3287		if (nla_get_u8(data[IFLA_VXLAN_RSC]))
3288			conf->flags |= VXLAN_F_RSC;
 
3289	}
3290
3291	if (data[IFLA_VXLAN_L2MISS]) {
3292		if (changelink)
3293			return -EOPNOTSUPP;
3294		if (nla_get_u8(data[IFLA_VXLAN_L2MISS]))
3295			conf->flags |= VXLAN_F_L2MISS;
 
3296	}
3297
3298	if (data[IFLA_VXLAN_L3MISS]) {
3299		if (changelink)
3300			return -EOPNOTSUPP;
3301		if (nla_get_u8(data[IFLA_VXLAN_L3MISS]))
3302			conf->flags |= VXLAN_F_L3MISS;
 
3303	}
3304
3305	if (data[IFLA_VXLAN_LIMIT]) {
3306		if (changelink)
 
 
3307			return -EOPNOTSUPP;
 
3308		conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3309	}
3310
3311	if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3312		if (changelink)
3313			return -EOPNOTSUPP;
3314		if (nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3315			conf->flags |= VXLAN_F_COLLECT_METADATA;
 
3316	}
3317
3318	if (data[IFLA_VXLAN_PORT_RANGE]) {
3319		if (!changelink) {
3320			const struct ifla_vxlan_port_range *p
3321				= nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3322			conf->port_min = ntohs(p->low);
3323			conf->port_max = ntohs(p->high);
3324		} else {
 
 
3325			return -EOPNOTSUPP;
3326		}
3327	}
3328
3329	if (data[IFLA_VXLAN_PORT]) {
3330		if (changelink)
 
 
3331			return -EOPNOTSUPP;
 
3332		conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3333	}
3334
3335	if (data[IFLA_VXLAN_UDP_CSUM]) {
3336		if (changelink)
 
 
3337			return -EOPNOTSUPP;
 
3338		if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3339			conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3340	}
3341
3342	if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3343		if (changelink)
3344			return -EOPNOTSUPP;
3345		if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
3346			conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
 
3347	}
3348
3349	if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3350		if (changelink)
3351			return -EOPNOTSUPP;
3352		if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
3353			conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
 
3354	}
3355
3356	if (data[IFLA_VXLAN_REMCSUM_TX]) {
3357		if (changelink)
3358			return -EOPNOTSUPP;
3359		if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
3360			conf->flags |= VXLAN_F_REMCSUM_TX;
 
3361	}
3362
3363	if (data[IFLA_VXLAN_REMCSUM_RX]) {
3364		if (changelink)
3365			return -EOPNOTSUPP;
3366		if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3367			conf->flags |= VXLAN_F_REMCSUM_RX;
 
3368	}
3369
3370	if (data[IFLA_VXLAN_GBP]) {
3371		if (changelink)
3372			return -EOPNOTSUPP;
3373		conf->flags |= VXLAN_F_GBP;
 
3374	}
3375
3376	if (data[IFLA_VXLAN_GPE]) {
3377		if (changelink)
3378			return -EOPNOTSUPP;
3379		conf->flags |= VXLAN_F_GPE;
 
 
3380	}
3381
3382	if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3383		if (changelink)
3384			return -EOPNOTSUPP;
3385		conf->flags |= VXLAN_F_REMCSUM_NOPARTIAL;
 
 
3386	}
3387
3388	if (tb[IFLA_MTU]) {
3389		if (changelink)
 
 
3390			return -EOPNOTSUPP;
 
3391		conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3392	}
3393
 
 
 
3394	return 0;
3395}
3396
3397static int vxlan_newlink(struct net *src_net, struct net_device *dev,
3398			 struct nlattr *tb[], struct nlattr *data[],
3399			 struct netlink_ext_ack *extack)
3400{
3401	struct vxlan_config conf;
3402	int err;
3403
3404	err = vxlan_nl2conf(tb, data, dev, &conf, false);
3405	if (err)
3406		return err;
3407
3408	return __vxlan_dev_create(src_net, dev, &conf, extack);
3409}
3410
3411static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
3412			    struct nlattr *data[],
3413			    struct netlink_ext_ack *extack)
3414{
3415	struct vxlan_dev *vxlan = netdev_priv(dev);
3416	struct vxlan_rdst *dst = &vxlan->default_dst;
3417	struct vxlan_rdst old_dst;
3418	struct vxlan_config conf;
 
3419	int err;
3420
3421	err = vxlan_nl2conf(tb, data,
3422			    dev, &conf, true);
3423	if (err)
3424		return err;
3425
3426	memcpy(&old_dst, dst, sizeof(struct vxlan_rdst));
 
 
 
 
 
 
3427
3428	err = vxlan_dev_configure(vxlan->net, dev, &conf, true, extack);
 
3429	if (err)
3430		return err;
3431
3432	/* handle default dst entry */
3433	if (!vxlan_addr_equal(&dst->remote_ip, &old_dst.remote_ip)) {
3434		spin_lock_bh(&vxlan->hash_lock);
3435		if (!vxlan_addr_any(&old_dst.remote_ip))
3436			__vxlan_fdb_delete(vxlan, all_zeros_mac,
3437					   old_dst.remote_ip,
3438					   vxlan->cfg.dst_port,
3439					   old_dst.remote_vni,
3440					   old_dst.remote_vni,
3441					   old_dst.remote_ifindex, 0);
3442
3443		if (!vxlan_addr_any(&dst->remote_ip)) {
3444			err = vxlan_fdb_create(vxlan, all_zeros_mac,
3445					       &dst->remote_ip,
3446					       NUD_REACHABLE | NUD_PERMANENT,
3447					       NLM_F_CREATE | NLM_F_APPEND,
3448					       vxlan->cfg.dst_port,
3449					       dst->remote_vni,
3450					       dst->remote_vni,
3451					       dst->remote_ifindex,
3452					       NTF_SELF);
3453			if (err) {
3454				spin_unlock_bh(&vxlan->hash_lock);
 
 
3455				return err;
3456			}
3457		}
3458		spin_unlock_bh(&vxlan->hash_lock);
 
 
 
 
 
 
 
 
3459	}
3460
 
 
 
 
 
 
 
3461	return 0;
3462}
3463
3464static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3465{
3466	struct vxlan_dev *vxlan = netdev_priv(dev);
3467
3468	vxlan_flush(vxlan, true);
3469
3470	gro_cells_destroy(&vxlan->gro_cells);
3471	list_del(&vxlan->next);
3472	unregister_netdevice_queue(dev, head);
 
 
3473}
3474
3475static size_t vxlan_get_size(const struct net_device *dev)
3476{
3477
3478	return nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_ID */
3479		nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
3480		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_LINK */
3481		nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
3482		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TTL */
 
3483		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TOS */
 
3484		nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
3485		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_LEARNING */
3486		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_PROXY */
3487		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_RSC */
3488		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_L2MISS */
3489		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_L3MISS */
3490		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_COLLECT_METADATA */
3491		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_AGEING */
3492		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_LIMIT */
3493		nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
3494		nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3495		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3496		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3497		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3498		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3499		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
3500		0;
3501}
3502
3503static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3504{
3505	const struct vxlan_dev *vxlan = netdev_priv(dev);
3506	const struct vxlan_rdst *dst = &vxlan->default_dst;
3507	struct ifla_vxlan_port_range ports = {
3508		.low =  htons(vxlan->cfg.port_min),
3509		.high = htons(vxlan->cfg.port_max),
3510	};
3511
3512	if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
3513		goto nla_put_failure;
3514
3515	if (!vxlan_addr_any(&dst->remote_ip)) {
3516		if (dst->remote_ip.sa.sa_family == AF_INET) {
3517			if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3518					    dst->remote_ip.sin.sin_addr.s_addr))
3519				goto nla_put_failure;
3520#if IS_ENABLED(CONFIG_IPV6)
3521		} else {
3522			if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3523					     &dst->remote_ip.sin6.sin6_addr))
3524				goto nla_put_failure;
3525#endif
3526		}
3527	}
3528
3529	if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
3530		goto nla_put_failure;
3531
3532	if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3533		if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
3534			if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
3535					    vxlan->cfg.saddr.sin.sin_addr.s_addr))
3536				goto nla_put_failure;
3537#if IS_ENABLED(CONFIG_IPV6)
3538		} else {
3539			if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
3540					     &vxlan->cfg.saddr.sin6.sin6_addr))
3541				goto nla_put_failure;
3542#endif
3543		}
3544	}
3545
3546	if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
 
 
3547	    nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
 
3548	    nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
3549	    nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3550			!!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
3551	    nla_put_u8(skb, IFLA_VXLAN_PROXY,
3552			!!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
3553	    nla_put_u8(skb, IFLA_VXLAN_RSC,
3554		       !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
3555	    nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3556			!!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
3557	    nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3558			!!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
3559	    nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3560		       !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
3561	    nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3562	    nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3563	    nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
3564	    nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
3565			!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
3566	    nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3567			!!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3568	    nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
3569			!!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3570	    nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3571			!!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
3572	    nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3573			!!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
3574		goto nla_put_failure;
3575
3576	if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3577		goto nla_put_failure;
3578
3579	if (vxlan->cfg.flags & VXLAN_F_GBP &&
3580	    nla_put_flag(skb, IFLA_VXLAN_GBP))
3581		goto nla_put_failure;
3582
3583	if (vxlan->cfg.flags & VXLAN_F_GPE &&
3584	    nla_put_flag(skb, IFLA_VXLAN_GPE))
3585		goto nla_put_failure;
3586
3587	if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3588	    nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3589		goto nla_put_failure;
3590
3591	return 0;
3592
3593nla_put_failure:
3594	return -EMSGSIZE;
3595}
3596
3597static struct net *vxlan_get_link_net(const struct net_device *dev)
3598{
3599	struct vxlan_dev *vxlan = netdev_priv(dev);
3600
3601	return vxlan->net;
3602}
3603
3604static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3605	.kind		= "vxlan",
3606	.maxtype	= IFLA_VXLAN_MAX,
3607	.policy		= vxlan_policy,
3608	.priv_size	= sizeof(struct vxlan_dev),
3609	.setup		= vxlan_setup,
3610	.validate	= vxlan_validate,
3611	.newlink	= vxlan_newlink,
3612	.changelink	= vxlan_changelink,
3613	.dellink	= vxlan_dellink,
3614	.get_size	= vxlan_get_size,
3615	.fill_info	= vxlan_fill_info,
3616	.get_link_net	= vxlan_get_link_net,
3617};
3618
3619struct net_device *vxlan_dev_create(struct net *net, const char *name,
3620				    u8 name_assign_type,
3621				    struct vxlan_config *conf)
3622{
3623	struct nlattr *tb[IFLA_MAX + 1];
3624	struct net_device *dev;
3625	int err;
3626
3627	memset(&tb, 0, sizeof(tb));
3628
3629	dev = rtnl_create_link(net, name, name_assign_type,
3630			       &vxlan_link_ops, tb);
3631	if (IS_ERR(dev))
3632		return dev;
3633
3634	err = __vxlan_dev_create(net, dev, conf, NULL);
3635	if (err < 0) {
3636		free_netdev(dev);
3637		return ERR_PTR(err);
3638	}
3639
3640	err = rtnl_configure_link(dev, NULL);
3641	if (err < 0) {
3642		LIST_HEAD(list_kill);
3643
3644		vxlan_dellink(dev, &list_kill);
3645		unregister_netdevice_many(&list_kill);
3646		return ERR_PTR(err);
3647	}
3648
3649	return dev;
3650}
3651EXPORT_SYMBOL_GPL(vxlan_dev_create);
3652
3653static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3654					     struct net_device *dev)
3655{
3656	struct vxlan_dev *vxlan, *next;
3657	LIST_HEAD(list_kill);
3658
3659	list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3660		struct vxlan_rdst *dst = &vxlan->default_dst;
3661
3662		/* In case we created vxlan device with carrier
3663		 * and we loose the carrier due to module unload
3664		 * we also need to remove vxlan device. In other
3665		 * cases, it's not necessary and remote_ifindex
3666		 * is 0 here, so no matches.
3667		 */
3668		if (dst->remote_ifindex == dev->ifindex)
3669			vxlan_dellink(vxlan->dev, &list_kill);
3670	}
3671
3672	unregister_netdevice_many(&list_kill);
3673}
3674
3675static int vxlan_netdevice_event(struct notifier_block *unused,
3676				 unsigned long event, void *ptr)
3677{
3678	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3679	struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
3680
3681	if (event == NETDEV_UNREGISTER) {
3682		vxlan_offload_rx_ports(dev, false);
 
3683		vxlan_handle_lowerdev_unregister(vn, dev);
3684	} else if (event == NETDEV_REGISTER) {
3685		vxlan_offload_rx_ports(dev, true);
 
3686	} else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
3687		   event == NETDEV_UDP_TUNNEL_DROP_INFO) {
3688		vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
3689	}
3690
3691	return NOTIFY_DONE;
3692}
3693
3694static struct notifier_block vxlan_notifier_block __read_mostly = {
3695	.notifier_call = vxlan_netdevice_event,
3696};
3697
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3698static __net_init int vxlan_init_net(struct net *net)
3699{
3700	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3701	unsigned int h;
3702
3703	INIT_LIST_HEAD(&vn->vxlan_list);
3704	spin_lock_init(&vn->sock_lock);
3705
3706	for (h = 0; h < PORT_HASH_SIZE; ++h)
3707		INIT_HLIST_HEAD(&vn->sock_list[h]);
3708
3709	return 0;
3710}
3711
3712static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
3713{
3714	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3715	struct vxlan_dev *vxlan, *next;
3716	struct net_device *dev, *aux;
3717	unsigned int h;
3718
3719	for_each_netdev_safe(net, dev, aux)
3720		if (dev->rtnl_link_ops == &vxlan_link_ops)
3721			unregister_netdevice_queue(dev, head);
3722
3723	list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3724		/* If vxlan->dev is in the same netns, it has already been added
3725		 * to the list by the previous loop.
3726		 */
3727		if (!net_eq(dev_net(vxlan->dev), net)) {
3728			gro_cells_destroy(&vxlan->gro_cells);
3729			unregister_netdevice_queue(vxlan->dev, head);
3730		}
3731	}
3732
3733	for (h = 0; h < PORT_HASH_SIZE; ++h)
3734		WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
3735}
3736
3737static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
3738{
3739	struct net *net;
3740	LIST_HEAD(list);
3741
3742	rtnl_lock();
3743	list_for_each_entry(net, net_list, exit_list)
 
 
3744		vxlan_destroy_tunnels(net, &list);
3745
3746	unregister_netdevice_many(&list);
3747	rtnl_unlock();
3748}
3749
3750static struct pernet_operations vxlan_net_ops = {
3751	.init = vxlan_init_net,
3752	.exit_batch = vxlan_exit_batch_net,
3753	.id   = &vxlan_net_id,
3754	.size = sizeof(struct vxlan_net),
3755};
3756
3757static int __init vxlan_init_module(void)
3758{
3759	int rc;
3760
3761	get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3762
3763	rc = register_pernet_subsys(&vxlan_net_ops);
3764	if (rc)
3765		goto out1;
3766
3767	rc = register_netdevice_notifier(&vxlan_notifier_block);
3768	if (rc)
3769		goto out2;
3770
3771	rc = rtnl_link_register(&vxlan_link_ops);
3772	if (rc)
3773		goto out3;
3774
 
 
 
 
3775	return 0;
 
 
3776out3:
3777	unregister_netdevice_notifier(&vxlan_notifier_block);
3778out2:
3779	unregister_pernet_subsys(&vxlan_net_ops);
3780out1:
3781	return rc;
3782}
3783late_initcall(vxlan_init_module);
3784
3785static void __exit vxlan_cleanup_module(void)
3786{
3787	rtnl_link_unregister(&vxlan_link_ops);
 
3788	unregister_netdevice_notifier(&vxlan_notifier_block);
3789	unregister_pernet_subsys(&vxlan_net_ops);
3790	/* rcu_barrier() is called by netns */
3791}
3792module_exit(vxlan_cleanup_module);
3793
3794MODULE_LICENSE("GPL");
3795MODULE_VERSION(VXLAN_VERSION);
3796MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3797MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3798MODULE_ALIAS_RTNL_LINK("vxlan");
v5.9
   1// SPDX-License-Identifier: GPL-2.0-only
   2/*
   3 * VXLAN: Virtual eXtensible Local Area Network
   4 *
   5 * Copyright (c) 2012-2013 Vyatta Inc.
 
 
 
 
   6 */
   7
   8#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
   9
  10#include <linux/kernel.h>
  11#include <linux/module.h>
  12#include <linux/errno.h>
  13#include <linux/slab.h>
  14#include <linux/udp.h>
  15#include <linux/igmp.h>
  16#include <linux/if_ether.h>
  17#include <linux/ethtool.h>
  18#include <net/arp.h>
  19#include <net/ndisc.h>
  20#include <net/ipv6_stubs.h>
  21#include <net/ip.h>
  22#include <net/icmp.h>
  23#include <net/rtnetlink.h>
  24#include <net/inet_ecn.h>
  25#include <net/net_namespace.h>
  26#include <net/netns/generic.h>
  27#include <net/tun_proto.h>
  28#include <net/vxlan.h>
  29#include <net/nexthop.h>
  30
  31#if IS_ENABLED(CONFIG_IPV6)
  32#include <net/ip6_tunnel.h>
  33#include <net/ip6_checksum.h>
  34#endif
  35
  36#define VXLAN_VERSION	"0.1"
  37
  38#define PORT_HASH_BITS	8
  39#define PORT_HASH_SIZE  (1<<PORT_HASH_BITS)
  40#define FDB_AGE_DEFAULT 300 /* 5 min */
  41#define FDB_AGE_INTERVAL (10 * HZ)	/* rescan interval */
  42
  43/* UDP port for VXLAN traffic.
  44 * The IANA assigned port is 4789, but the Linux default is 8472
  45 * for compatibility with early adopters.
  46 */
  47static unsigned short vxlan_port __read_mostly = 8472;
  48module_param_named(udp_port, vxlan_port, ushort, 0444);
  49MODULE_PARM_DESC(udp_port, "Destination UDP port");
  50
  51static bool log_ecn_error = true;
  52module_param(log_ecn_error, bool, 0644);
  53MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  54
  55static unsigned int vxlan_net_id;
  56static struct rtnl_link_ops vxlan_link_ops;
  57
  58static const u8 all_zeros_mac[ETH_ALEN + 2];
  59
  60static int vxlan_sock_add(struct vxlan_dev *vxlan);
  61
  62static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
  63
  64/* per-network namespace private data for this module */
  65struct vxlan_net {
  66	struct list_head  vxlan_list;
  67	struct hlist_head sock_list[PORT_HASH_SIZE];
  68	spinlock_t	  sock_lock;
  69};
  70
  71/* Forwarding table entry */
  72struct vxlan_fdb {
  73	struct hlist_node hlist;	/* linked list of entries */
  74	struct rcu_head	  rcu;
  75	unsigned long	  updated;	/* jiffies */
  76	unsigned long	  used;
  77	struct list_head  remotes;
  78	u8		  eth_addr[ETH_ALEN];
  79	u16		  state;	/* see ndm_state */
  80	__be32		  vni;
  81	u16		  flags;	/* see ndm_flags and below */
  82	struct list_head  nh_list;
  83	struct nexthop __rcu *nh;
  84	struct vxlan_dev  __rcu *vdev;
  85};
  86
  87#define NTF_VXLAN_ADDED_BY_USER 0x100
  88
  89/* salt for hash table */
  90static u32 vxlan_salt __read_mostly;
  91
  92static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
  93{
  94	return vs->flags & VXLAN_F_COLLECT_METADATA ||
  95	       ip_tunnel_collect_metadata();
  96}
  97
  98#if IS_ENABLED(CONFIG_IPV6)
  99static inline
 100bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
 101{
 102	if (a->sa.sa_family != b->sa.sa_family)
 103		return false;
 104	if (a->sa.sa_family == AF_INET6)
 105		return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
 106	else
 107		return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
 108}
 109
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 110static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
 111{
 112	if (nla_len(nla) >= sizeof(struct in6_addr)) {
 113		ip->sin6.sin6_addr = nla_get_in6_addr(nla);
 114		ip->sa.sa_family = AF_INET6;
 115		return 0;
 116	} else if (nla_len(nla) >= sizeof(__be32)) {
 117		ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
 118		ip->sa.sa_family = AF_INET;
 119		return 0;
 120	} else {
 121		return -EAFNOSUPPORT;
 122	}
 123}
 124
 125static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
 126			      const union vxlan_addr *ip)
 127{
 128	if (ip->sa.sa_family == AF_INET6)
 129		return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
 130	else
 131		return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
 132}
 133
 134#else /* !CONFIG_IPV6 */
 135
 136static inline
 137bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
 138{
 139	return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
 140}
 141
 
 
 
 
 
 
 
 
 
 
 142static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
 143{
 144	if (nla_len(nla) >= sizeof(struct in6_addr)) {
 145		return -EAFNOSUPPORT;
 146	} else if (nla_len(nla) >= sizeof(__be32)) {
 147		ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
 148		ip->sa.sa_family = AF_INET;
 149		return 0;
 150	} else {
 151		return -EAFNOSUPPORT;
 152	}
 153}
 154
 155static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
 156			      const union vxlan_addr *ip)
 157{
 158	return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
 159}
 160#endif
 161
 162/* Virtual Network hash table head */
 163static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
 164{
 165	return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
 166}
 167
 168/* Socket hash table head */
 169static inline struct hlist_head *vs_head(struct net *net, __be16 port)
 170{
 171	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
 172
 173	return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
 174}
 175
 176/* First remote destination for a forwarding entry.
 177 * Guaranteed to be non-NULL because remotes are never deleted.
 178 */
 179static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
 180{
 181	if (rcu_access_pointer(fdb->nh))
 182		return NULL;
 183	return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
 184}
 185
 186static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
 187{
 188	if (rcu_access_pointer(fdb->nh))
 189		return NULL;
 190	return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
 191}
 192
 193/* Find VXLAN socket based on network namespace, address family and UDP port
 194 * and enabled unshareable flags.
 195 */
 196static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
 197					  __be16 port, u32 flags, int ifindex)
 198{
 199	struct vxlan_sock *vs;
 200
 201	flags &= VXLAN_F_RCV_FLAGS;
 202
 203	hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
 204		if (inet_sk(vs->sock->sk)->inet_sport == port &&
 205		    vxlan_get_sk_family(vs) == family &&
 206		    vs->flags == flags &&
 207		    vs->sock->sk->sk_bound_dev_if == ifindex)
 208			return vs;
 209	}
 210	return NULL;
 211}
 212
 213static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
 214					   __be32 vni)
 215{
 216	struct vxlan_dev_node *node;
 217
 218	/* For flow based devices, map all packets to VNI 0 */
 219	if (vs->flags & VXLAN_F_COLLECT_METADATA)
 220		vni = 0;
 221
 222	hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
 223		if (node->vxlan->default_dst.remote_vni != vni)
 224			continue;
 225
 226		if (IS_ENABLED(CONFIG_IPV6)) {
 227			const struct vxlan_config *cfg = &node->vxlan->cfg;
 228
 229			if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
 230			    cfg->remote_ifindex != ifindex)
 231				continue;
 232		}
 233
 234		return node->vxlan;
 235	}
 236
 237	return NULL;
 238}
 239
 240/* Look up VNI in a per net namespace table */
 241static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
 242					__be32 vni, sa_family_t family,
 243					__be16 port, u32 flags)
 244{
 245	struct vxlan_sock *vs;
 246
 247	vs = vxlan_find_sock(net, family, port, flags, ifindex);
 248	if (!vs)
 249		return NULL;
 250
 251	return vxlan_vs_find_vni(vs, ifindex, vni);
 252}
 253
 254/* Fill in neighbour message in skbuff. */
 255static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
 256			  const struct vxlan_fdb *fdb,
 257			  u32 portid, u32 seq, int type, unsigned int flags,
 258			  const struct vxlan_rdst *rdst)
 259{
 260	unsigned long now = jiffies;
 261	struct nda_cacheinfo ci;
 262	bool send_ip, send_eth;
 263	struct nlmsghdr *nlh;
 264	struct nexthop *nh;
 265	struct ndmsg *ndm;
 266	int nh_family;
 267	u32 nh_id;
 268
 269	nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
 270	if (nlh == NULL)
 271		return -EMSGSIZE;
 272
 273	ndm = nlmsg_data(nlh);
 274	memset(ndm, 0, sizeof(*ndm));
 275
 276	send_eth = send_ip = true;
 277
 278	rcu_read_lock();
 279	nh = rcu_dereference(fdb->nh);
 280	if (nh) {
 281		nh_family = nexthop_get_family(nh);
 282		nh_id = nh->id;
 283	}
 284	rcu_read_unlock();
 285
 286	if (type == RTM_GETNEIGH) {
 287		if (rdst) {
 288			send_ip = !vxlan_addr_any(&rdst->remote_ip);
 289			ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
 290		} else if (nh) {
 291			ndm->ndm_family = nh_family;
 292		}
 293		send_eth = !is_zero_ether_addr(fdb->eth_addr);
 
 294	} else
 295		ndm->ndm_family	= AF_BRIDGE;
 296	ndm->ndm_state = fdb->state;
 297	ndm->ndm_ifindex = vxlan->dev->ifindex;
 298	ndm->ndm_flags = fdb->flags;
 299	if (rdst && rdst->offloaded)
 300		ndm->ndm_flags |= NTF_OFFLOADED;
 301	ndm->ndm_type = RTN_UNICAST;
 302
 303	if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
 304	    nla_put_s32(skb, NDA_LINK_NETNSID,
 305			peernet2id(dev_net(vxlan->dev), vxlan->net)))
 306		goto nla_put_failure;
 307
 308	if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
 309		goto nla_put_failure;
 310	if (nh) {
 311		if (nla_put_u32(skb, NDA_NH_ID, nh_id))
 312			goto nla_put_failure;
 313	} else if (rdst) {
 314		if (send_ip && vxlan_nla_put_addr(skb, NDA_DST,
 315						  &rdst->remote_ip))
 316			goto nla_put_failure;
 317
 318		if (rdst->remote_port &&
 319		    rdst->remote_port != vxlan->cfg.dst_port &&
 320		    nla_put_be16(skb, NDA_PORT, rdst->remote_port))
 321			goto nla_put_failure;
 322		if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
 323		    nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
 324			goto nla_put_failure;
 325		if (rdst->remote_ifindex &&
 326		    nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
 327			goto nla_put_failure;
 328	}
 329
 
 
 
 
 
 
 
 
 
 330	if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
 331	    nla_put_u32(skb, NDA_SRC_VNI,
 332			be32_to_cpu(fdb->vni)))
 333		goto nla_put_failure;
 
 
 
 334
 335	ci.ndm_used	 = jiffies_to_clock_t(now - fdb->used);
 336	ci.ndm_confirmed = 0;
 337	ci.ndm_updated	 = jiffies_to_clock_t(now - fdb->updated);
 338	ci.ndm_refcnt	 = 0;
 339
 340	if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
 341		goto nla_put_failure;
 342
 343	nlmsg_end(skb, nlh);
 344	return 0;
 345
 346nla_put_failure:
 347	nlmsg_cancel(skb, nlh);
 348	return -EMSGSIZE;
 349}
 350
 351static inline size_t vxlan_nlmsg_size(void)
 352{
 353	return NLMSG_ALIGN(sizeof(struct ndmsg))
 354		+ nla_total_size(ETH_ALEN) /* NDA_LLADDR */
 355		+ nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
 356		+ nla_total_size(sizeof(__be16)) /* NDA_PORT */
 357		+ nla_total_size(sizeof(__be32)) /* NDA_VNI */
 358		+ nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
 359		+ nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
 360		+ nla_total_size(sizeof(struct nda_cacheinfo));
 361}
 362
 363static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
 364			       struct vxlan_rdst *rd, int type)
 365{
 366	struct net *net = dev_net(vxlan->dev);
 367	struct sk_buff *skb;
 368	int err = -ENOBUFS;
 369
 370	skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
 371	if (skb == NULL)
 372		goto errout;
 373
 374	err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
 375	if (err < 0) {
 376		/* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
 377		WARN_ON(err == -EMSGSIZE);
 378		kfree_skb(skb);
 379		goto errout;
 380	}
 381
 382	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
 383	return;
 384errout:
 385	if (err < 0)
 386		rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
 387}
 388
 389static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
 390			    const struct vxlan_fdb *fdb,
 391			    const struct vxlan_rdst *rd,
 392			    struct netlink_ext_ack *extack,
 393			    struct switchdev_notifier_vxlan_fdb_info *fdb_info)
 394{
 395	fdb_info->info.dev = vxlan->dev;
 396	fdb_info->info.extack = extack;
 397	fdb_info->remote_ip = rd->remote_ip;
 398	fdb_info->remote_port = rd->remote_port;
 399	fdb_info->remote_vni = rd->remote_vni;
 400	fdb_info->remote_ifindex = rd->remote_ifindex;
 401	memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
 402	fdb_info->vni = fdb->vni;
 403	fdb_info->offloaded = rd->offloaded;
 404	fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
 405}
 406
 407static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
 408					      struct vxlan_fdb *fdb,
 409					      struct vxlan_rdst *rd,
 410					      bool adding,
 411					      struct netlink_ext_ack *extack)
 412{
 413	struct switchdev_notifier_vxlan_fdb_info info;
 414	enum switchdev_notifier_type notifier_type;
 415	int ret;
 416
 417	if (WARN_ON(!rd))
 418		return 0;
 419
 420	notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
 421			       : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
 422	vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
 423	ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
 424				       &info.info, extack);
 425	return notifier_to_errno(ret);
 426}
 427
 428static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
 429			    struct vxlan_rdst *rd, int type, bool swdev_notify,
 430			    struct netlink_ext_ack *extack)
 431{
 432	int err;
 433
 434	if (swdev_notify && rd) {
 435		switch (type) {
 436		case RTM_NEWNEIGH:
 437			err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
 438								 true, extack);
 439			if (err)
 440				return err;
 441			break;
 442		case RTM_DELNEIGH:
 443			vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
 444							   false, extack);
 445			break;
 446		}
 447	}
 448
 449	__vxlan_fdb_notify(vxlan, fdb, rd, type);
 450	return 0;
 451}
 452
 453static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
 454{
 455	struct vxlan_dev *vxlan = netdev_priv(dev);
 456	struct vxlan_fdb f = {
 457		.state = NUD_STALE,
 458	};
 459	struct vxlan_rdst remote = {
 460		.remote_ip = *ipa, /* goes to NDA_DST */
 461		.remote_vni = cpu_to_be32(VXLAN_N_VID),
 462	};
 463
 464	vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
 465}
 466
 467static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
 468{
 469	struct vxlan_fdb f = {
 470		.state = NUD_STALE,
 471	};
 472	struct vxlan_rdst remote = { };
 473
 474	memcpy(f.eth_addr, eth_addr, ETH_ALEN);
 475
 476	vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
 477}
 478
 479/* Hash Ethernet address */
 480static u32 eth_hash(const unsigned char *addr)
 481{
 482	u64 value = get_unaligned((u64 *)addr);
 483
 484	/* only want 6 bytes */
 485#ifdef __BIG_ENDIAN
 486	value >>= 16;
 487#else
 488	value <<= 16;
 489#endif
 490	return hash_64(value, FDB_HASH_BITS);
 491}
 492
 493static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
 494{
 495	/* use 1 byte of OUI and 3 bytes of NIC */
 496	u32 key = get_unaligned((u32 *)(addr + 2));
 497
 498	return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
 499}
 500
 501static u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
 502{
 503	if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
 504		return eth_vni_hash(mac, vni);
 505	else
 506		return eth_hash(mac);
 507}
 508
 509/* Hash chain to use given mac address */
 510static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
 511						const u8 *mac, __be32 vni)
 512{
 513	return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
 
 
 
 514}
 515
 516/* Look up Ethernet address in forwarding table */
 517static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
 518					  const u8 *mac, __be32 vni)
 519{
 520	struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
 521	struct vxlan_fdb *f;
 522
 523	hlist_for_each_entry_rcu(f, head, hlist) {
 524		if (ether_addr_equal(mac, f->eth_addr)) {
 525			if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
 526				if (vni == f->vni)
 527					return f;
 528			} else {
 529				return f;
 530			}
 531		}
 532	}
 533
 534	return NULL;
 535}
 536
 537static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
 538					const u8 *mac, __be32 vni)
 539{
 540	struct vxlan_fdb *f;
 541
 542	f = __vxlan_find_mac(vxlan, mac, vni);
 543	if (f && f->used != jiffies)
 544		f->used = jiffies;
 545
 546	return f;
 547}
 548
 549/* caller should hold vxlan->hash_lock */
 550static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
 551					      union vxlan_addr *ip, __be16 port,
 552					      __be32 vni, __u32 ifindex)
 553{
 554	struct vxlan_rdst *rd;
 555
 556	list_for_each_entry(rd, &f->remotes, list) {
 557		if (vxlan_addr_equal(&rd->remote_ip, ip) &&
 558		    rd->remote_port == port &&
 559		    rd->remote_vni == vni &&
 560		    rd->remote_ifindex == ifindex)
 561			return rd;
 562	}
 563
 564	return NULL;
 565}
 566
 567int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
 568		      struct switchdev_notifier_vxlan_fdb_info *fdb_info)
 569{
 570	struct vxlan_dev *vxlan = netdev_priv(dev);
 571	u8 eth_addr[ETH_ALEN + 2] = { 0 };
 572	struct vxlan_rdst *rdst;
 573	struct vxlan_fdb *f;
 574	int rc = 0;
 575
 576	if (is_multicast_ether_addr(mac) ||
 577	    is_zero_ether_addr(mac))
 578		return -EINVAL;
 579
 580	ether_addr_copy(eth_addr, mac);
 581
 582	rcu_read_lock();
 583
 584	f = __vxlan_find_mac(vxlan, eth_addr, vni);
 585	if (!f) {
 586		rc = -ENOENT;
 587		goto out;
 588	}
 589
 590	rdst = first_remote_rcu(f);
 591	vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
 592
 593out:
 594	rcu_read_unlock();
 595	return rc;
 596}
 597EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
 598
 599static int vxlan_fdb_notify_one(struct notifier_block *nb,
 600				const struct vxlan_dev *vxlan,
 601				const struct vxlan_fdb *f,
 602				const struct vxlan_rdst *rdst,
 603				struct netlink_ext_ack *extack)
 604{
 605	struct switchdev_notifier_vxlan_fdb_info fdb_info;
 606	int rc;
 607
 608	vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
 609	rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
 610			       &fdb_info);
 611	return notifier_to_errno(rc);
 612}
 613
 614int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
 615		     struct notifier_block *nb,
 616		     struct netlink_ext_ack *extack)
 617{
 618	struct vxlan_dev *vxlan;
 619	struct vxlan_rdst *rdst;
 620	struct vxlan_fdb *f;
 621	unsigned int h;
 622	int rc = 0;
 623
 624	if (!netif_is_vxlan(dev))
 625		return -EINVAL;
 626	vxlan = netdev_priv(dev);
 627
 628	for (h = 0; h < FDB_HASH_SIZE; ++h) {
 629		spin_lock_bh(&vxlan->hash_lock[h]);
 630		hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
 631			if (f->vni == vni) {
 632				list_for_each_entry(rdst, &f->remotes, list) {
 633					rc = vxlan_fdb_notify_one(nb, vxlan,
 634								  f, rdst,
 635								  extack);
 636					if (rc)
 637						goto unlock;
 638				}
 639			}
 640		}
 641		spin_unlock_bh(&vxlan->hash_lock[h]);
 642	}
 643	return 0;
 644
 645unlock:
 646	spin_unlock_bh(&vxlan->hash_lock[h]);
 647	return rc;
 648}
 649EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
 650
 651void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
 652{
 653	struct vxlan_dev *vxlan;
 654	struct vxlan_rdst *rdst;
 655	struct vxlan_fdb *f;
 656	unsigned int h;
 657
 658	if (!netif_is_vxlan(dev))
 659		return;
 660	vxlan = netdev_priv(dev);
 661
 662	for (h = 0; h < FDB_HASH_SIZE; ++h) {
 663		spin_lock_bh(&vxlan->hash_lock[h]);
 664		hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
 665			if (f->vni == vni)
 666				list_for_each_entry(rdst, &f->remotes, list)
 667					rdst->offloaded = false;
 668		spin_unlock_bh(&vxlan->hash_lock[h]);
 669	}
 670
 671}
 672EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
 673
 674/* Replace destination of unicast mac */
 675static int vxlan_fdb_replace(struct vxlan_fdb *f,
 676			     union vxlan_addr *ip, __be16 port, __be32 vni,
 677			     __u32 ifindex, struct vxlan_rdst *oldrd)
 678{
 679	struct vxlan_rdst *rd;
 680
 681	rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
 682	if (rd)
 683		return 0;
 684
 685	rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
 686	if (!rd)
 687		return 0;
 688
 689	*oldrd = *rd;
 690	dst_cache_reset(&rd->dst_cache);
 691	rd->remote_ip = *ip;
 692	rd->remote_port = port;
 693	rd->remote_vni = vni;
 694	rd->remote_ifindex = ifindex;
 695	rd->offloaded = false;
 696	return 1;
 697}
 698
 699/* Add/update destinations for multicast */
 700static int vxlan_fdb_append(struct vxlan_fdb *f,
 701			    union vxlan_addr *ip, __be16 port, __be32 vni,
 702			    __u32 ifindex, struct vxlan_rdst **rdp)
 703{
 704	struct vxlan_rdst *rd;
 705
 706	rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
 707	if (rd)
 708		return 0;
 709
 710	rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
 711	if (rd == NULL)
 712		return -ENOBUFS;
 713
 714	if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
 715		kfree(rd);
 716		return -ENOBUFS;
 717	}
 718
 719	rd->remote_ip = *ip;
 720	rd->remote_port = port;
 721	rd->offloaded = false;
 722	rd->remote_vni = vni;
 723	rd->remote_ifindex = ifindex;
 724
 725	list_add_tail_rcu(&rd->list, &f->remotes);
 726
 727	*rdp = rd;
 728	return 1;
 729}
 730
 731static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
 732					  unsigned int off,
 733					  struct vxlanhdr *vh, size_t hdrlen,
 734					  __be32 vni_field,
 735					  struct gro_remcsum *grc,
 736					  bool nopartial)
 737{
 738	size_t start, offset;
 739
 740	if (skb->remcsum_offload)
 741		return vh;
 742
 743	if (!NAPI_GRO_CB(skb)->csum_valid)
 744		return NULL;
 745
 746	start = vxlan_rco_start(vni_field);
 747	offset = start + vxlan_rco_offset(vni_field);
 748
 749	vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
 750				     start, offset, grc, nopartial);
 751
 752	skb->remcsum_offload = 1;
 753
 754	return vh;
 755}
 756
 757static struct sk_buff *vxlan_gro_receive(struct sock *sk,
 758					 struct list_head *head,
 759					 struct sk_buff *skb)
 760{
 761	struct sk_buff *pp = NULL;
 762	struct sk_buff *p;
 763	struct vxlanhdr *vh, *vh2;
 764	unsigned int hlen, off_vx;
 765	int flush = 1;
 766	struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
 767	__be32 flags;
 768	struct gro_remcsum grc;
 769
 770	skb_gro_remcsum_init(&grc);
 771
 772	off_vx = skb_gro_offset(skb);
 773	hlen = off_vx + sizeof(*vh);
 774	vh   = skb_gro_header_fast(skb, off_vx);
 775	if (skb_gro_header_hard(skb, hlen)) {
 776		vh = skb_gro_header_slow(skb, hlen, off_vx);
 777		if (unlikely(!vh))
 778			goto out;
 779	}
 780
 781	skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
 782
 783	flags = vh->vx_flags;
 784
 785	if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
 786		vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
 787				       vh->vx_vni, &grc,
 788				       !!(vs->flags &
 789					  VXLAN_F_REMCSUM_NOPARTIAL));
 790
 791		if (!vh)
 792			goto out;
 793	}
 794
 795	skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
 796
 797	list_for_each_entry(p, head, list) {
 798		if (!NAPI_GRO_CB(p)->same_flow)
 799			continue;
 800
 801		vh2 = (struct vxlanhdr *)(p->data + off_vx);
 802		if (vh->vx_flags != vh2->vx_flags ||
 803		    vh->vx_vni != vh2->vx_vni) {
 804			NAPI_GRO_CB(p)->same_flow = 0;
 805			continue;
 806		}
 807	}
 808
 809	pp = call_gro_receive(eth_gro_receive, head, skb);
 810	flush = 0;
 811
 812out:
 813	skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
 
 
 814
 815	return pp;
 816}
 817
 818static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
 819{
 820	/* Sets 'skb->inner_mac_header' since we are always called with
 821	 * 'skb->encapsulation' set.
 822	 */
 823	return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
 824}
 825
 826static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
 827					 __u16 state, __be32 src_vni,
 828					 __u16 ndm_flags)
 
 
 
 829{
 
 830	struct vxlan_fdb *f;
 
 
 831
 832	f = kmalloc(sizeof(*f), GFP_ATOMIC);
 833	if (!f)
 834		return NULL;
 835	f->state = state;
 836	f->flags = ndm_flags;
 837	f->updated = f->used = jiffies;
 838	f->vni = src_vni;
 839	f->nh = NULL;
 840	RCU_INIT_POINTER(f->vdev, vxlan);
 841	INIT_LIST_HEAD(&f->nh_list);
 842	INIT_LIST_HEAD(&f->remotes);
 843	memcpy(f->eth_addr, mac, ETH_ALEN);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 844
 845	return f;
 846}
 
 
 
 
 
 847
 848static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
 849			     __be32 src_vni, struct vxlan_fdb *f)
 850{
 851	++vxlan->addrcnt;
 852	hlist_add_head_rcu(&f->hlist,
 853			   vxlan_fdb_head(vxlan, mac, src_vni));
 854}
 
 855
 856static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
 857			       u32 nhid, struct netlink_ext_ack *extack)
 858{
 859	struct nexthop *old_nh = rtnl_dereference(fdb->nh);
 860	struct nexthop *nh;
 861	int err = -EINVAL;
 
 
 
 
 
 
 862
 863	if (old_nh && old_nh->id == nhid)
 864		return 0;
 
 
 
 865
 866	nh = nexthop_find_by_id(vxlan->net, nhid);
 867	if (!nh) {
 868		NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
 869		goto err_inval;
 870	}
 871
 872	if (nh) {
 873		if (!nexthop_get(nh)) {
 874			NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
 875			nh = NULL;
 876			goto err_inval;
 877		}
 878		if (!nexthop_is_fdb(nh)) {
 879			NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
 880			goto err_inval;
 881		}
 882
 883		if (!nexthop_is_multipath(nh)) {
 884			NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
 885			goto err_inval;
 886		}
 887
 888		/* check nexthop group family */
 889		switch (vxlan->default_dst.remote_ip.sa.sa_family) {
 890		case AF_INET:
 891			if (!nexthop_has_v4(nh)) {
 892				err = -EAFNOSUPPORT;
 893				NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
 894				goto err_inval;
 895			}
 896			break;
 897		case AF_INET6:
 898			if (nexthop_has_v4(nh)) {
 899				err = -EAFNOSUPPORT;
 900				NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
 901				goto err_inval;
 902			}
 903		}
 904	}
 905
 906	if (old_nh) {
 907		list_del_rcu(&fdb->nh_list);
 908		nexthop_put(old_nh);
 
 909	}
 910	rcu_assign_pointer(fdb->nh, nh);
 911	list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
 912	return 1;
 913
 914err_inval:
 915	if (nh)
 916		nexthop_put(nh);
 917	return err;
 918}
 919
 920static int vxlan_fdb_create(struct vxlan_dev *vxlan,
 921			    const u8 *mac, union vxlan_addr *ip,
 922			    __u16 state, __be16 port, __be32 src_vni,
 923			    __be32 vni, __u32 ifindex, __u16 ndm_flags,
 924			    u32 nhid, struct vxlan_fdb **fdb,
 925			    struct netlink_ext_ack *extack)
 926{
 927	struct vxlan_rdst *rd = NULL;
 928	struct vxlan_fdb *f;
 929	int rc;
 930
 931	if (vxlan->cfg.addrmax &&
 932	    vxlan->addrcnt >= vxlan->cfg.addrmax)
 933		return -ENOSPC;
 934
 935	netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
 936	f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
 937	if (!f)
 938		return -ENOMEM;
 939
 940	if (nhid)
 941		rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
 942	else
 943		rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
 944	if (rc < 0)
 945		goto errout;
 946
 947	*fdb = f;
 948
 949	return 0;
 950
 951errout:
 952	kfree(f);
 953	return rc;
 954}
 955
 956static void __vxlan_fdb_free(struct vxlan_fdb *f)
 957{
 
 958	struct vxlan_rdst *rd, *nd;
 959	struct nexthop *nh;
 960
 961	nh = rcu_dereference_raw(f->nh);
 962	if (nh) {
 963		rcu_assign_pointer(f->nh, NULL);
 964		rcu_assign_pointer(f->vdev, NULL);
 965		nexthop_put(nh);
 966	}
 967
 968	list_for_each_entry_safe(rd, nd, &f->remotes, list) {
 969		dst_cache_destroy(&rd->dst_cache);
 970		kfree(rd);
 971	}
 972	kfree(f);
 973}
 974
 975static void vxlan_fdb_free(struct rcu_head *head)
 976{
 977	struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
 978
 979	__vxlan_fdb_free(f);
 980}
 981
 982static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
 983			      bool do_notify, bool swdev_notify)
 984{
 985	struct vxlan_rdst *rd;
 986
 987	netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
 988
 989	--vxlan->addrcnt;
 990	if (do_notify) {
 991		if (rcu_access_pointer(f->nh))
 992			vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
 993					 swdev_notify, NULL);
 994		else
 995			list_for_each_entry(rd, &f->remotes, list)
 996				vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
 997						 swdev_notify, NULL);
 998	}
 999
1000	hlist_del_rcu(&f->hlist);
1001	list_del_rcu(&f->nh_list);
1002	call_rcu(&f->rcu, vxlan_fdb_free);
1003}
1004
1005static void vxlan_dst_free(struct rcu_head *head)
1006{
1007	struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
1008
1009	dst_cache_destroy(&rd->dst_cache);
1010	kfree(rd);
1011}
1012
1013static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
1014				     union vxlan_addr *ip,
1015				     __u16 state, __u16 flags,
1016				     __be16 port, __be32 vni,
1017				     __u32 ifindex, __u16 ndm_flags,
1018				     struct vxlan_fdb *f, u32 nhid,
1019				     bool swdev_notify,
1020				     struct netlink_ext_ack *extack)
1021{
1022	__u16 fdb_flags = (ndm_flags & ~NTF_USE);
1023	struct vxlan_rdst *rd = NULL;
1024	struct vxlan_rdst oldrd;
1025	int notify = 0;
1026	int rc = 0;
1027	int err;
1028
1029	if (nhid && !rcu_access_pointer(f->nh)) {
1030		NL_SET_ERR_MSG(extack,
1031			       "Cannot replace an existing non nexthop fdb with a nexthop");
1032		return -EOPNOTSUPP;
1033	}
1034
1035	if (nhid && (flags & NLM_F_APPEND)) {
1036		NL_SET_ERR_MSG(extack,
1037			       "Cannot append to a nexthop fdb");
1038		return -EOPNOTSUPP;
1039	}
1040
1041	/* Do not allow an externally learned entry to take over an entry added
1042	 * by the user.
1043	 */
1044	if (!(fdb_flags & NTF_EXT_LEARNED) ||
1045	    !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1046		if (f->state != state) {
1047			f->state = state;
1048			f->updated = jiffies;
1049			notify = 1;
1050		}
1051		if (f->flags != fdb_flags) {
1052			f->flags = fdb_flags;
1053			f->updated = jiffies;
1054			notify = 1;
1055		}
1056	}
1057
1058	if ((flags & NLM_F_REPLACE)) {
1059		/* Only change unicasts */
1060		if (!(is_multicast_ether_addr(f->eth_addr) ||
1061		      is_zero_ether_addr(f->eth_addr))) {
1062			if (nhid) {
1063				rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1064				if (rc < 0)
1065					return rc;
1066			} else {
1067				rc = vxlan_fdb_replace(f, ip, port, vni,
1068						       ifindex, &oldrd);
1069			}
1070			notify |= rc;
1071		} else {
1072			NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
1073			return -EOPNOTSUPP;
1074		}
1075	}
1076	if ((flags & NLM_F_APPEND) &&
1077	    (is_multicast_ether_addr(f->eth_addr) ||
1078	     is_zero_ether_addr(f->eth_addr))) {
1079		rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1080
1081		if (rc < 0)
1082			return rc;
1083		notify |= rc;
1084	}
1085
1086	if (ndm_flags & NTF_USE)
1087		f->used = jiffies;
1088
1089	if (notify) {
1090		if (rd == NULL)
1091			rd = first_remote_rtnl(f);
1092
1093		err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
1094				       swdev_notify, extack);
1095		if (err)
1096			goto err_notify;
1097	}
1098
1099	return 0;
1100
1101err_notify:
1102	if (nhid)
1103		return err;
1104	if ((flags & NLM_F_REPLACE) && rc)
1105		*rd = oldrd;
1106	else if ((flags & NLM_F_APPEND) && rc) {
1107		list_del_rcu(&rd->list);
1108		call_rcu(&rd->rcu, vxlan_dst_free);
1109	}
1110	return err;
1111}
1112
1113static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
1114				   const u8 *mac, union vxlan_addr *ip,
1115				   __u16 state, __u16 flags,
1116				   __be16 port, __be32 src_vni, __be32 vni,
1117				   __u32 ifindex, __u16 ndm_flags, u32 nhid,
1118				   bool swdev_notify,
1119				   struct netlink_ext_ack *extack)
1120{
1121	__u16 fdb_flags = (ndm_flags & ~NTF_USE);
1122	struct vxlan_fdb *f;
1123	int rc;
1124
1125	/* Disallow replace to add a multicast entry */
1126	if ((flags & NLM_F_REPLACE) &&
1127	    (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1128		return -EOPNOTSUPP;
1129
1130	netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1131	rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1132			      vni, ifindex, fdb_flags, nhid, &f, extack);
1133	if (rc < 0)
1134		return rc;
1135
1136	vxlan_fdb_insert(vxlan, mac, src_vni, f);
1137	rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1138			      swdev_notify, extack);
1139	if (rc)
1140		goto err_notify;
1141
1142	return 0;
1143
1144err_notify:
1145	vxlan_fdb_destroy(vxlan, f, false, false);
1146	return rc;
1147}
1148
1149/* Add new entry to forwarding table -- assumes lock held */
1150static int vxlan_fdb_update(struct vxlan_dev *vxlan,
1151			    const u8 *mac, union vxlan_addr *ip,
1152			    __u16 state, __u16 flags,
1153			    __be16 port, __be32 src_vni, __be32 vni,
1154			    __u32 ifindex, __u16 ndm_flags, u32 nhid,
1155			    bool swdev_notify,
1156			    struct netlink_ext_ack *extack)
1157{
1158	struct vxlan_fdb *f;
1159
1160	f = __vxlan_find_mac(vxlan, mac, src_vni);
1161	if (f) {
1162		if (flags & NLM_F_EXCL) {
1163			netdev_dbg(vxlan->dev,
1164				   "lost race to create %pM\n", mac);
1165			return -EEXIST;
1166		}
1167
1168		return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1169						 vni, ifindex, ndm_flags, f,
1170						 nhid, swdev_notify, extack);
1171	} else {
1172		if (!(flags & NLM_F_CREATE))
1173			return -ENOENT;
1174
1175		return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1176					       port, src_vni, vni, ifindex,
1177					       ndm_flags, nhid, swdev_notify,
1178					       extack);
1179	}
1180}
1181
1182static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1183				  struct vxlan_rdst *rd, bool swdev_notify)
1184{
1185	list_del_rcu(&rd->list);
1186	vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1187	call_rcu(&rd->rcu, vxlan_dst_free);
1188}
1189
1190static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1191			   union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1192			   __be32 *vni, u32 *ifindex, u32 *nhid)
1193{
1194	struct net *net = dev_net(vxlan->dev);
1195	int err;
1196
1197	if (tb[NDA_NH_ID] && (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] ||
1198	    tb[NDA_PORT]))
1199		return -EINVAL;
1200
1201	if (tb[NDA_DST]) {
1202		err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1203		if (err)
1204			return err;
1205	} else {
1206		union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1207
1208		if (remote->sa.sa_family == AF_INET) {
1209			ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1210			ip->sa.sa_family = AF_INET;
1211#if IS_ENABLED(CONFIG_IPV6)
1212		} else {
1213			ip->sin6.sin6_addr = in6addr_any;
1214			ip->sa.sa_family = AF_INET6;
1215#endif
1216		}
1217	}
1218
1219	if (tb[NDA_PORT]) {
1220		if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
1221			return -EINVAL;
1222		*port = nla_get_be16(tb[NDA_PORT]);
1223	} else {
1224		*port = vxlan->cfg.dst_port;
1225	}
1226
1227	if (tb[NDA_VNI]) {
1228		if (nla_len(tb[NDA_VNI]) != sizeof(u32))
1229			return -EINVAL;
1230		*vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1231	} else {
1232		*vni = vxlan->default_dst.remote_vni;
1233	}
1234
1235	if (tb[NDA_SRC_VNI]) {
1236		if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
1237			return -EINVAL;
1238		*src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1239	} else {
1240		*src_vni = vxlan->default_dst.remote_vni;
1241	}
1242
1243	if (tb[NDA_IFINDEX]) {
1244		struct net_device *tdev;
1245
1246		if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
1247			return -EINVAL;
1248		*ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1249		tdev = __dev_get_by_index(net, *ifindex);
1250		if (!tdev)
1251			return -EADDRNOTAVAIL;
1252	} else {
1253		*ifindex = 0;
1254	}
1255
1256	if (tb[NDA_NH_ID])
1257		*nhid = nla_get_u32(tb[NDA_NH_ID]);
1258	else
1259		*nhid = 0;
1260
1261	return 0;
1262}
1263
1264/* Add static entry (via netlink) */
1265static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1266			 struct net_device *dev,
1267			 const unsigned char *addr, u16 vid, u16 flags,
1268			 struct netlink_ext_ack *extack)
1269{
1270	struct vxlan_dev *vxlan = netdev_priv(dev);
1271	/* struct net *net = dev_net(vxlan->dev); */
1272	union vxlan_addr ip;
1273	__be16 port;
1274	__be32 src_vni, vni;
1275	u32 ifindex, nhid;
1276	u32 hash_index;
1277	int err;
1278
1279	if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1280		pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1281			ndm->ndm_state);
1282		return -EINVAL;
1283	}
1284
1285	if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
1286		return -EINVAL;
1287
1288	err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1289			      &nhid);
1290	if (err)
1291		return err;
1292
1293	if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1294		return -EAFNOSUPPORT;
1295
1296	hash_index = fdb_head_index(vxlan, addr, src_vni);
1297	spin_lock_bh(&vxlan->hash_lock[hash_index]);
1298	err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1299			       port, src_vni, vni, ifindex,
1300			       ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1301			       nhid, true, extack);
1302	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1303
1304	return err;
1305}
1306
1307static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1308			      const unsigned char *addr, union vxlan_addr ip,
1309			      __be16 port, __be32 src_vni, __be32 vni,
1310			      u32 ifindex, bool swdev_notify)
1311{
 
1312	struct vxlan_rdst *rd = NULL;
1313	struct vxlan_fdb *f;
1314	int err = -ENOENT;
1315
1316	f = vxlan_find_mac(vxlan, addr, src_vni);
1317	if (!f)
1318		return err;
1319
1320	if (!vxlan_addr_any(&ip)) {
1321		rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1322		if (!rd)
1323			goto out;
1324	}
1325
1326	/* remove a destination if it's not the only one on the list,
1327	 * otherwise destroy the fdb entry
1328	 */
1329	if (rd && !list_is_singular(&f->remotes)) {
1330		vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1331		goto out;
1332	}
1333
1334	vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1335
1336out:
1337	return 0;
1338}
1339
1340/* Delete entry (via netlink) */
1341static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1342			    struct net_device *dev,
1343			    const unsigned char *addr, u16 vid)
1344{
1345	struct vxlan_dev *vxlan = netdev_priv(dev);
1346	union vxlan_addr ip;
1347	__be32 src_vni, vni;
1348	u32 ifindex, nhid;
1349	u32 hash_index;
1350	__be16 port;
 
1351	int err;
1352
1353	err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1354			      &nhid);
1355	if (err)
1356		return err;
1357
1358	hash_index = fdb_head_index(vxlan, addr, src_vni);
1359	spin_lock_bh(&vxlan->hash_lock[hash_index]);
1360	err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1361				 true);
1362	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1363
1364	return err;
1365}
1366
1367/* Dump forwarding table */
1368static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1369			  struct net_device *dev,
1370			  struct net_device *filter_dev, int *idx)
1371{
1372	struct vxlan_dev *vxlan = netdev_priv(dev);
1373	unsigned int h;
1374	int err = 0;
1375
1376	for (h = 0; h < FDB_HASH_SIZE; ++h) {
1377		struct vxlan_fdb *f;
1378
1379		rcu_read_lock();
1380		hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1381			struct vxlan_rdst *rd;
1382
1383			if (rcu_access_pointer(f->nh)) {
1384				if (*idx < cb->args[2])
1385					goto skip_nh;
1386				err = vxlan_fdb_info(skb, vxlan, f,
1387						     NETLINK_CB(cb->skb).portid,
1388						     cb->nlh->nlmsg_seq,
1389						     RTM_NEWNEIGH,
1390						     NLM_F_MULTI, NULL);
1391				if (err < 0) {
1392					rcu_read_unlock();
1393					goto out;
1394				}
1395skip_nh:
1396				*idx += 1;
1397				continue;
1398			}
1399
1400			list_for_each_entry_rcu(rd, &f->remotes, list) {
1401				if (*idx < cb->args[2])
1402					goto skip;
1403
1404				err = vxlan_fdb_info(skb, vxlan, f,
1405						     NETLINK_CB(cb->skb).portid,
1406						     cb->nlh->nlmsg_seq,
1407						     RTM_NEWNEIGH,
1408						     NLM_F_MULTI, rd);
1409				if (err < 0) {
1410					rcu_read_unlock();
1411					goto out;
1412				}
1413skip:
1414				*idx += 1;
1415			}
1416		}
1417		rcu_read_unlock();
1418	}
1419out:
1420	return err;
1421}
1422
1423static int vxlan_fdb_get(struct sk_buff *skb,
1424			 struct nlattr *tb[],
1425			 struct net_device *dev,
1426			 const unsigned char *addr,
1427			 u16 vid, u32 portid, u32 seq,
1428			 struct netlink_ext_ack *extack)
1429{
1430	struct vxlan_dev *vxlan = netdev_priv(dev);
1431	struct vxlan_fdb *f;
1432	__be32 vni;
1433	int err;
1434
1435	if (tb[NDA_VNI])
1436		vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1437	else
1438		vni = vxlan->default_dst.remote_vni;
1439
1440	rcu_read_lock();
1441
1442	f = __vxlan_find_mac(vxlan, addr, vni);
1443	if (!f) {
1444		NL_SET_ERR_MSG(extack, "Fdb entry not found");
1445		err = -ENOENT;
1446		goto errout;
1447	}
1448
1449	err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1450			     RTM_NEWNEIGH, 0, first_remote_rcu(f));
1451errout:
1452	rcu_read_unlock();
1453	return err;
1454}
1455
1456/* Watch incoming packets to learn mapping between Ethernet address
1457 * and Tunnel endpoint.
1458 * Return true if packet is bogus and should be dropped.
1459 */
1460static bool vxlan_snoop(struct net_device *dev,
1461			union vxlan_addr *src_ip, const u8 *src_mac,
1462			u32 src_ifindex, __be32 vni)
1463{
1464	struct vxlan_dev *vxlan = netdev_priv(dev);
1465	struct vxlan_fdb *f;
1466	u32 ifindex = 0;
1467
1468#if IS_ENABLED(CONFIG_IPV6)
1469	if (src_ip->sa.sa_family == AF_INET6 &&
1470	    (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1471		ifindex = src_ifindex;
1472#endif
1473
1474	f = vxlan_find_mac(vxlan, src_mac, vni);
1475	if (likely(f)) {
1476		struct vxlan_rdst *rdst = first_remote_rcu(f);
1477
1478		if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1479			   rdst->remote_ifindex == ifindex))
1480			return false;
1481
1482		/* Don't migrate static entries, drop packets */
1483		if (f->state & (NUD_PERMANENT | NUD_NOARP))
1484			return true;
1485
1486		/* Don't override an fdb with nexthop with a learnt entry */
1487		if (rcu_access_pointer(f->nh))
1488			return true;
1489
1490		if (net_ratelimit())
1491			netdev_info(dev,
1492				    "%pM migrated from %pIS to %pIS\n",
1493				    src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1494
1495		rdst->remote_ip = *src_ip;
1496		f->updated = jiffies;
1497		vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1498	} else {
1499		u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1500
1501		/* learned new entry */
1502		spin_lock(&vxlan->hash_lock[hash_index]);
1503
1504		/* close off race between vxlan_flush and incoming packets */
1505		if (netif_running(dev))
1506			vxlan_fdb_update(vxlan, src_mac, src_ip,
1507					 NUD_REACHABLE,
1508					 NLM_F_EXCL|NLM_F_CREATE,
1509					 vxlan->cfg.dst_port,
1510					 vni,
1511					 vxlan->default_dst.remote_vni,
1512					 ifindex, NTF_SELF, 0, true, NULL);
1513		spin_unlock(&vxlan->hash_lock[hash_index]);
1514	}
1515
1516	return false;
1517}
1518
1519/* See if multicast group is already in use by other ID */
1520static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1521{
1522	struct vxlan_dev *vxlan;
1523	struct vxlan_sock *sock4;
1524#if IS_ENABLED(CONFIG_IPV6)
1525	struct vxlan_sock *sock6;
1526#endif
1527	unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1528
1529	sock4 = rtnl_dereference(dev->vn4_sock);
1530
1531	/* The vxlan_sock is only used by dev, leaving group has
1532	 * no effect on other vxlan devices.
1533	 */
1534	if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
1535		return false;
1536#if IS_ENABLED(CONFIG_IPV6)
1537	sock6 = rtnl_dereference(dev->vn6_sock);
1538	if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
1539		return false;
1540#endif
1541
1542	list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1543		if (!netif_running(vxlan->dev) || vxlan == dev)
1544			continue;
1545
1546		if (family == AF_INET &&
1547		    rtnl_dereference(vxlan->vn4_sock) != sock4)
1548			continue;
1549#if IS_ENABLED(CONFIG_IPV6)
1550		if (family == AF_INET6 &&
1551		    rtnl_dereference(vxlan->vn6_sock) != sock6)
1552			continue;
1553#endif
1554
1555		if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1556				      &dev->default_dst.remote_ip))
1557			continue;
1558
1559		if (vxlan->default_dst.remote_ifindex !=
1560		    dev->default_dst.remote_ifindex)
1561			continue;
1562
1563		return true;
1564	}
1565
1566	return false;
1567}
1568
1569static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1570{
1571	struct vxlan_net *vn;
1572
1573	if (!vs)
1574		return false;
1575	if (!refcount_dec_and_test(&vs->refcnt))
1576		return false;
1577
1578	vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1579	spin_lock(&vn->sock_lock);
1580	hlist_del_rcu(&vs->hlist);
1581	udp_tunnel_notify_del_rx_port(vs->sock,
1582				      (vs->flags & VXLAN_F_GPE) ?
1583				      UDP_TUNNEL_TYPE_VXLAN_GPE :
1584				      UDP_TUNNEL_TYPE_VXLAN);
1585	spin_unlock(&vn->sock_lock);
1586
1587	return true;
1588}
1589
1590static void vxlan_sock_release(struct vxlan_dev *vxlan)
1591{
1592	struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1593#if IS_ENABLED(CONFIG_IPV6)
1594	struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1595
1596	RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1597#endif
1598
1599	RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1600	synchronize_net();
1601
1602	vxlan_vs_del_dev(vxlan);
1603
1604	if (__vxlan_sock_release_prep(sock4)) {
1605		udp_tunnel_sock_release(sock4->sock);
1606		kfree(sock4);
1607	}
1608
1609#if IS_ENABLED(CONFIG_IPV6)
1610	if (__vxlan_sock_release_prep(sock6)) {
1611		udp_tunnel_sock_release(sock6->sock);
1612		kfree(sock6);
1613	}
1614#endif
1615}
1616
1617/* Update multicast group membership when first VNI on
1618 * multicast address is brought up
1619 */
1620static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1621{
1622	struct sock *sk;
1623	union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1624	int ifindex = vxlan->default_dst.remote_ifindex;
1625	int ret = -EINVAL;
1626
1627	if (ip->sa.sa_family == AF_INET) {
1628		struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1629		struct ip_mreqn mreq = {
1630			.imr_multiaddr.s_addr	= ip->sin.sin_addr.s_addr,
1631			.imr_ifindex		= ifindex,
1632		};
1633
1634		sk = sock4->sock->sk;
1635		lock_sock(sk);
1636		ret = ip_mc_join_group(sk, &mreq);
1637		release_sock(sk);
1638#if IS_ENABLED(CONFIG_IPV6)
1639	} else {
1640		struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1641
1642		sk = sock6->sock->sk;
1643		lock_sock(sk);
1644		ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1645						   &ip->sin6.sin6_addr);
1646		release_sock(sk);
1647#endif
1648	}
1649
1650	return ret;
1651}
1652
1653/* Inverse of vxlan_igmp_join when last VNI is brought down */
1654static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1655{
1656	struct sock *sk;
1657	union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1658	int ifindex = vxlan->default_dst.remote_ifindex;
1659	int ret = -EINVAL;
1660
1661	if (ip->sa.sa_family == AF_INET) {
1662		struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1663		struct ip_mreqn mreq = {
1664			.imr_multiaddr.s_addr	= ip->sin.sin_addr.s_addr,
1665			.imr_ifindex		= ifindex,
1666		};
1667
1668		sk = sock4->sock->sk;
1669		lock_sock(sk);
1670		ret = ip_mc_leave_group(sk, &mreq);
1671		release_sock(sk);
1672#if IS_ENABLED(CONFIG_IPV6)
1673	} else {
1674		struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1675
1676		sk = sock6->sock->sk;
1677		lock_sock(sk);
1678		ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1679						   &ip->sin6.sin6_addr);
1680		release_sock(sk);
1681#endif
1682	}
1683
1684	return ret;
1685}
1686
1687static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1688			  struct sk_buff *skb, u32 vxflags)
1689{
1690	size_t start, offset;
1691
1692	if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1693		goto out;
1694
1695	start = vxlan_rco_start(unparsed->vx_vni);
1696	offset = start + vxlan_rco_offset(unparsed->vx_vni);
1697
1698	if (!pskb_may_pull(skb, offset + sizeof(u16)))
1699		return false;
1700
1701	skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1702			    !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1703out:
1704	unparsed->vx_flags &= ~VXLAN_HF_RCO;
1705	unparsed->vx_vni &= VXLAN_VNI_MASK;
1706	return true;
1707}
1708
1709static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1710				struct sk_buff *skb, u32 vxflags,
1711				struct vxlan_metadata *md)
1712{
1713	struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1714	struct metadata_dst *tun_dst;
1715
1716	if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1717		goto out;
1718
1719	md->gbp = ntohs(gbp->policy_id);
1720
1721	tun_dst = (struct metadata_dst *)skb_dst(skb);
1722	if (tun_dst) {
1723		tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1724		tun_dst->u.tun_info.options_len = sizeof(*md);
1725	}
1726	if (gbp->dont_learn)
1727		md->gbp |= VXLAN_GBP_DONT_LEARN;
1728
1729	if (gbp->policy_applied)
1730		md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1731
1732	/* In flow-based mode, GBP is carried in dst_metadata */
1733	if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1734		skb->mark = md->gbp;
1735out:
1736	unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1737}
1738
1739static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1740				__be16 *protocol,
1741				struct sk_buff *skb, u32 vxflags)
1742{
1743	struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1744
1745	/* Need to have Next Protocol set for interfaces in GPE mode. */
1746	if (!gpe->np_applied)
1747		return false;
1748	/* "The initial version is 0. If a receiver does not support the
1749	 * version indicated it MUST drop the packet.
1750	 */
1751	if (gpe->version != 0)
1752		return false;
1753	/* "When the O bit is set to 1, the packet is an OAM packet and OAM
1754	 * processing MUST occur." However, we don't implement OAM
1755	 * processing, thus drop the packet.
1756	 */
1757	if (gpe->oam_flag)
1758		return false;
1759
1760	*protocol = tun_p_to_eth_p(gpe->next_protocol);
1761	if (!*protocol)
1762		return false;
1763
1764	unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1765	return true;
1766}
1767
1768static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1769			  struct vxlan_sock *vs,
1770			  struct sk_buff *skb, __be32 vni)
1771{
1772	union vxlan_addr saddr;
1773	u32 ifindex = skb->dev->ifindex;
1774
1775	skb_reset_mac_header(skb);
1776	skb->protocol = eth_type_trans(skb, vxlan->dev);
1777	skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1778
1779	/* Ignore packet loops (and multicast echo) */
1780	if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1781		return false;
1782
1783	/* Get address from the outer IP header */
1784	if (vxlan_get_sk_family(vs) == AF_INET) {
1785		saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1786		saddr.sa.sa_family = AF_INET;
1787#if IS_ENABLED(CONFIG_IPV6)
1788	} else {
1789		saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1790		saddr.sa.sa_family = AF_INET6;
1791#endif
1792	}
1793
1794	if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1795	    vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1796		return false;
1797
1798	return true;
1799}
1800
1801static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1802				  struct sk_buff *skb)
1803{
1804	int err = 0;
1805
1806	if (vxlan_get_sk_family(vs) == AF_INET)
1807		err = IP_ECN_decapsulate(oiph, skb);
1808#if IS_ENABLED(CONFIG_IPV6)
1809	else
1810		err = IP6_ECN_decapsulate(oiph, skb);
1811#endif
1812
1813	if (unlikely(err) && log_ecn_error) {
1814		if (vxlan_get_sk_family(vs) == AF_INET)
1815			net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1816					     &((struct iphdr *)oiph)->saddr,
1817					     ((struct iphdr *)oiph)->tos);
1818		else
1819			net_info_ratelimited("non-ECT from %pI6\n",
1820					     &((struct ipv6hdr *)oiph)->saddr);
1821	}
1822	return err <= 1;
1823}
1824
1825/* Callback from net/ipv4/udp.c to receive packets */
1826static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1827{
1828	struct pcpu_sw_netstats *stats;
1829	struct vxlan_dev *vxlan;
1830	struct vxlan_sock *vs;
1831	struct vxlanhdr unparsed;
1832	struct vxlan_metadata _md;
1833	struct vxlan_metadata *md = &_md;
1834	__be16 protocol = htons(ETH_P_TEB);
1835	bool raw_proto = false;
1836	void *oiph;
1837	__be32 vni = 0;
1838
1839	/* Need UDP and VXLAN header to be present */
1840	if (!pskb_may_pull(skb, VXLAN_HLEN))
1841		goto drop;
1842
1843	unparsed = *vxlan_hdr(skb);
1844	/* VNI flag always required to be set */
1845	if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1846		netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1847			   ntohl(vxlan_hdr(skb)->vx_flags),
1848			   ntohl(vxlan_hdr(skb)->vx_vni));
1849		/* Return non vxlan pkt */
1850		goto drop;
1851	}
1852	unparsed.vx_flags &= ~VXLAN_HF_VNI;
1853	unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1854
1855	vs = rcu_dereference_sk_user_data(sk);
1856	if (!vs)
1857		goto drop;
1858
1859	vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1860
1861	vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1862	if (!vxlan)
1863		goto drop;
1864
1865	/* For backwards compatibility, only allow reserved fields to be
1866	 * used by VXLAN extensions if explicitly requested.
1867	 */
1868	if (vs->flags & VXLAN_F_GPE) {
1869		if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1870			goto drop;
1871		raw_proto = true;
1872	}
1873
1874	if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1875				   !net_eq(vxlan->net, dev_net(vxlan->dev))))
1876		goto drop;
1877
1878	if (vxlan_collect_metadata(vs)) {
1879		struct metadata_dst *tun_dst;
1880
1881		tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1882					 key32_to_tunnel_id(vni), sizeof(*md));
1883
1884		if (!tun_dst)
1885			goto drop;
1886
1887		md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1888
1889		skb_dst_set(skb, (struct dst_entry *)tun_dst);
1890	} else {
1891		memset(md, 0, sizeof(*md));
1892	}
1893
1894	if (vs->flags & VXLAN_F_REMCSUM_RX)
1895		if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1896			goto drop;
1897	if (vs->flags & VXLAN_F_GBP)
1898		vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1899	/* Note that GBP and GPE can never be active together. This is
1900	 * ensured in vxlan_dev_configure.
1901	 */
1902
1903	if (unparsed.vx_flags || unparsed.vx_vni) {
1904		/* If there are any unprocessed flags remaining treat
1905		 * this as a malformed packet. This behavior diverges from
1906		 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1907		 * in reserved fields are to be ignored. The approach here
1908		 * maintains compatibility with previous stack code, and also
1909		 * is more robust and provides a little more security in
1910		 * adding extensions to VXLAN.
1911		 */
1912		goto drop;
1913	}
1914
1915	if (!raw_proto) {
1916		if (!vxlan_set_mac(vxlan, vs, skb, vni))
1917			goto drop;
1918	} else {
1919		skb_reset_mac_header(skb);
1920		skb->dev = vxlan->dev;
1921		skb->pkt_type = PACKET_HOST;
1922	}
1923
1924	oiph = skb_network_header(skb);
1925	skb_reset_network_header(skb);
1926
1927	if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1928		++vxlan->dev->stats.rx_frame_errors;
1929		++vxlan->dev->stats.rx_errors;
1930		goto drop;
1931	}
1932
1933	rcu_read_lock();
1934
1935	if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1936		rcu_read_unlock();
1937		atomic_long_inc(&vxlan->dev->rx_dropped);
1938		goto drop;
1939	}
1940
1941	stats = this_cpu_ptr(vxlan->dev->tstats);
1942	u64_stats_update_begin(&stats->syncp);
1943	stats->rx_packets++;
1944	stats->rx_bytes += skb->len;
1945	u64_stats_update_end(&stats->syncp);
1946
1947	gro_cells_receive(&vxlan->gro_cells, skb);
1948
1949	rcu_read_unlock();
1950
1951	return 0;
1952
1953drop:
1954	/* Consume bad packet */
1955	kfree_skb(skb);
1956	return 0;
1957}
1958
1959/* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1960static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1961{
1962	struct vxlan_dev *vxlan;
1963	struct vxlan_sock *vs;
1964	struct vxlanhdr *hdr;
1965	__be32 vni;
1966
1967	if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1968		return -EINVAL;
1969
1970	hdr = vxlan_hdr(skb);
1971
1972	if (!(hdr->vx_flags & VXLAN_HF_VNI))
1973		return -EINVAL;
1974
1975	vs = rcu_dereference_sk_user_data(sk);
1976	if (!vs)
1977		return -ENOENT;
1978
1979	vni = vxlan_vni(hdr->vx_vni);
1980	vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1981	if (!vxlan)
1982		return -ENOENT;
1983
1984	return 0;
1985}
1986
1987static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1988{
1989	struct vxlan_dev *vxlan = netdev_priv(dev);
1990	struct arphdr *parp;
1991	u8 *arpptr, *sha;
1992	__be32 sip, tip;
1993	struct neighbour *n;
1994
1995	if (dev->flags & IFF_NOARP)
1996		goto out;
1997
1998	if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1999		dev->stats.tx_dropped++;
2000		goto out;
2001	}
2002	parp = arp_hdr(skb);
2003
2004	if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
2005	     parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
2006	    parp->ar_pro != htons(ETH_P_IP) ||
2007	    parp->ar_op != htons(ARPOP_REQUEST) ||
2008	    parp->ar_hln != dev->addr_len ||
2009	    parp->ar_pln != 4)
2010		goto out;
2011	arpptr = (u8 *)parp + sizeof(struct arphdr);
2012	sha = arpptr;
2013	arpptr += dev->addr_len;	/* sha */
2014	memcpy(&sip, arpptr, sizeof(sip));
2015	arpptr += sizeof(sip);
2016	arpptr += dev->addr_len;	/* tha */
2017	memcpy(&tip, arpptr, sizeof(tip));
2018
2019	if (ipv4_is_loopback(tip) ||
2020	    ipv4_is_multicast(tip))
2021		goto out;
2022
2023	n = neigh_lookup(&arp_tbl, &tip, dev);
2024
2025	if (n) {
2026		struct vxlan_fdb *f;
2027		struct sk_buff	*reply;
2028
2029		if (!(n->nud_state & NUD_CONNECTED)) {
2030			neigh_release(n);
2031			goto out;
2032		}
2033
2034		f = vxlan_find_mac(vxlan, n->ha, vni);
2035		if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2036			/* bridge-local neighbor */
2037			neigh_release(n);
2038			goto out;
2039		}
2040
2041		reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
2042				n->ha, sha);
2043
2044		neigh_release(n);
2045
2046		if (reply == NULL)
2047			goto out;
2048
2049		skb_reset_mac_header(reply);
2050		__skb_pull(reply, skb_network_offset(reply));
2051		reply->ip_summed = CHECKSUM_UNNECESSARY;
2052		reply->pkt_type = PACKET_HOST;
2053
2054		if (netif_rx_ni(reply) == NET_RX_DROP)
2055			dev->stats.rx_dropped++;
2056	} else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2057		union vxlan_addr ipa = {
2058			.sin.sin_addr.s_addr = tip,
2059			.sin.sin_family = AF_INET,
2060		};
2061
2062		vxlan_ip_miss(dev, &ipa);
2063	}
2064out:
2065	consume_skb(skb);
2066	return NETDEV_TX_OK;
2067}
2068
2069#if IS_ENABLED(CONFIG_IPV6)
2070static struct sk_buff *vxlan_na_create(struct sk_buff *request,
2071	struct neighbour *n, bool isrouter)
2072{
2073	struct net_device *dev = request->dev;
2074	struct sk_buff *reply;
2075	struct nd_msg *ns, *na;
2076	struct ipv6hdr *pip6;
2077	u8 *daddr;
2078	int na_olen = 8; /* opt hdr + ETH_ALEN for target */
2079	int ns_olen;
2080	int i, len;
2081
2082	if (dev == NULL || !pskb_may_pull(request, request->len))
2083		return NULL;
2084
2085	len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
2086		sizeof(*na) + na_olen + dev->needed_tailroom;
2087	reply = alloc_skb(len, GFP_ATOMIC);
2088	if (reply == NULL)
2089		return NULL;
2090
2091	reply->protocol = htons(ETH_P_IPV6);
2092	reply->dev = dev;
2093	skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
2094	skb_push(reply, sizeof(struct ethhdr));
2095	skb_reset_mac_header(reply);
2096
2097	ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
2098
2099	daddr = eth_hdr(request)->h_source;
2100	ns_olen = request->len - skb_network_offset(request) -
2101		sizeof(struct ipv6hdr) - sizeof(*ns);
2102	for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
2103		if (!ns->opt[i + 1]) {
2104			kfree_skb(reply);
2105			return NULL;
2106		}
2107		if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
2108			daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
2109			break;
2110		}
2111	}
2112
2113	/* Ethernet header */
2114	ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
2115	ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
2116	eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
2117	reply->protocol = htons(ETH_P_IPV6);
2118
2119	skb_pull(reply, sizeof(struct ethhdr));
2120	skb_reset_network_header(reply);
2121	skb_put(reply, sizeof(struct ipv6hdr));
2122
2123	/* IPv6 header */
2124
2125	pip6 = ipv6_hdr(reply);
2126	memset(pip6, 0, sizeof(struct ipv6hdr));
2127	pip6->version = 6;
2128	pip6->priority = ipv6_hdr(request)->priority;
2129	pip6->nexthdr = IPPROTO_ICMPV6;
2130	pip6->hop_limit = 255;
2131	pip6->daddr = ipv6_hdr(request)->saddr;
2132	pip6->saddr = *(struct in6_addr *)n->primary_key;
2133
2134	skb_pull(reply, sizeof(struct ipv6hdr));
2135	skb_reset_transport_header(reply);
2136
2137	/* Neighbor Advertisement */
2138	na = skb_put_zero(reply, sizeof(*na) + na_olen);
2139	na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
2140	na->icmph.icmp6_router = isrouter;
2141	na->icmph.icmp6_override = 1;
2142	na->icmph.icmp6_solicited = 1;
2143	na->target = ns->target;
2144	ether_addr_copy(&na->opt[2], n->ha);
2145	na->opt[0] = ND_OPT_TARGET_LL_ADDR;
2146	na->opt[1] = na_olen >> 3;
2147
2148	na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
2149		&pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
2150		csum_partial(na, sizeof(*na)+na_olen, 0));
2151
2152	pip6->payload_len = htons(sizeof(*na)+na_olen);
2153
2154	skb_push(reply, sizeof(struct ipv6hdr));
2155
2156	reply->ip_summed = CHECKSUM_UNNECESSARY;
2157
2158	return reply;
2159}
2160
2161static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2162{
2163	struct vxlan_dev *vxlan = netdev_priv(dev);
2164	const struct in6_addr *daddr;
2165	const struct ipv6hdr *iphdr;
2166	struct inet6_dev *in6_dev;
2167	struct neighbour *n;
2168	struct nd_msg *msg;
2169
2170	in6_dev = __in6_dev_get(dev);
2171	if (!in6_dev)
2172		goto out;
2173
2174	iphdr = ipv6_hdr(skb);
2175	daddr = &iphdr->daddr;
2176	msg = (struct nd_msg *)(iphdr + 1);
2177
2178	if (ipv6_addr_loopback(daddr) ||
2179	    ipv6_addr_is_multicast(&msg->target))
2180		goto out;
2181
2182	n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2183
2184	if (n) {
2185		struct vxlan_fdb *f;
2186		struct sk_buff *reply;
2187
2188		if (!(n->nud_state & NUD_CONNECTED)) {
2189			neigh_release(n);
2190			goto out;
2191		}
2192
2193		f = vxlan_find_mac(vxlan, n->ha, vni);
2194		if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2195			/* bridge-local neighbor */
2196			neigh_release(n);
2197			goto out;
2198		}
2199
2200		reply = vxlan_na_create(skb, n,
2201					!!(f ? f->flags & NTF_ROUTER : 0));
2202
2203		neigh_release(n);
2204
2205		if (reply == NULL)
2206			goto out;
2207
2208		if (netif_rx_ni(reply) == NET_RX_DROP)
2209			dev->stats.rx_dropped++;
2210
2211	} else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2212		union vxlan_addr ipa = {
2213			.sin6.sin6_addr = msg->target,
2214			.sin6.sin6_family = AF_INET6,
2215		};
2216
2217		vxlan_ip_miss(dev, &ipa);
2218	}
2219
2220out:
2221	consume_skb(skb);
2222	return NETDEV_TX_OK;
2223}
2224#endif
2225
2226static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2227{
2228	struct vxlan_dev *vxlan = netdev_priv(dev);
2229	struct neighbour *n;
2230
2231	if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2232		return false;
2233
2234	n = NULL;
2235	switch (ntohs(eth_hdr(skb)->h_proto)) {
2236	case ETH_P_IP:
2237	{
2238		struct iphdr *pip;
2239
2240		if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2241			return false;
2242		pip = ip_hdr(skb);
2243		n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2244		if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2245			union vxlan_addr ipa = {
2246				.sin.sin_addr.s_addr = pip->daddr,
2247				.sin.sin_family = AF_INET,
2248			};
2249
2250			vxlan_ip_miss(dev, &ipa);
2251			return false;
2252		}
2253
2254		break;
2255	}
2256#if IS_ENABLED(CONFIG_IPV6)
2257	case ETH_P_IPV6:
2258	{
2259		struct ipv6hdr *pip6;
2260
2261		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2262			return false;
2263		pip6 = ipv6_hdr(skb);
2264		n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2265		if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2266			union vxlan_addr ipa = {
2267				.sin6.sin6_addr = pip6->daddr,
2268				.sin6.sin6_family = AF_INET6,
2269			};
2270
2271			vxlan_ip_miss(dev, &ipa);
2272			return false;
2273		}
2274
2275		break;
2276	}
2277#endif
2278	default:
2279		return false;
2280	}
2281
2282	if (n) {
2283		bool diff;
2284
2285		diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2286		if (diff) {
2287			memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2288				dev->addr_len);
2289			memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2290		}
2291		neigh_release(n);
2292		return diff;
2293	}
2294
2295	return false;
2296}
2297
2298static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
2299				struct vxlan_metadata *md)
2300{
2301	struct vxlanhdr_gbp *gbp;
2302
2303	if (!md->gbp)
2304		return;
2305
2306	gbp = (struct vxlanhdr_gbp *)vxh;
2307	vxh->vx_flags |= VXLAN_HF_GBP;
2308
2309	if (md->gbp & VXLAN_GBP_DONT_LEARN)
2310		gbp->dont_learn = 1;
2311
2312	if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
2313		gbp->policy_applied = 1;
2314
2315	gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
2316}
2317
2318static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
2319			       __be16 protocol)
2320{
2321	struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2322
2323	gpe->np_applied = 1;
2324	gpe->next_protocol = tun_p_from_eth_p(protocol);
2325	if (!gpe->next_protocol)
2326		return -EPFNOSUPPORT;
2327	return 0;
2328}
2329
2330static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2331			   int iphdr_len, __be32 vni,
2332			   struct vxlan_metadata *md, u32 vxflags,
2333			   bool udp_sum)
2334{
2335	struct vxlanhdr *vxh;
2336	int min_headroom;
2337	int err;
2338	int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2339	__be16 inner_protocol = htons(ETH_P_TEB);
2340
2341	if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2342	    skb->ip_summed == CHECKSUM_PARTIAL) {
2343		int csum_start = skb_checksum_start_offset(skb);
2344
2345		if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2346		    !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2347		    (skb->csum_offset == offsetof(struct udphdr, check) ||
2348		     skb->csum_offset == offsetof(struct tcphdr, check)))
2349			type |= SKB_GSO_TUNNEL_REMCSUM;
2350	}
2351
2352	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2353			+ VXLAN_HLEN + iphdr_len;
2354
2355	/* Need space for new headers (invalidates iph ptr) */
2356	err = skb_cow_head(skb, min_headroom);
2357	if (unlikely(err))
2358		return err;
2359
2360	err = iptunnel_handle_offloads(skb, type);
2361	if (err)
2362		return err;
2363
2364	vxh = __skb_push(skb, sizeof(*vxh));
2365	vxh->vx_flags = VXLAN_HF_VNI;
2366	vxh->vx_vni = vxlan_vni_field(vni);
2367
2368	if (type & SKB_GSO_TUNNEL_REMCSUM) {
2369		unsigned int start;
2370
2371		start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2372		vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2373		vxh->vx_flags |= VXLAN_HF_RCO;
2374
2375		if (!skb_is_gso(skb)) {
2376			skb->ip_summed = CHECKSUM_NONE;
2377			skb->encapsulation = 0;
2378		}
2379	}
2380
2381	if (vxflags & VXLAN_F_GBP)
2382		vxlan_build_gbp_hdr(vxh, vxflags, md);
2383	if (vxflags & VXLAN_F_GPE) {
2384		err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2385		if (err < 0)
2386			return err;
2387		inner_protocol = skb->protocol;
2388	}
2389
2390	skb_set_inner_protocol(skb, inner_protocol);
2391	return 0;
2392}
2393
2394static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2395				      struct vxlan_sock *sock4,
2396				      struct sk_buff *skb, int oif, u8 tos,
2397				      __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
2398				      struct dst_cache *dst_cache,
2399				      const struct ip_tunnel_info *info)
2400{
2401	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2402	struct rtable *rt = NULL;
2403	struct flowi4 fl4;
2404
2405	if (!sock4)
2406		return ERR_PTR(-EIO);
2407
2408	if (tos && !info)
2409		use_cache = false;
2410	if (use_cache) {
2411		rt = dst_cache_get_ip4(dst_cache, saddr);
2412		if (rt)
2413			return rt;
2414	}
2415
2416	memset(&fl4, 0, sizeof(fl4));
2417	fl4.flowi4_oif = oif;
2418	fl4.flowi4_tos = RT_TOS(tos);
2419	fl4.flowi4_mark = skb->mark;
2420	fl4.flowi4_proto = IPPROTO_UDP;
2421	fl4.daddr = daddr;
2422	fl4.saddr = *saddr;
2423	fl4.fl4_dport = dport;
2424	fl4.fl4_sport = sport;
2425
2426	rt = ip_route_output_key(vxlan->net, &fl4);
2427	if (!IS_ERR(rt)) {
2428		if (rt->dst.dev == dev) {
2429			netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2430			ip_rt_put(rt);
2431			return ERR_PTR(-ELOOP);
2432		}
2433
2434		*saddr = fl4.saddr;
2435		if (use_cache)
2436			dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2437	} else {
2438		netdev_dbg(dev, "no route to %pI4\n", &daddr);
2439		return ERR_PTR(-ENETUNREACH);
2440	}
2441	return rt;
2442}
2443
2444#if IS_ENABLED(CONFIG_IPV6)
2445static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2446					  struct net_device *dev,
2447					  struct vxlan_sock *sock6,
2448					  struct sk_buff *skb, int oif, u8 tos,
2449					  __be32 label,
2450					  const struct in6_addr *daddr,
2451					  struct in6_addr *saddr,
2452					  __be16 dport, __be16 sport,
2453					  struct dst_cache *dst_cache,
2454					  const struct ip_tunnel_info *info)
2455{
2456	bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2457	struct dst_entry *ndst;
2458	struct flowi6 fl6;
 
2459
2460	if (!sock6)
2461		return ERR_PTR(-EIO);
2462
2463	if (tos && !info)
2464		use_cache = false;
2465	if (use_cache) {
2466		ndst = dst_cache_get_ip6(dst_cache, saddr);
2467		if (ndst)
2468			return ndst;
2469	}
2470
2471	memset(&fl6, 0, sizeof(fl6));
2472	fl6.flowi6_oif = oif;
2473	fl6.daddr = *daddr;
2474	fl6.saddr = *saddr;
2475	fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
2476	fl6.flowi6_mark = skb->mark;
2477	fl6.flowi6_proto = IPPROTO_UDP;
2478	fl6.fl6_dport = dport;
2479	fl6.fl6_sport = sport;
2480
2481	ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
2482					       &fl6, NULL);
2483	if (unlikely(IS_ERR(ndst))) {
 
2484		netdev_dbg(dev, "no route to %pI6\n", daddr);
2485		return ERR_PTR(-ENETUNREACH);
2486	}
2487
2488	if (unlikely(ndst->dev == dev)) {
2489		netdev_dbg(dev, "circular route to %pI6\n", daddr);
2490		dst_release(ndst);
2491		return ERR_PTR(-ELOOP);
2492	}
2493
2494	*saddr = fl6.saddr;
2495	if (use_cache)
2496		dst_cache_set_ip6(dst_cache, ndst, saddr);
2497	return ndst;
2498}
2499#endif
2500
2501/* Bypass encapsulation if the destination is local */
2502static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2503			       struct vxlan_dev *dst_vxlan, __be32 vni,
2504			       bool snoop)
2505{
2506	struct pcpu_sw_netstats *tx_stats, *rx_stats;
2507	union vxlan_addr loopback;
2508	union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2509	struct net_device *dev;
2510	int len = skb->len;
2511
2512	tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2513	rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2514	skb->pkt_type = PACKET_HOST;
2515	skb->encapsulation = 0;
2516	skb->dev = dst_vxlan->dev;
2517	__skb_pull(skb, skb_network_offset(skb));
2518
2519	if (remote_ip->sa.sa_family == AF_INET) {
2520		loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2521		loopback.sa.sa_family =  AF_INET;
2522#if IS_ENABLED(CONFIG_IPV6)
2523	} else {
2524		loopback.sin6.sin6_addr = in6addr_loopback;
2525		loopback.sa.sa_family =  AF_INET6;
2526#endif
2527	}
2528
2529	rcu_read_lock();
2530	dev = skb->dev;
2531	if (unlikely(!(dev->flags & IFF_UP))) {
2532		kfree_skb(skb);
2533		goto drop;
2534	}
2535
2536	if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
2537		vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2538
2539	u64_stats_update_begin(&tx_stats->syncp);
2540	tx_stats->tx_packets++;
2541	tx_stats->tx_bytes += len;
2542	u64_stats_update_end(&tx_stats->syncp);
2543
2544	if (netif_rx(skb) == NET_RX_SUCCESS) {
2545		u64_stats_update_begin(&rx_stats->syncp);
2546		rx_stats->rx_packets++;
2547		rx_stats->rx_bytes += len;
2548		u64_stats_update_end(&rx_stats->syncp);
2549	} else {
2550drop:
2551		dev->stats.rx_dropped++;
2552	}
2553	rcu_read_unlock();
2554}
2555
2556static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2557				 struct vxlan_dev *vxlan,
2558				 union vxlan_addr *daddr,
2559				 __be16 dst_port, int dst_ifindex, __be32 vni,
2560				 struct dst_entry *dst,
2561				 u32 rt_flags)
2562{
2563#if IS_ENABLED(CONFIG_IPV6)
2564	/* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2565	 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2566	 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2567	 */
2568	BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2569#endif
2570	/* Bypass encapsulation if the destination is local */
2571	if (rt_flags & RTCF_LOCAL &&
2572	    !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2573		struct vxlan_dev *dst_vxlan;
2574
2575		dst_release(dst);
2576		dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2577					   daddr->sa.sa_family, dst_port,
2578					   vxlan->cfg.flags);
2579		if (!dst_vxlan) {
2580			dev->stats.tx_errors++;
2581			kfree_skb(skb);
2582
2583			return -ENOENT;
2584		}
2585		vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
2586		return 1;
2587	}
2588
2589	return 0;
2590}
2591
2592static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2593			   __be32 default_vni, struct vxlan_rdst *rdst,
2594			   bool did_rsc)
2595{
2596	struct dst_cache *dst_cache;
2597	struct ip_tunnel_info *info;
2598	struct vxlan_dev *vxlan = netdev_priv(dev);
2599	const struct iphdr *old_iph = ip_hdr(skb);
2600	union vxlan_addr *dst;
2601	union vxlan_addr remote_ip, local_ip;
2602	struct vxlan_metadata _md;
2603	struct vxlan_metadata *md = &_md;
2604	__be16 src_port = 0, dst_port;
2605	struct dst_entry *ndst = NULL;
2606	__be32 vni, label;
2607	__u8 tos, ttl;
2608	int ifindex;
2609	int err;
2610	u32 flags = vxlan->cfg.flags;
2611	bool udp_sum = false;
2612	bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2613
2614	info = skb_tunnel_info(skb);
2615
2616	if (rdst) {
2617		dst = &rdst->remote_ip;
2618		if (vxlan_addr_any(dst)) {
2619			if (did_rsc) {
2620				/* short-circuited back to local bridge */
2621				vxlan_encap_bypass(skb, vxlan, vxlan,
2622						   default_vni, true);
2623				return;
2624			}
2625			goto drop;
2626		}
2627
2628		dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2629		vni = (rdst->remote_vni) ? : default_vni;
2630		ifindex = rdst->remote_ifindex;
2631		local_ip = vxlan->cfg.saddr;
2632		dst_cache = &rdst->dst_cache;
2633		md->gbp = skb->mark;
2634		if (flags & VXLAN_F_TTL_INHERIT) {
2635			ttl = ip_tunnel_get_ttl(old_iph, skb);
2636		} else {
2637			ttl = vxlan->cfg.ttl;
2638			if (!ttl && vxlan_addr_multicast(dst))
2639				ttl = 1;
2640		}
2641
2642		tos = vxlan->cfg.tos;
2643		if (tos == 1)
2644			tos = ip_tunnel_get_dsfield(old_iph, skb);
2645
2646		if (dst->sa.sa_family == AF_INET)
2647			udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2648		else
2649			udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2650		label = vxlan->cfg.label;
2651	} else {
2652		if (!info) {
2653			WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2654				  dev->name);
2655			goto drop;
2656		}
2657		remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2658		if (remote_ip.sa.sa_family == AF_INET) {
2659			remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2660			local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2661		} else {
2662			remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2663			local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2664		}
2665		dst = &remote_ip;
2666		dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2667		vni = tunnel_id_to_key32(info->key.tun_id);
2668		ifindex = 0;
2669		dst_cache = &info->dst_cache;
2670		if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2671			if (info->options_len < sizeof(*md))
2672				goto drop;
2673			md = ip_tunnel_info_opts(info);
2674		}
2675		ttl = info->key.ttl;
2676		tos = info->key.tos;
2677		label = info->key.label;
2678		udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2679	}
2680	src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2681				     vxlan->cfg.port_max, true);
2682
2683	rcu_read_lock();
2684	if (dst->sa.sa_family == AF_INET) {
2685		struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2686		struct rtable *rt;
2687		__be16 df = 0;
2688
2689		if (!ifindex)
2690			ifindex = sock4->sock->sk->sk_bound_dev_if;
2691
2692		rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2693				     dst->sin.sin_addr.s_addr,
2694				     &local_ip.sin.sin_addr.s_addr,
2695				     dst_port, src_port,
2696				     dst_cache, info);
2697		if (IS_ERR(rt)) {
2698			err = PTR_ERR(rt);
2699			goto tx_error;
2700		}
2701
 
2702		if (!info) {
2703			/* Bypass encapsulation if the destination is local */
2704			err = encap_bypass_if_local(skb, dev, vxlan, dst,
2705						    dst_port, ifindex, vni,
2706						    &rt->dst, rt->rt_flags);
2707			if (err)
2708				goto out_unlock;
2709
2710			if (vxlan->cfg.df == VXLAN_DF_SET) {
2711				df = htons(IP_DF);
2712			} else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2713				struct ethhdr *eth = eth_hdr(skb);
2714
2715				if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2716				    (ntohs(eth->h_proto) == ETH_P_IP &&
2717				     old_iph->frag_off & htons(IP_DF)))
2718					df = htons(IP_DF);
2719			}
2720		} else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2721			df = htons(IP_DF);
2722		}
2723
2724		ndst = &rt->dst;
2725		err = skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM,
2726					    netif_is_any_bridge_port(dev));
2727		if (err < 0) {
2728			goto tx_error;
2729		} else if (err) {
2730			if (info) {
2731				struct in_addr src, dst;
2732
2733				src = remote_ip.sin.sin_addr;
2734				dst = local_ip.sin.sin_addr;
2735				info->key.u.ipv4.src = src.s_addr;
2736				info->key.u.ipv4.dst = dst.s_addr;
2737			}
2738			vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2739			dst_release(ndst);
2740			goto out_unlock;
2741		}
2742
2743		tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2744		ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2745		err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2746				      vni, md, flags, udp_sum);
2747		if (err < 0)
2748			goto tx_error;
2749
2750		udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2751				    dst->sin.sin_addr.s_addr, tos, ttl, df,
2752				    src_port, dst_port, xnet, !udp_sum);
2753#if IS_ENABLED(CONFIG_IPV6)
2754	} else {
2755		struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2756
2757		if (!ifindex)
2758			ifindex = sock6->sock->sk->sk_bound_dev_if;
2759
2760		ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2761					label, &dst->sin6.sin6_addr,
2762					&local_ip.sin6.sin6_addr,
2763					dst_port, src_port,
2764					dst_cache, info);
2765		if (IS_ERR(ndst)) {
2766			err = PTR_ERR(ndst);
2767			ndst = NULL;
2768			goto tx_error;
2769		}
2770
2771		if (!info) {
2772			u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2773
2774			err = encap_bypass_if_local(skb, dev, vxlan, dst,
2775						    dst_port, ifindex, vni,
2776						    ndst, rt6i_flags);
2777			if (err)
2778				goto out_unlock;
2779		}
2780
2781		err = skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM,
2782					    netif_is_any_bridge_port(dev));
2783		if (err < 0) {
2784			goto tx_error;
2785		} else if (err) {
2786			if (info) {
2787				struct in6_addr src, dst;
2788
2789				src = remote_ip.sin6.sin6_addr;
2790				dst = local_ip.sin6.sin6_addr;
2791				info->key.u.ipv6.src = src;
2792				info->key.u.ipv6.dst = dst;
2793			}
2794
2795			vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2796			dst_release(ndst);
2797			goto out_unlock;
2798		}
2799
2800		tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2801		ttl = ttl ? : ip6_dst_hoplimit(ndst);
2802		skb_scrub_packet(skb, xnet);
2803		err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2804				      vni, md, flags, udp_sum);
2805		if (err < 0)
2806			goto tx_error;
2807
2808		udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2809				     &local_ip.sin6.sin6_addr,
2810				     &dst->sin6.sin6_addr, tos, ttl,
2811				     label, src_port, dst_port, !udp_sum);
2812#endif
2813	}
2814out_unlock:
2815	rcu_read_unlock();
2816	return;
2817
2818drop:
2819	dev->stats.tx_dropped++;
2820	dev_kfree_skb(skb);
2821	return;
2822
2823tx_error:
2824	rcu_read_unlock();
2825	if (err == -ELOOP)
2826		dev->stats.collisions++;
2827	else if (err == -ENETUNREACH)
2828		dev->stats.tx_carrier_errors++;
2829	dst_release(ndst);
2830	dev->stats.tx_errors++;
2831	kfree_skb(skb);
2832}
2833
2834static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2835			  struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2836{
2837	struct vxlan_rdst nh_rdst;
2838	struct nexthop *nh;
2839	bool do_xmit;
2840	u32 hash;
2841
2842	memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2843	hash = skb_get_hash(skb);
2844
2845	rcu_read_lock();
2846	nh = rcu_dereference(f->nh);
2847	if (!nh) {
2848		rcu_read_unlock();
2849		goto drop;
2850	}
2851	do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2852	rcu_read_unlock();
2853
2854	if (likely(do_xmit))
2855		vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2856	else
2857		goto drop;
2858
2859	return;
2860
2861drop:
2862	dev->stats.tx_dropped++;
2863	dev_kfree_skb(skb);
2864}
2865
2866/* Transmit local packets over Vxlan
2867 *
2868 * Outer IP header inherits ECN and DF from inner header.
2869 * Outer UDP destination is the VXLAN assigned port.
2870 *           source port is based on hash of flow
2871 */
2872static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2873{
2874	struct vxlan_dev *vxlan = netdev_priv(dev);
2875	struct vxlan_rdst *rdst, *fdst = NULL;
2876	const struct ip_tunnel_info *info;
2877	bool did_rsc = false;
2878	struct vxlan_fdb *f;
2879	struct ethhdr *eth;
2880	__be32 vni = 0;
2881
2882	info = skb_tunnel_info(skb);
2883
2884	skb_reset_mac_header(skb);
2885
2886	if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2887		if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2888		    info->mode & IP_TUNNEL_INFO_TX) {
2889			vni = tunnel_id_to_key32(info->key.tun_id);
2890		} else {
2891			if (info && info->mode & IP_TUNNEL_INFO_TX)
2892				vxlan_xmit_one(skb, dev, vni, NULL, false);
2893			else
2894				kfree_skb(skb);
2895			return NETDEV_TX_OK;
2896		}
2897	}
2898
2899	if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2900		eth = eth_hdr(skb);
2901		if (ntohs(eth->h_proto) == ETH_P_ARP)
2902			return arp_reduce(dev, skb, vni);
2903#if IS_ENABLED(CONFIG_IPV6)
2904		else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2905			 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2906					    sizeof(struct nd_msg)) &&
2907			 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2908			struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2909
2910			if (m->icmph.icmp6_code == 0 &&
2911			    m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2912				return neigh_reduce(dev, skb, vni);
2913		}
2914#endif
2915	}
2916
2917	eth = eth_hdr(skb);
2918	f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2919	did_rsc = false;
2920
2921	if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2922	    (ntohs(eth->h_proto) == ETH_P_IP ||
2923	     ntohs(eth->h_proto) == ETH_P_IPV6)) {
2924		did_rsc = route_shortcircuit(dev, skb);
2925		if (did_rsc)
2926			f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2927	}
2928
2929	if (f == NULL) {
2930		f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2931		if (f == NULL) {
2932			if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2933			    !is_multicast_ether_addr(eth->h_dest))
2934				vxlan_fdb_miss(vxlan, eth->h_dest);
2935
2936			dev->stats.tx_dropped++;
2937			kfree_skb(skb);
2938			return NETDEV_TX_OK;
2939		}
2940	}
2941
2942	if (rcu_access_pointer(f->nh)) {
2943		vxlan_xmit_nh(skb, dev, f,
2944			      (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2945	} else {
2946		list_for_each_entry_rcu(rdst, &f->remotes, list) {
2947			struct sk_buff *skb1;
2948
2949			if (!fdst) {
2950				fdst = rdst;
2951				continue;
2952			}
2953			skb1 = skb_clone(skb, GFP_ATOMIC);
2954			if (skb1)
2955				vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2956		}
2957		if (fdst)
2958			vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2959		else
2960			kfree_skb(skb);
2961	}
2962
 
 
 
 
2963	return NETDEV_TX_OK;
2964}
2965
2966/* Walk the forwarding table and purge stale entries */
2967static void vxlan_cleanup(struct timer_list *t)
2968{
2969	struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2970	unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2971	unsigned int h;
2972
2973	if (!netif_running(vxlan->dev))
2974		return;
2975
2976	for (h = 0; h < FDB_HASH_SIZE; ++h) {
2977		struct hlist_node *p, *n;
2978
2979		spin_lock(&vxlan->hash_lock[h]);
2980		hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2981			struct vxlan_fdb *f
2982				= container_of(p, struct vxlan_fdb, hlist);
2983			unsigned long timeout;
2984
2985			if (f->state & (NUD_PERMANENT | NUD_NOARP))
2986				continue;
2987
2988			if (f->flags & NTF_EXT_LEARNED)
2989				continue;
2990
2991			timeout = f->used + vxlan->cfg.age_interval * HZ;
2992			if (time_before_eq(timeout, jiffies)) {
2993				netdev_dbg(vxlan->dev,
2994					   "garbage collect %pM\n",
2995					   f->eth_addr);
2996				f->state = NUD_STALE;
2997				vxlan_fdb_destroy(vxlan, f, true, true);
2998			} else if (time_before(timeout, next_timer))
2999				next_timer = timeout;
3000		}
3001		spin_unlock(&vxlan->hash_lock[h]);
3002	}
3003
3004	mod_timer(&vxlan->age_timer, next_timer);
3005}
3006
3007static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
3008{
3009	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3010
3011	spin_lock(&vn->sock_lock);
3012	hlist_del_init_rcu(&vxlan->hlist4.hlist);
3013#if IS_ENABLED(CONFIG_IPV6)
3014	hlist_del_init_rcu(&vxlan->hlist6.hlist);
3015#endif
3016	spin_unlock(&vn->sock_lock);
3017}
3018
3019static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
3020			     struct vxlan_dev_node *node)
3021{
3022	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3023	__be32 vni = vxlan->default_dst.remote_vni;
3024
3025	node->vxlan = vxlan;
3026	spin_lock(&vn->sock_lock);
3027	hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
3028	spin_unlock(&vn->sock_lock);
3029}
3030
3031/* Setup stats when device is created */
3032static int vxlan_init(struct net_device *dev)
3033{
3034	struct vxlan_dev *vxlan = netdev_priv(dev);
3035	int err;
3036
3037	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
3038	if (!dev->tstats)
3039		return -ENOMEM;
3040
3041	err = gro_cells_init(&vxlan->gro_cells, dev);
3042	if (err) {
3043		free_percpu(dev->tstats);
3044		return err;
3045	}
3046
3047	return 0;
3048}
3049
3050static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
3051{
3052	struct vxlan_fdb *f;
3053	u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
3054
3055	spin_lock_bh(&vxlan->hash_lock[hash_index]);
3056	f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
3057	if (f)
3058		vxlan_fdb_destroy(vxlan, f, true, true);
3059	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
3060}
3061
3062static void vxlan_uninit(struct net_device *dev)
3063{
3064	struct vxlan_dev *vxlan = netdev_priv(dev);
3065
3066	gro_cells_destroy(&vxlan->gro_cells);
3067
3068	vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
3069
3070	free_percpu(dev->tstats);
3071}
3072
3073/* Start ageing timer and join group when device is brought up */
3074static int vxlan_open(struct net_device *dev)
3075{
3076	struct vxlan_dev *vxlan = netdev_priv(dev);
3077	int ret;
3078
3079	ret = vxlan_sock_add(vxlan);
3080	if (ret < 0)
3081		return ret;
3082
3083	if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
3084		ret = vxlan_igmp_join(vxlan);
3085		if (ret == -EADDRINUSE)
3086			ret = 0;
3087		if (ret) {
3088			vxlan_sock_release(vxlan);
3089			return ret;
3090		}
3091	}
3092
3093	if (vxlan->cfg.age_interval)
3094		mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
3095
3096	return ret;
3097}
3098
3099/* Purge the forwarding table */
3100static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
3101{
3102	unsigned int h;
3103
 
3104	for (h = 0; h < FDB_HASH_SIZE; ++h) {
3105		struct hlist_node *p, *n;
3106
3107		spin_lock_bh(&vxlan->hash_lock[h]);
3108		hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3109			struct vxlan_fdb *f
3110				= container_of(p, struct vxlan_fdb, hlist);
3111			if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
3112				continue;
3113			/* the all_zeros_mac entry is deleted at vxlan_uninit */
3114			if (is_zero_ether_addr(f->eth_addr) &&
3115			    f->vni == vxlan->cfg.vni)
3116				continue;
3117			vxlan_fdb_destroy(vxlan, f, true, true);
3118		}
3119		spin_unlock_bh(&vxlan->hash_lock[h]);
3120	}
 
3121}
3122
3123/* Cleanup timer and forwarding table on shutdown */
3124static int vxlan_stop(struct net_device *dev)
3125{
3126	struct vxlan_dev *vxlan = netdev_priv(dev);
3127	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3128	int ret = 0;
3129
3130	if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
3131	    !vxlan_group_used(vn, vxlan))
3132		ret = vxlan_igmp_leave(vxlan);
3133
3134	del_timer_sync(&vxlan->age_timer);
3135
3136	vxlan_flush(vxlan, false);
3137	vxlan_sock_release(vxlan);
3138
3139	return ret;
3140}
3141
3142/* Stub, nothing needs to be done. */
3143static void vxlan_set_multicast_list(struct net_device *dev)
3144{
3145}
3146
3147static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3148{
3149	struct vxlan_dev *vxlan = netdev_priv(dev);
3150	struct vxlan_rdst *dst = &vxlan->default_dst;
3151	struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3152							 dst->remote_ifindex);
3153	bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
3154
3155	/* This check is different than dev->max_mtu, because it looks at
3156	 * the lowerdev->mtu, rather than the static dev->max_mtu
3157	 */
3158	if (lowerdev) {
3159		int max_mtu = lowerdev->mtu -
3160			      (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
3161		if (new_mtu > max_mtu)
3162			return -EINVAL;
3163	}
3164
3165	dev->mtu = new_mtu;
3166	return 0;
3167}
3168
3169static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3170{
3171	struct vxlan_dev *vxlan = netdev_priv(dev);
3172	struct ip_tunnel_info *info = skb_tunnel_info(skb);
3173	__be16 sport, dport;
3174
3175	sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3176				  vxlan->cfg.port_max, true);
3177	dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3178
3179	if (ip_tunnel_info_af(info) == AF_INET) {
3180		struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3181		struct rtable *rt;
3182
3183		rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
3184				     info->key.u.ipv4.dst,
3185				     &info->key.u.ipv4.src, dport, sport,
3186				     &info->dst_cache, info);
3187		if (IS_ERR(rt))
3188			return PTR_ERR(rt);
3189		ip_rt_put(rt);
3190	} else {
3191#if IS_ENABLED(CONFIG_IPV6)
3192		struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3193		struct dst_entry *ndst;
3194
3195		ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
3196					info->key.label, &info->key.u.ipv6.dst,
3197					&info->key.u.ipv6.src, dport, sport,
3198					&info->dst_cache, info);
3199		if (IS_ERR(ndst))
3200			return PTR_ERR(ndst);
3201		dst_release(ndst);
3202#else /* !CONFIG_IPV6 */
3203		return -EPFNOSUPPORT;
3204#endif
3205	}
3206	info->key.tp_src = sport;
3207	info->key.tp_dst = dport;
3208	return 0;
3209}
3210
3211static const struct net_device_ops vxlan_netdev_ether_ops = {
3212	.ndo_init		= vxlan_init,
3213	.ndo_uninit		= vxlan_uninit,
3214	.ndo_open		= vxlan_open,
3215	.ndo_stop		= vxlan_stop,
3216	.ndo_start_xmit		= vxlan_xmit,
3217	.ndo_get_stats64	= ip_tunnel_get_stats64,
3218	.ndo_set_rx_mode	= vxlan_set_multicast_list,
3219	.ndo_change_mtu		= vxlan_change_mtu,
3220	.ndo_validate_addr	= eth_validate_addr,
3221	.ndo_set_mac_address	= eth_mac_addr,
3222	.ndo_fdb_add		= vxlan_fdb_add,
3223	.ndo_fdb_del		= vxlan_fdb_delete,
3224	.ndo_fdb_dump		= vxlan_fdb_dump,
3225	.ndo_fdb_get		= vxlan_fdb_get,
3226	.ndo_fill_metadata_dst	= vxlan_fill_metadata_dst,
3227	.ndo_change_proto_down  = dev_change_proto_down_generic,
3228};
3229
3230static const struct net_device_ops vxlan_netdev_raw_ops = {
3231	.ndo_init		= vxlan_init,
3232	.ndo_uninit		= vxlan_uninit,
3233	.ndo_open		= vxlan_open,
3234	.ndo_stop		= vxlan_stop,
3235	.ndo_start_xmit		= vxlan_xmit,
3236	.ndo_get_stats64	= ip_tunnel_get_stats64,
3237	.ndo_change_mtu		= vxlan_change_mtu,
3238	.ndo_fill_metadata_dst	= vxlan_fill_metadata_dst,
3239};
3240
3241/* Info for udev, that this is a virtual tunnel endpoint */
3242static struct device_type vxlan_type = {
3243	.name = "vxlan",
3244};
3245
3246/* Calls the ndo_udp_tunnel_add of the caller in order to
3247 * supply the listening VXLAN udp ports. Callers are expected
3248 * to implement the ndo_udp_tunnel_add.
3249 */
3250static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3251{
3252	struct vxlan_sock *vs;
3253	struct net *net = dev_net(dev);
3254	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3255	unsigned int i;
3256
3257	spin_lock(&vn->sock_lock);
3258	for (i = 0; i < PORT_HASH_SIZE; ++i) {
3259		hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3260			unsigned short type;
3261
3262			if (vs->flags & VXLAN_F_GPE)
3263				type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3264			else
3265				type = UDP_TUNNEL_TYPE_VXLAN;
3266
3267			if (push)
3268				udp_tunnel_push_rx_port(dev, vs->sock, type);
3269			else
3270				udp_tunnel_drop_rx_port(dev, vs->sock, type);
3271		}
3272	}
3273	spin_unlock(&vn->sock_lock);
3274}
3275
3276/* Initialize the device structure. */
3277static void vxlan_setup(struct net_device *dev)
3278{
3279	struct vxlan_dev *vxlan = netdev_priv(dev);
3280	unsigned int h;
3281
3282	eth_hw_addr_random(dev);
3283	ether_setup(dev);
3284
3285	dev->needs_free_netdev = true;
3286	SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3287
3288	dev->features	|= NETIF_F_LLTX;
3289	dev->features	|= NETIF_F_SG | NETIF_F_HW_CSUM;
3290	dev->features   |= NETIF_F_RXCSUM;
3291	dev->features   |= NETIF_F_GSO_SOFTWARE;
3292
3293	dev->vlan_features = dev->features;
3294	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
3295	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3296	netif_keep_dst(dev);
3297	dev->priv_flags |= IFF_NO_QUEUE;
3298
3299	/* MTU range: 68 - 65535 */
3300	dev->min_mtu = ETH_MIN_MTU;
3301	dev->max_mtu = ETH_MAX_MTU;
3302
3303	INIT_LIST_HEAD(&vxlan->next);
 
3304
3305	timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3306
3307	vxlan->dev = dev;
3308
3309	for (h = 0; h < FDB_HASH_SIZE; ++h) {
3310		spin_lock_init(&vxlan->hash_lock[h]);
 
3311		INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3312	}
3313}
3314
3315static void vxlan_ether_setup(struct net_device *dev)
3316{
3317	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3318	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3319	dev->netdev_ops = &vxlan_netdev_ether_ops;
3320}
3321
3322static void vxlan_raw_setup(struct net_device *dev)
3323{
3324	dev->header_ops = NULL;
3325	dev->type = ARPHRD_NONE;
3326	dev->hard_header_len = 0;
3327	dev->addr_len = 0;
3328	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3329	dev->netdev_ops = &vxlan_netdev_raw_ops;
3330}
3331
3332static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3333	[IFLA_VXLAN_ID]		= { .type = NLA_U32 },
3334	[IFLA_VXLAN_GROUP]	= { .len = sizeof_field(struct iphdr, daddr) },
3335	[IFLA_VXLAN_GROUP6]	= { .len = sizeof(struct in6_addr) },
3336	[IFLA_VXLAN_LINK]	= { .type = NLA_U32 },
3337	[IFLA_VXLAN_LOCAL]	= { .len = sizeof_field(struct iphdr, saddr) },
3338	[IFLA_VXLAN_LOCAL6]	= { .len = sizeof(struct in6_addr) },
3339	[IFLA_VXLAN_TOS]	= { .type = NLA_U8 },
3340	[IFLA_VXLAN_TTL]	= { .type = NLA_U8 },
3341	[IFLA_VXLAN_LABEL]	= { .type = NLA_U32 },
3342	[IFLA_VXLAN_LEARNING]	= { .type = NLA_U8 },
3343	[IFLA_VXLAN_AGEING]	= { .type = NLA_U32 },
3344	[IFLA_VXLAN_LIMIT]	= { .type = NLA_U32 },
3345	[IFLA_VXLAN_PORT_RANGE] = { .len  = sizeof(struct ifla_vxlan_port_range) },
3346	[IFLA_VXLAN_PROXY]	= { .type = NLA_U8 },
3347	[IFLA_VXLAN_RSC]	= { .type = NLA_U8 },
3348	[IFLA_VXLAN_L2MISS]	= { .type = NLA_U8 },
3349	[IFLA_VXLAN_L3MISS]	= { .type = NLA_U8 },
3350	[IFLA_VXLAN_COLLECT_METADATA]	= { .type = NLA_U8 },
3351	[IFLA_VXLAN_PORT]	= { .type = NLA_U16 },
3352	[IFLA_VXLAN_UDP_CSUM]	= { .type = NLA_U8 },
3353	[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]	= { .type = NLA_U8 },
3354	[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]	= { .type = NLA_U8 },
3355	[IFLA_VXLAN_REMCSUM_TX]	= { .type = NLA_U8 },
3356	[IFLA_VXLAN_REMCSUM_RX]	= { .type = NLA_U8 },
3357	[IFLA_VXLAN_GBP]	= { .type = NLA_FLAG, },
3358	[IFLA_VXLAN_GPE]	= { .type = NLA_FLAG, },
3359	[IFLA_VXLAN_REMCSUM_NOPARTIAL]	= { .type = NLA_FLAG },
3360	[IFLA_VXLAN_TTL_INHERIT]	= { .type = NLA_FLAG },
3361	[IFLA_VXLAN_DF]		= { .type = NLA_U8 },
3362};
3363
3364static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3365			  struct netlink_ext_ack *extack)
3366{
3367	if (tb[IFLA_ADDRESS]) {
3368		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3369			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3370					    "Provided link layer address is not Ethernet");
3371			return -EINVAL;
3372		}
3373
3374		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3375			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3376					    "Provided Ethernet address is not unicast");
3377			return -EADDRNOTAVAIL;
3378		}
3379	}
3380
3381	if (tb[IFLA_MTU]) {
3382		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3383
3384		if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3385			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3386					    "MTU must be between 68 and 65535");
3387			return -EINVAL;
3388		}
3389	}
3390
3391	if (!data) {
3392		NL_SET_ERR_MSG(extack,
3393			       "Required attributes not provided to perform the operation");
3394		return -EINVAL;
3395	}
3396
3397	if (data[IFLA_VXLAN_ID]) {
3398		u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3399
3400		if (id >= VXLAN_N_VID) {
3401			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3402					    "VXLAN ID must be lower than 16777216");
3403			return -ERANGE;
3404		}
3405	}
3406
3407	if (data[IFLA_VXLAN_PORT_RANGE]) {
3408		const struct ifla_vxlan_port_range *p
3409			= nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3410
3411		if (ntohs(p->high) < ntohs(p->low)) {
3412			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3413					    "Invalid source port range");
3414			return -EINVAL;
3415		}
3416	}
3417
3418	if (data[IFLA_VXLAN_DF]) {
3419		enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3420
3421		if (df < 0 || df > VXLAN_DF_MAX) {
3422			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3423					    "Invalid DF attribute");
3424			return -EINVAL;
3425		}
3426	}
3427
3428	return 0;
3429}
3430
3431static void vxlan_get_drvinfo(struct net_device *netdev,
3432			      struct ethtool_drvinfo *drvinfo)
3433{
3434	strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3435	strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3436}
3437
3438static int vxlan_get_link_ksettings(struct net_device *dev,
3439				    struct ethtool_link_ksettings *cmd)
3440{
3441	struct vxlan_dev *vxlan = netdev_priv(dev);
3442	struct vxlan_rdst *dst = &vxlan->default_dst;
3443	struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3444							 dst->remote_ifindex);
3445
3446	if (!lowerdev) {
3447		cmd->base.duplex = DUPLEX_UNKNOWN;
3448		cmd->base.port = PORT_OTHER;
3449		cmd->base.speed = SPEED_UNKNOWN;
3450
3451		return 0;
3452	}
3453
3454	return __ethtool_get_link_ksettings(lowerdev, cmd);
3455}
3456
3457static const struct ethtool_ops vxlan_ethtool_ops = {
3458	.get_drvinfo		= vxlan_get_drvinfo,
3459	.get_link		= ethtool_op_get_link,
3460	.get_link_ksettings	= vxlan_get_link_ksettings,
3461};
3462
3463static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3464					__be16 port, u32 flags, int ifindex)
3465{
3466	struct socket *sock;
3467	struct udp_port_cfg udp_conf;
3468	int err;
3469
3470	memset(&udp_conf, 0, sizeof(udp_conf));
3471
3472	if (ipv6) {
3473		udp_conf.family = AF_INET6;
3474		udp_conf.use_udp6_rx_checksums =
3475		    !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3476		udp_conf.ipv6_v6only = 1;
3477	} else {
3478		udp_conf.family = AF_INET;
3479	}
3480
3481	udp_conf.local_udp_port = port;
3482	udp_conf.bind_ifindex = ifindex;
3483
3484	/* Open UDP socket */
3485	err = udp_sock_create(net, &udp_conf, &sock);
3486	if (err < 0)
3487		return ERR_PTR(err);
3488
3489	return sock;
3490}
3491
3492/* Create new listen socket if needed */
3493static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3494					      __be16 port, u32 flags,
3495					      int ifindex)
3496{
3497	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3498	struct vxlan_sock *vs;
3499	struct socket *sock;
3500	unsigned int h;
3501	struct udp_tunnel_sock_cfg tunnel_cfg;
3502
3503	vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3504	if (!vs)
3505		return ERR_PTR(-ENOMEM);
3506
3507	for (h = 0; h < VNI_HASH_SIZE; ++h)
3508		INIT_HLIST_HEAD(&vs->vni_list[h]);
3509
3510	sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3511	if (IS_ERR(sock)) {
3512		kfree(vs);
3513		return ERR_CAST(sock);
3514	}
3515
3516	vs->sock = sock;
3517	refcount_set(&vs->refcnt, 1);
3518	vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3519
3520	spin_lock(&vn->sock_lock);
3521	hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3522	udp_tunnel_notify_add_rx_port(sock,
3523				      (vs->flags & VXLAN_F_GPE) ?
3524				      UDP_TUNNEL_TYPE_VXLAN_GPE :
3525				      UDP_TUNNEL_TYPE_VXLAN);
3526	spin_unlock(&vn->sock_lock);
3527
3528	/* Mark socket as an encapsulation socket. */
3529	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3530	tunnel_cfg.sk_user_data = vs;
3531	tunnel_cfg.encap_type = 1;
3532	tunnel_cfg.encap_rcv = vxlan_rcv;
3533	tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3534	tunnel_cfg.encap_destroy = NULL;
3535	tunnel_cfg.gro_receive = vxlan_gro_receive;
3536	tunnel_cfg.gro_complete = vxlan_gro_complete;
3537
3538	setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3539
3540	return vs;
3541}
3542
3543static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3544{
3545	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3546	struct vxlan_sock *vs = NULL;
3547	struct vxlan_dev_node *node;
3548	int l3mdev_index = 0;
3549
3550	if (vxlan->cfg.remote_ifindex)
3551		l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3552			vxlan->net, vxlan->cfg.remote_ifindex);
3553
3554	if (!vxlan->cfg.no_share) {
3555		spin_lock(&vn->sock_lock);
3556		vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3557				     vxlan->cfg.dst_port, vxlan->cfg.flags,
3558				     l3mdev_index);
3559		if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3560			spin_unlock(&vn->sock_lock);
3561			return -EBUSY;
3562		}
3563		spin_unlock(&vn->sock_lock);
3564	}
3565	if (!vs)
3566		vs = vxlan_socket_create(vxlan->net, ipv6,
3567					 vxlan->cfg.dst_port, vxlan->cfg.flags,
3568					 l3mdev_index);
3569	if (IS_ERR(vs))
3570		return PTR_ERR(vs);
3571#if IS_ENABLED(CONFIG_IPV6)
3572	if (ipv6) {
3573		rcu_assign_pointer(vxlan->vn6_sock, vs);
3574		node = &vxlan->hlist6;
3575	} else
3576#endif
3577	{
3578		rcu_assign_pointer(vxlan->vn4_sock, vs);
3579		node = &vxlan->hlist4;
3580	}
3581	vxlan_vs_add_dev(vs, vxlan, node);
3582	return 0;
3583}
3584
3585static int vxlan_sock_add(struct vxlan_dev *vxlan)
3586{
3587	bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3588	bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3589	bool ipv4 = !ipv6 || metadata;
3590	int ret = 0;
3591
3592	RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3593#if IS_ENABLED(CONFIG_IPV6)
3594	RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3595	if (ipv6) {
3596		ret = __vxlan_sock_add(vxlan, true);
3597		if (ret < 0 && ret != -EAFNOSUPPORT)
3598			ipv4 = false;
3599	}
3600#endif
3601	if (ipv4)
3602		ret = __vxlan_sock_add(vxlan, false);
3603	if (ret < 0)
3604		vxlan_sock_release(vxlan);
3605	return ret;
3606}
3607
3608static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3609				 struct net_device **lower,
3610				 struct vxlan_dev *old,
3611				 struct netlink_ext_ack *extack)
3612{
3613	struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3614	struct vxlan_dev *tmp;
3615	bool use_ipv6 = false;
3616
3617	if (conf->flags & VXLAN_F_GPE) {
3618		/* For now, allow GPE only together with
3619		 * COLLECT_METADATA. This can be relaxed later; in such
3620		 * case, the other side of the PtP link will have to be
3621		 * provided.
3622		 */
3623		if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3624		    !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3625			NL_SET_ERR_MSG(extack,
3626				       "VXLAN GPE does not support this combination of attributes");
3627			return -EINVAL;
3628		}
3629	}
3630
3631	if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3632		/* Unless IPv6 is explicitly requested, assume IPv4 */
3633		conf->remote_ip.sa.sa_family = AF_INET;
3634		conf->saddr.sa.sa_family = AF_INET;
3635	} else if (!conf->remote_ip.sa.sa_family) {
3636		conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3637	} else if (!conf->saddr.sa.sa_family) {
3638		conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3639	}
3640
3641	if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3642		NL_SET_ERR_MSG(extack,
3643			       "Local and remote address must be from the same family");
3644		return -EINVAL;
3645	}
3646
3647	if (vxlan_addr_multicast(&conf->saddr)) {
3648		NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3649		return -EINVAL;
3650	}
3651
3652	if (conf->saddr.sa.sa_family == AF_INET6) {
3653		if (!IS_ENABLED(CONFIG_IPV6)) {
3654			NL_SET_ERR_MSG(extack,
3655				       "IPv6 support not enabled in the kernel");
3656			return -EPFNOSUPPORT;
3657		}
3658		use_ipv6 = true;
3659		conf->flags |= VXLAN_F_IPV6;
3660
3661		if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3662			int local_type =
3663				ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3664			int remote_type =
3665				ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3666
3667			if (local_type & IPV6_ADDR_LINKLOCAL) {
3668				if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3669				    (remote_type != IPV6_ADDR_ANY)) {
3670					NL_SET_ERR_MSG(extack,
3671						       "Invalid combination of local and remote address scopes");
3672					return -EINVAL;
3673				}
3674
3675				conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3676			} else {
3677				if (remote_type ==
3678				    (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3679					NL_SET_ERR_MSG(extack,
3680						       "Invalid combination of local and remote address scopes");
3681					return -EINVAL;
3682				}
3683
3684				conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3685			}
3686		}
3687	}
3688
3689	if (conf->label && !use_ipv6) {
3690		NL_SET_ERR_MSG(extack,
3691			       "Label attribute only applies to IPv6 VXLAN devices");
3692		return -EINVAL;
3693	}
3694
3695	if (conf->remote_ifindex) {
3696		struct net_device *lowerdev;
3697
3698		lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3699		if (!lowerdev) {
3700			NL_SET_ERR_MSG(extack,
3701				       "Invalid local interface, device not found");
3702			return -ENODEV;
3703		}
3704
3705#if IS_ENABLED(CONFIG_IPV6)
3706		if (use_ipv6) {
3707			struct inet6_dev *idev = __in6_dev_get(lowerdev);
3708			if (idev && idev->cnf.disable_ipv6) {
3709				NL_SET_ERR_MSG(extack,
3710					       "IPv6 support disabled by administrator");
3711				return -EPERM;
3712			}
3713		}
3714#endif
3715
3716		*lower = lowerdev;
3717	} else {
3718		if (vxlan_addr_multicast(&conf->remote_ip)) {
3719			NL_SET_ERR_MSG(extack,
3720				       "Local interface required for multicast remote destination");
3721
3722			return -EINVAL;
3723		}
3724
3725#if IS_ENABLED(CONFIG_IPV6)
3726		if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3727			NL_SET_ERR_MSG(extack,
3728				       "Local interface required for link-local local/remote addresses");
3729			return -EINVAL;
3730		}
3731#endif
3732
3733		*lower = NULL;
3734	}
3735
3736	if (!conf->dst_port) {
3737		if (conf->flags & VXLAN_F_GPE)
3738			conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3739		else
3740			conf->dst_port = htons(vxlan_port);
3741	}
3742
3743	if (!conf->age_interval)
3744		conf->age_interval = FDB_AGE_DEFAULT;
3745
3746	list_for_each_entry(tmp, &vn->vxlan_list, next) {
3747		if (tmp == old)
3748			continue;
3749
3750		if (tmp->cfg.vni != conf->vni)
3751			continue;
3752		if (tmp->cfg.dst_port != conf->dst_port)
3753			continue;
3754		if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3755		    (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3756			continue;
3757
3758		if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3759		    tmp->cfg.remote_ifindex != conf->remote_ifindex)
3760			continue;
3761
3762		NL_SET_ERR_MSG(extack,
3763			       "A VXLAN device with the specified VNI already exists");
3764		return -EEXIST;
3765	}
3766
3767	return 0;
3768}
3769
3770static void vxlan_config_apply(struct net_device *dev,
3771			       struct vxlan_config *conf,
3772			       struct net_device *lowerdev,
3773			       struct net *src_net,
3774			       bool changelink)
3775{
3776	struct vxlan_dev *vxlan = netdev_priv(dev);
3777	struct vxlan_rdst *dst = &vxlan->default_dst;
3778	unsigned short needed_headroom = ETH_HLEN;
3779	bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3780	int max_mtu = ETH_MAX_MTU;
3781
3782	if (!changelink) {
3783		if (conf->flags & VXLAN_F_GPE)
3784			vxlan_raw_setup(dev);
3785		else
3786			vxlan_ether_setup(dev);
3787
3788		if (conf->mtu)
3789			dev->mtu = conf->mtu;
3790
3791		vxlan->net = src_net;
3792	}
3793
3794	dst->remote_vni = conf->vni;
3795
3796	memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3797
3798	if (lowerdev) {
3799		dst->remote_ifindex = conf->remote_ifindex;
3800
3801		dev->gso_max_size = lowerdev->gso_max_size;
3802		dev->gso_max_segs = lowerdev->gso_max_segs;
3803
3804		needed_headroom = lowerdev->hard_header_len;
3805
3806		max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3807					   VXLAN_HEADROOM);
3808		if (max_mtu < ETH_MIN_MTU)
3809			max_mtu = ETH_MIN_MTU;
3810
3811		if (!changelink && !conf->mtu)
3812			dev->mtu = max_mtu;
3813	}
3814
3815	if (dev->mtu > max_mtu)
3816		dev->mtu = max_mtu;
3817
3818	if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3819		needed_headroom += VXLAN6_HEADROOM;
3820	else
3821		needed_headroom += VXLAN_HEADROOM;
3822	dev->needed_headroom = needed_headroom;
3823
3824	memcpy(&vxlan->cfg, conf, sizeof(*conf));
3825}
3826
3827static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3828			       struct vxlan_config *conf, bool changelink,
3829			       struct netlink_ext_ack *extack)
3830{
3831	struct vxlan_dev *vxlan = netdev_priv(dev);
3832	struct net_device *lowerdev;
3833	int ret;
3834
3835	ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3836	if (ret)
3837		return ret;
3838
3839	vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3840
3841	return 0;
3842}
3843
3844static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3845			      struct vxlan_config *conf,
3846			      struct netlink_ext_ack *extack)
3847{
3848	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3849	struct vxlan_dev *vxlan = netdev_priv(dev);
3850	struct net_device *remote_dev = NULL;
3851	struct vxlan_fdb *f = NULL;
3852	bool unregister = false;
3853	struct vxlan_rdst *dst;
3854	int err;
3855
3856	dst = &vxlan->default_dst;
3857	err = vxlan_dev_configure(net, dev, conf, false, extack);
3858	if (err)
3859		return err;
3860
3861	dev->ethtool_ops = &vxlan_ethtool_ops;
3862
3863	/* create an fdb entry for a valid default destination */
3864	if (!vxlan_addr_any(&dst->remote_ip)) {
3865		err = vxlan_fdb_create(vxlan, all_zeros_mac,
3866				       &dst->remote_ip,
3867				       NUD_REACHABLE | NUD_PERMANENT,
 
3868				       vxlan->cfg.dst_port,
3869				       dst->remote_vni,
3870				       dst->remote_vni,
3871				       dst->remote_ifindex,
3872				       NTF_SELF, 0, &f, extack);
3873		if (err)
3874			return err;
3875	}
3876
3877	err = register_netdevice(dev);
3878	if (err)
3879		goto errout;
3880	unregister = true;
3881
3882	if (dst->remote_ifindex) {
3883		remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3884		if (!remote_dev)
3885			goto errout;
3886
3887		err = netdev_upper_dev_link(remote_dev, dev, extack);
3888		if (err)
3889			goto errout;
3890	}
3891
3892	err = rtnl_configure_link(dev, NULL);
3893	if (err)
3894		goto unlink;
3895
3896	if (f) {
3897		vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
3898
3899		/* notify default fdb entry */
3900		err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
3901				       RTM_NEWNEIGH, true, extack);
3902		if (err) {
3903			vxlan_fdb_destroy(vxlan, f, false, false);
3904			if (remote_dev)
3905				netdev_upper_dev_unlink(remote_dev, dev);
3906			goto unregister;
3907		}
3908	}
3909
3910	list_add(&vxlan->next, &vn->vxlan_list);
3911	if (remote_dev)
3912		dst->remote_dev = remote_dev;
3913	return 0;
3914unlink:
3915	if (remote_dev)
3916		netdev_upper_dev_unlink(remote_dev, dev);
3917errout:
3918	/* unregister_netdevice() destroys the default FDB entry with deletion
3919	 * notification. But the addition notification was not sent yet, so
3920	 * destroy the entry by hand here.
3921	 */
3922	if (f)
3923		__vxlan_fdb_free(f);
3924unregister:
3925	if (unregister)
3926		unregister_netdevice(dev);
3927	return err;
3928}
3929
3930/* Set/clear flags based on attribute */
3931static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3932			  int attrtype, unsigned long mask, bool changelink,
3933			  bool changelink_supported,
3934			  struct netlink_ext_ack *extack)
3935{
3936	unsigned long flags;
3937
3938	if (!tb[attrtype])
3939		return 0;
3940
3941	if (changelink && !changelink_supported) {
3942		vxlan_flag_attr_error(attrtype, extack);
3943		return -EOPNOTSUPP;
3944	}
3945
3946	if (vxlan_policy[attrtype].type == NLA_FLAG)
3947		flags = conf->flags | mask;
3948	else if (nla_get_u8(tb[attrtype]))
3949		flags = conf->flags | mask;
3950	else
3951		flags = conf->flags & ~mask;
3952
3953	conf->flags = flags;
3954
3955	return 0;
3956}
3957
3958static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3959			 struct net_device *dev, struct vxlan_config *conf,
3960			 bool changelink, struct netlink_ext_ack *extack)
3961{
3962	struct vxlan_dev *vxlan = netdev_priv(dev);
3963	int err = 0;
3964
3965	memset(conf, 0, sizeof(*conf));
3966
3967	/* if changelink operation, start with old existing cfg */
3968	if (changelink)
3969		memcpy(conf, &vxlan->cfg, sizeof(*conf));
3970
3971	if (data[IFLA_VXLAN_ID]) {
3972		__be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3973
3974		if (changelink && (vni != conf->vni)) {
3975			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
3976			return -EOPNOTSUPP;
3977		}
3978		conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3979	}
3980
3981	if (data[IFLA_VXLAN_GROUP]) {
3982		if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
3983			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
3984			return -EOPNOTSUPP;
3985		}
3986
3987		conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
3988		conf->remote_ip.sa.sa_family = AF_INET;
3989	} else if (data[IFLA_VXLAN_GROUP6]) {
3990		if (!IS_ENABLED(CONFIG_IPV6)) {
3991			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
3992			return -EPFNOSUPPORT;
3993		}
3994
3995		if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
3996			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
3997			return -EOPNOTSUPP;
3998		}
3999
4000		conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
4001		conf->remote_ip.sa.sa_family = AF_INET6;
4002	}
4003
4004	if (data[IFLA_VXLAN_LOCAL]) {
4005		if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
4006			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
4007			return -EOPNOTSUPP;
4008		}
4009
4010		conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
4011		conf->saddr.sa.sa_family = AF_INET;
4012	} else if (data[IFLA_VXLAN_LOCAL6]) {
4013		if (!IS_ENABLED(CONFIG_IPV6)) {
4014			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
4015			return -EPFNOSUPPORT;
4016		}
4017
4018		if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
4019			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
4020			return -EOPNOTSUPP;
4021		}
4022
4023		/* TODO: respect scope id */
4024		conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
4025		conf->saddr.sa.sa_family = AF_INET6;
4026	}
4027
4028	if (data[IFLA_VXLAN_LINK])
4029		conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
4030
4031	if (data[IFLA_VXLAN_TOS])
4032		conf->tos  = nla_get_u8(data[IFLA_VXLAN_TOS]);
4033
4034	if (data[IFLA_VXLAN_TTL])
4035		conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
4036
4037	if (data[IFLA_VXLAN_TTL_INHERIT]) {
4038		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4039				    VXLAN_F_TTL_INHERIT, changelink, false,
4040				    extack);
4041		if (err)
4042			return err;
4043
4044	}
4045
4046	if (data[IFLA_VXLAN_LABEL])
4047		conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
4048			     IPV6_FLOWLABEL_MASK;
4049
4050	if (data[IFLA_VXLAN_LEARNING]) {
4051		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4052				    VXLAN_F_LEARN, changelink, true,
4053				    extack);
4054		if (err)
4055			return err;
4056	} else if (!changelink) {
4057		/* default to learn on a new device */
4058		conf->flags |= VXLAN_F_LEARN;
4059	}
4060
4061	if (data[IFLA_VXLAN_AGEING])
 
 
4062		conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
 
4063
4064	if (data[IFLA_VXLAN_PROXY]) {
4065		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4066				    VXLAN_F_PROXY, changelink, false,
4067				    extack);
4068		if (err)
4069			return err;
4070	}
4071
4072	if (data[IFLA_VXLAN_RSC]) {
4073		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4074				    VXLAN_F_RSC, changelink, false,
4075				    extack);
4076		if (err)
4077			return err;
4078	}
4079
4080	if (data[IFLA_VXLAN_L2MISS]) {
4081		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4082				    VXLAN_F_L2MISS, changelink, false,
4083				    extack);
4084		if (err)
4085			return err;
4086	}
4087
4088	if (data[IFLA_VXLAN_L3MISS]) {
4089		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4090				    VXLAN_F_L3MISS, changelink, false,
4091				    extack);
4092		if (err)
4093			return err;
4094	}
4095
4096	if (data[IFLA_VXLAN_LIMIT]) {
4097		if (changelink) {
4098			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4099					    "Cannot change limit");
4100			return -EOPNOTSUPP;
4101		}
4102		conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4103	}
4104
4105	if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4106		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4107				    VXLAN_F_COLLECT_METADATA, changelink, false,
4108				    extack);
4109		if (err)
4110			return err;
4111	}
4112
4113	if (data[IFLA_VXLAN_PORT_RANGE]) {
4114		if (!changelink) {
4115			const struct ifla_vxlan_port_range *p
4116				= nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4117			conf->port_min = ntohs(p->low);
4118			conf->port_max = ntohs(p->high);
4119		} else {
4120			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4121					    "Cannot change port range");
4122			return -EOPNOTSUPP;
4123		}
4124	}
4125
4126	if (data[IFLA_VXLAN_PORT]) {
4127		if (changelink) {
4128			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4129					    "Cannot change port");
4130			return -EOPNOTSUPP;
4131		}
4132		conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4133	}
4134
4135	if (data[IFLA_VXLAN_UDP_CSUM]) {
4136		if (changelink) {
4137			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4138					    "Cannot change UDP_CSUM flag");
4139			return -EOPNOTSUPP;
4140		}
4141		if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4142			conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4143	}
4144
4145	if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4146		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4147				    VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4148				    false, extack);
4149		if (err)
4150			return err;
4151	}
4152
4153	if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4154		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4155				    VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4156				    false, extack);
4157		if (err)
4158			return err;
4159	}
4160
4161	if (data[IFLA_VXLAN_REMCSUM_TX]) {
4162		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4163				    VXLAN_F_REMCSUM_TX, changelink, false,
4164				    extack);
4165		if (err)
4166			return err;
4167	}
4168
4169	if (data[IFLA_VXLAN_REMCSUM_RX]) {
4170		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4171				    VXLAN_F_REMCSUM_RX, changelink, false,
4172				    extack);
4173		if (err)
4174			return err;
4175	}
4176
4177	if (data[IFLA_VXLAN_GBP]) {
4178		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4179				    VXLAN_F_GBP, changelink, false, extack);
4180		if (err)
4181			return err;
4182	}
4183
4184	if (data[IFLA_VXLAN_GPE]) {
4185		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4186				    VXLAN_F_GPE, changelink, false,
4187				    extack);
4188		if (err)
4189			return err;
4190	}
4191
4192	if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4193		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4194				    VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4195				    false, extack);
4196		if (err)
4197			return err;
4198	}
4199
4200	if (tb[IFLA_MTU]) {
4201		if (changelink) {
4202			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4203					    "Cannot change mtu");
4204			return -EOPNOTSUPP;
4205		}
4206		conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4207	}
4208
4209	if (data[IFLA_VXLAN_DF])
4210		conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4211
4212	return 0;
4213}
4214
4215static int vxlan_newlink(struct net *src_net, struct net_device *dev,
4216			 struct nlattr *tb[], struct nlattr *data[],
4217			 struct netlink_ext_ack *extack)
4218{
4219	struct vxlan_config conf;
4220	int err;
4221
4222	err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4223	if (err)
4224		return err;
4225
4226	return __vxlan_dev_create(src_net, dev, &conf, extack);
4227}
4228
4229static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4230			    struct nlattr *data[],
4231			    struct netlink_ext_ack *extack)
4232{
4233	struct vxlan_dev *vxlan = netdev_priv(dev);
4234	struct net_device *lowerdev;
 
4235	struct vxlan_config conf;
4236	struct vxlan_rdst *dst;
4237	int err;
4238
4239	dst = &vxlan->default_dst;
4240	err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4241	if (err)
4242		return err;
4243
4244	err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4245				    vxlan, extack);
4246	if (err)
4247		return err;
4248
4249	if (dst->remote_dev == lowerdev)
4250		lowerdev = NULL;
4251
4252	err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4253					     extack);
4254	if (err)
4255		return err;
4256
4257	/* handle default dst entry */
4258	if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
4259		u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4260
4261		spin_lock_bh(&vxlan->hash_lock[hash_index]);
4262		if (!vxlan_addr_any(&conf.remote_ip)) {
4263			err = vxlan_fdb_update(vxlan, all_zeros_mac,
4264					       &conf.remote_ip,
 
 
 
 
 
 
4265					       NUD_REACHABLE | NUD_PERMANENT,
4266					       NLM_F_APPEND | NLM_F_CREATE,
4267					       vxlan->cfg.dst_port,
4268					       conf.vni, conf.vni,
4269					       conf.remote_ifindex,
4270					       NTF_SELF, 0, true, extack);
 
4271			if (err) {
4272				spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4273				netdev_adjacent_change_abort(dst->remote_dev,
4274							     lowerdev, dev);
4275				return err;
4276			}
4277		}
4278		if (!vxlan_addr_any(&dst->remote_ip))
4279			__vxlan_fdb_delete(vxlan, all_zeros_mac,
4280					   dst->remote_ip,
4281					   vxlan->cfg.dst_port,
4282					   dst->remote_vni,
4283					   dst->remote_vni,
4284					   dst->remote_ifindex,
4285					   true);
4286		spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4287	}
4288
4289	if (conf.age_interval != vxlan->cfg.age_interval)
4290		mod_timer(&vxlan->age_timer, jiffies);
4291
4292	netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4293	if (lowerdev && lowerdev != dst->remote_dev)
4294		dst->remote_dev = lowerdev;
4295	vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4296	return 0;
4297}
4298
4299static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4300{
4301	struct vxlan_dev *vxlan = netdev_priv(dev);
4302
4303	vxlan_flush(vxlan, true);
4304
 
4305	list_del(&vxlan->next);
4306	unregister_netdevice_queue(dev, head);
4307	if (vxlan->default_dst.remote_dev)
4308		netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4309}
4310
4311static size_t vxlan_get_size(const struct net_device *dev)
4312{
4313
4314	return nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_ID */
4315		nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4316		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_LINK */
4317		nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4318		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TTL */
4319		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TTL_INHERIT */
4320		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TOS */
4321		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_DF */
4322		nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4323		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_LEARNING */
4324		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_PROXY */
4325		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_RSC */
4326		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_L2MISS */
4327		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_L3MISS */
4328		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_COLLECT_METADATA */
4329		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_AGEING */
4330		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_LIMIT */
4331		nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4332		nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4333		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4334		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4335		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4336		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4337		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4338		0;
4339}
4340
4341static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4342{
4343	const struct vxlan_dev *vxlan = netdev_priv(dev);
4344	const struct vxlan_rdst *dst = &vxlan->default_dst;
4345	struct ifla_vxlan_port_range ports = {
4346		.low =  htons(vxlan->cfg.port_min),
4347		.high = htons(vxlan->cfg.port_max),
4348	};
4349
4350	if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4351		goto nla_put_failure;
4352
4353	if (!vxlan_addr_any(&dst->remote_ip)) {
4354		if (dst->remote_ip.sa.sa_family == AF_INET) {
4355			if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4356					    dst->remote_ip.sin.sin_addr.s_addr))
4357				goto nla_put_failure;
4358#if IS_ENABLED(CONFIG_IPV6)
4359		} else {
4360			if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4361					     &dst->remote_ip.sin6.sin6_addr))
4362				goto nla_put_failure;
4363#endif
4364		}
4365	}
4366
4367	if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4368		goto nla_put_failure;
4369
4370	if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4371		if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4372			if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4373					    vxlan->cfg.saddr.sin.sin_addr.s_addr))
4374				goto nla_put_failure;
4375#if IS_ENABLED(CONFIG_IPV6)
4376		} else {
4377			if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4378					     &vxlan->cfg.saddr.sin6.sin6_addr))
4379				goto nla_put_failure;
4380#endif
4381		}
4382	}
4383
4384	if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4385	    nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4386		       !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4387	    nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4388	    nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4389	    nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4390	    nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4391		       !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4392	    nla_put_u8(skb, IFLA_VXLAN_PROXY,
4393		       !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4394	    nla_put_u8(skb, IFLA_VXLAN_RSC,
4395		       !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4396	    nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4397		       !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4398	    nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4399		       !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4400	    nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4401		       !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4402	    nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4403	    nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4404	    nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4405	    nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4406		       !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4407	    nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4408		       !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4409	    nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4410		       !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4411	    nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4412		       !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4413	    nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4414		       !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
4415		goto nla_put_failure;
4416
4417	if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4418		goto nla_put_failure;
4419
4420	if (vxlan->cfg.flags & VXLAN_F_GBP &&
4421	    nla_put_flag(skb, IFLA_VXLAN_GBP))
4422		goto nla_put_failure;
4423
4424	if (vxlan->cfg.flags & VXLAN_F_GPE &&
4425	    nla_put_flag(skb, IFLA_VXLAN_GPE))
4426		goto nla_put_failure;
4427
4428	if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4429	    nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4430		goto nla_put_failure;
4431
4432	return 0;
4433
4434nla_put_failure:
4435	return -EMSGSIZE;
4436}
4437
4438static struct net *vxlan_get_link_net(const struct net_device *dev)
4439{
4440	struct vxlan_dev *vxlan = netdev_priv(dev);
4441
4442	return vxlan->net;
4443}
4444
4445static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4446	.kind		= "vxlan",
4447	.maxtype	= IFLA_VXLAN_MAX,
4448	.policy		= vxlan_policy,
4449	.priv_size	= sizeof(struct vxlan_dev),
4450	.setup		= vxlan_setup,
4451	.validate	= vxlan_validate,
4452	.newlink	= vxlan_newlink,
4453	.changelink	= vxlan_changelink,
4454	.dellink	= vxlan_dellink,
4455	.get_size	= vxlan_get_size,
4456	.fill_info	= vxlan_fill_info,
4457	.get_link_net	= vxlan_get_link_net,
4458};
4459
4460struct net_device *vxlan_dev_create(struct net *net, const char *name,
4461				    u8 name_assign_type,
4462				    struct vxlan_config *conf)
4463{
4464	struct nlattr *tb[IFLA_MAX + 1];
4465	struct net_device *dev;
4466	int err;
4467
4468	memset(&tb, 0, sizeof(tb));
4469
4470	dev = rtnl_create_link(net, name, name_assign_type,
4471			       &vxlan_link_ops, tb, NULL);
4472	if (IS_ERR(dev))
4473		return dev;
4474
4475	err = __vxlan_dev_create(net, dev, conf, NULL);
4476	if (err < 0) {
4477		free_netdev(dev);
4478		return ERR_PTR(err);
4479	}
4480
4481	err = rtnl_configure_link(dev, NULL);
4482	if (err < 0) {
4483		LIST_HEAD(list_kill);
4484
4485		vxlan_dellink(dev, &list_kill);
4486		unregister_netdevice_many(&list_kill);
4487		return ERR_PTR(err);
4488	}
4489
4490	return dev;
4491}
4492EXPORT_SYMBOL_GPL(vxlan_dev_create);
4493
4494static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4495					     struct net_device *dev)
4496{
4497	struct vxlan_dev *vxlan, *next;
4498	LIST_HEAD(list_kill);
4499
4500	list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4501		struct vxlan_rdst *dst = &vxlan->default_dst;
4502
4503		/* In case we created vxlan device with carrier
4504		 * and we loose the carrier due to module unload
4505		 * we also need to remove vxlan device. In other
4506		 * cases, it's not necessary and remote_ifindex
4507		 * is 0 here, so no matches.
4508		 */
4509		if (dst->remote_ifindex == dev->ifindex)
4510			vxlan_dellink(vxlan->dev, &list_kill);
4511	}
4512
4513	unregister_netdevice_many(&list_kill);
4514}
4515
4516static int vxlan_netdevice_event(struct notifier_block *unused,
4517				 unsigned long event, void *ptr)
4518{
4519	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4520	struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4521
4522	if (event == NETDEV_UNREGISTER) {
4523		if (!dev->udp_tunnel_nic_info)
4524			vxlan_offload_rx_ports(dev, false);
4525		vxlan_handle_lowerdev_unregister(vn, dev);
4526	} else if (event == NETDEV_REGISTER) {
4527		if (!dev->udp_tunnel_nic_info)
4528			vxlan_offload_rx_ports(dev, true);
4529	} else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
4530		   event == NETDEV_UDP_TUNNEL_DROP_INFO) {
4531		vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
4532	}
4533
4534	return NOTIFY_DONE;
4535}
4536
4537static struct notifier_block vxlan_notifier_block __read_mostly = {
4538	.notifier_call = vxlan_netdevice_event,
4539};
4540
4541static void
4542vxlan_fdb_offloaded_set(struct net_device *dev,
4543			struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4544{
4545	struct vxlan_dev *vxlan = netdev_priv(dev);
4546	struct vxlan_rdst *rdst;
4547	struct vxlan_fdb *f;
4548	u32 hash_index;
4549
4550	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4551
4552	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4553
4554	f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4555	if (!f)
4556		goto out;
4557
4558	rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4559				   fdb_info->remote_port,
4560				   fdb_info->remote_vni,
4561				   fdb_info->remote_ifindex);
4562	if (!rdst)
4563		goto out;
4564
4565	rdst->offloaded = fdb_info->offloaded;
4566
4567out:
4568	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4569}
4570
4571static int
4572vxlan_fdb_external_learn_add(struct net_device *dev,
4573			     struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4574{
4575	struct vxlan_dev *vxlan = netdev_priv(dev);
4576	struct netlink_ext_ack *extack;
4577	u32 hash_index;
4578	int err;
4579
4580	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4581	extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4582
4583	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4584	err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4585			       NUD_REACHABLE,
4586			       NLM_F_CREATE | NLM_F_REPLACE,
4587			       fdb_info->remote_port,
4588			       fdb_info->vni,
4589			       fdb_info->remote_vni,
4590			       fdb_info->remote_ifindex,
4591			       NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4592			       0, false, extack);
4593	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4594
4595	return err;
4596}
4597
4598static int
4599vxlan_fdb_external_learn_del(struct net_device *dev,
4600			     struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4601{
4602	struct vxlan_dev *vxlan = netdev_priv(dev);
4603	struct vxlan_fdb *f;
4604	u32 hash_index;
4605	int err = 0;
4606
4607	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4608	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4609
4610	f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4611	if (!f)
4612		err = -ENOENT;
4613	else if (f->flags & NTF_EXT_LEARNED)
4614		err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4615					 fdb_info->remote_ip,
4616					 fdb_info->remote_port,
4617					 fdb_info->vni,
4618					 fdb_info->remote_vni,
4619					 fdb_info->remote_ifindex,
4620					 false);
4621
4622	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4623
4624	return err;
4625}
4626
4627static int vxlan_switchdev_event(struct notifier_block *unused,
4628				 unsigned long event, void *ptr)
4629{
4630	struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4631	struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4632	int err = 0;
4633
4634	switch (event) {
4635	case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4636		vxlan_fdb_offloaded_set(dev, ptr);
4637		break;
4638	case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4639		fdb_info = ptr;
4640		err = vxlan_fdb_external_learn_add(dev, fdb_info);
4641		if (err) {
4642			err = notifier_from_errno(err);
4643			break;
4644		}
4645		fdb_info->offloaded = true;
4646		vxlan_fdb_offloaded_set(dev, fdb_info);
4647		break;
4648	case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4649		fdb_info = ptr;
4650		err = vxlan_fdb_external_learn_del(dev, fdb_info);
4651		if (err) {
4652			err = notifier_from_errno(err);
4653			break;
4654		}
4655		fdb_info->offloaded = false;
4656		vxlan_fdb_offloaded_set(dev, fdb_info);
4657		break;
4658	}
4659
4660	return err;
4661}
4662
4663static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4664	.notifier_call = vxlan_switchdev_event,
4665};
4666
4667static void vxlan_fdb_nh_flush(struct nexthop *nh)
4668{
4669	struct vxlan_fdb *fdb;
4670	struct vxlan_dev *vxlan;
4671	u32 hash_index;
4672
4673	rcu_read_lock();
4674	list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4675		vxlan = rcu_dereference(fdb->vdev);
4676		WARN_ON(!vxlan);
4677		hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4678					    vxlan->default_dst.remote_vni);
4679		spin_lock_bh(&vxlan->hash_lock[hash_index]);
4680		if (!hlist_unhashed(&fdb->hlist))
4681			vxlan_fdb_destroy(vxlan, fdb, false, false);
4682		spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4683	}
4684	rcu_read_unlock();
4685}
4686
4687static int vxlan_nexthop_event(struct notifier_block *nb,
4688			       unsigned long event, void *ptr)
4689{
4690	struct nexthop *nh = ptr;
4691
4692	if (!nh || event != NEXTHOP_EVENT_DEL)
4693		return NOTIFY_DONE;
4694
4695	vxlan_fdb_nh_flush(nh);
4696
4697	return NOTIFY_DONE;
4698}
4699
4700static struct notifier_block vxlan_nexthop_notifier_block __read_mostly = {
4701	.notifier_call = vxlan_nexthop_event,
4702};
4703
4704static __net_init int vxlan_init_net(struct net *net)
4705{
4706	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4707	unsigned int h;
4708
4709	INIT_LIST_HEAD(&vn->vxlan_list);
4710	spin_lock_init(&vn->sock_lock);
4711
4712	for (h = 0; h < PORT_HASH_SIZE; ++h)
4713		INIT_HLIST_HEAD(&vn->sock_list[h]);
4714
4715	return register_nexthop_notifier(net, &vxlan_nexthop_notifier_block);
4716}
4717
4718static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
4719{
4720	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4721	struct vxlan_dev *vxlan, *next;
4722	struct net_device *dev, *aux;
4723	unsigned int h;
4724
4725	for_each_netdev_safe(net, dev, aux)
4726		if (dev->rtnl_link_ops == &vxlan_link_ops)
4727			unregister_netdevice_queue(dev, head);
4728
4729	list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4730		/* If vxlan->dev is in the same netns, it has already been added
4731		 * to the list by the previous loop.
4732		 */
4733		if (!net_eq(dev_net(vxlan->dev), net))
 
4734			unregister_netdevice_queue(vxlan->dev, head);
 
4735	}
4736
4737	for (h = 0; h < PORT_HASH_SIZE; ++h)
4738		WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4739}
4740
4741static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4742{
4743	struct net *net;
4744	LIST_HEAD(list);
4745
4746	rtnl_lock();
4747	list_for_each_entry(net, net_list, exit_list)
4748		unregister_nexthop_notifier(net, &vxlan_nexthop_notifier_block);
4749	list_for_each_entry(net, net_list, exit_list)
4750		vxlan_destroy_tunnels(net, &list);
4751
4752	unregister_netdevice_many(&list);
4753	rtnl_unlock();
4754}
4755
4756static struct pernet_operations vxlan_net_ops = {
4757	.init = vxlan_init_net,
4758	.exit_batch = vxlan_exit_batch_net,
4759	.id   = &vxlan_net_id,
4760	.size = sizeof(struct vxlan_net),
4761};
4762
4763static int __init vxlan_init_module(void)
4764{
4765	int rc;
4766
4767	get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4768
4769	rc = register_pernet_subsys(&vxlan_net_ops);
4770	if (rc)
4771		goto out1;
4772
4773	rc = register_netdevice_notifier(&vxlan_notifier_block);
4774	if (rc)
4775		goto out2;
4776
4777	rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4778	if (rc)
4779		goto out3;
4780
4781	rc = rtnl_link_register(&vxlan_link_ops);
4782	if (rc)
4783		goto out4;
4784
4785	return 0;
4786out4:
4787	unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4788out3:
4789	unregister_netdevice_notifier(&vxlan_notifier_block);
4790out2:
4791	unregister_pernet_subsys(&vxlan_net_ops);
4792out1:
4793	return rc;
4794}
4795late_initcall(vxlan_init_module);
4796
4797static void __exit vxlan_cleanup_module(void)
4798{
4799	rtnl_link_unregister(&vxlan_link_ops);
4800	unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4801	unregister_netdevice_notifier(&vxlan_notifier_block);
4802	unregister_pernet_subsys(&vxlan_net_ops);
4803	/* rcu_barrier() is called by netns */
4804}
4805module_exit(vxlan_cleanup_module);
4806
4807MODULE_LICENSE("GPL");
4808MODULE_VERSION(VXLAN_VERSION);
4809MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4810MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4811MODULE_ALIAS_RTNL_LINK("vxlan");