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