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/gro.h>
  21#include <net/ipv6_stubs.h>
  22#include <net/ip.h>
  23#include <net/icmp.h>
  24#include <net/rtnetlink.h>
  25#include <net/inet_ecn.h>
  26#include <net/net_namespace.h>
  27#include <net/netns/generic.h>
  28#include <net/tun_proto.h>
  29#include <net/vxlan.h>
  30#include <net/nexthop.h>
  31
  32#if IS_ENABLED(CONFIG_IPV6)
  33#include <net/ip6_tunnel.h>
  34#include <net/ip6_checksum.h>
  35#endif
  36
  37#include "vxlan_private.h"
  38
  39#define VXLAN_VERSION	"0.1"
  40
  41#define FDB_AGE_DEFAULT 300 /* 5 min */
  42#define FDB_AGE_INTERVAL (10 * HZ)	/* rescan interval */
  43
  44/* UDP port for VXLAN traffic.
  45 * The IANA assigned port is 4789, but the Linux default is 8472
  46 * for compatibility with early adopters.
  47 */
  48static unsigned short vxlan_port __read_mostly = 8472;
  49module_param_named(udp_port, vxlan_port, ushort, 0444);
  50MODULE_PARM_DESC(udp_port, "Destination UDP port");
  51
  52static bool log_ecn_error = true;
  53module_param(log_ecn_error, bool, 0644);
  54MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  55
  56unsigned int vxlan_net_id;
  57
  58const u8 all_zeros_mac[ETH_ALEN + 2];
  59static struct rtnl_link_ops vxlan_link_ops;
  60
  61static int vxlan_sock_add(struct vxlan_dev *vxlan);
  62
  63static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
  64
  65/* salt for hash table */
  66static u32 vxlan_salt __read_mostly;
  67
  68static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
  69{
  70	return vs->flags & VXLAN_F_COLLECT_METADATA ||
  71	       ip_tunnel_collect_metadata();
  72}
  73
  74/* Find VXLAN socket based on network namespace, address family, UDP port,
  75 * enabled unshareable flags and socket device binding (see l3mdev with
  76 * non-default VRF).
  77 */
  78static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
  79					  __be16 port, u32 flags, int ifindex)
  80{
  81	struct vxlan_sock *vs;
  82
  83	flags &= VXLAN_F_RCV_FLAGS;
  84
  85	hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
  86		if (inet_sk(vs->sock->sk)->inet_sport == port &&
  87		    vxlan_get_sk_family(vs) == family &&
  88		    vs->flags == flags &&
  89		    vs->sock->sk->sk_bound_dev_if == ifindex)
  90			return vs;
  91	}
  92	return NULL;
  93}
  94
  95static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs,
  96					   int ifindex, __be32 vni,
  97					   struct vxlan_vni_node **vninode)
  98{
  99	struct vxlan_vni_node *vnode;
 100	struct vxlan_dev_node *node;
 101
 102	/* For flow based devices, map all packets to VNI 0 */
 103	if (vs->flags & VXLAN_F_COLLECT_METADATA &&
 104	    !(vs->flags & VXLAN_F_VNIFILTER))
 105		vni = 0;
 106
 107	hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
 108		if (!node->vxlan)
 109			continue;
 110		vnode = NULL;
 111		if (node->vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
 112			vnode = vxlan_vnifilter_lookup(node->vxlan, vni);
 113			if (!vnode)
 114				continue;
 115		} else if (node->vxlan->default_dst.remote_vni != vni) {
 116			continue;
 117		}
 118
 119		if (IS_ENABLED(CONFIG_IPV6)) {
 120			const struct vxlan_config *cfg = &node->vxlan->cfg;
 121
 122			if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
 123			    cfg->remote_ifindex != ifindex)
 124				continue;
 125		}
 126
 127		if (vninode)
 128			*vninode = vnode;
 129		return node->vxlan;
 130	}
 131
 132	return NULL;
 133}
 134
 135/* Look up VNI in a per net namespace table */
 136static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
 137					__be32 vni, sa_family_t family,
 138					__be16 port, u32 flags)
 139{
 140	struct vxlan_sock *vs;
 141
 142	vs = vxlan_find_sock(net, family, port, flags, ifindex);
 143	if (!vs)
 144		return NULL;
 145
 146	return vxlan_vs_find_vni(vs, ifindex, vni, NULL);
 147}
 148
 149/* Fill in neighbour message in skbuff. */
 150static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
 151			  const struct vxlan_fdb *fdb,
 152			  u32 portid, u32 seq, int type, unsigned int flags,
 153			  const struct vxlan_rdst *rdst)
 154{
 155	unsigned long now = jiffies;
 156	struct nda_cacheinfo ci;
 157	bool send_ip, send_eth;
 158	struct nlmsghdr *nlh;
 159	struct nexthop *nh;
 160	struct ndmsg *ndm;
 161	int nh_family;
 162	u32 nh_id;
 163
 164	nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
 165	if (nlh == NULL)
 166		return -EMSGSIZE;
 167
 168	ndm = nlmsg_data(nlh);
 169	memset(ndm, 0, sizeof(*ndm));
 170
 171	send_eth = send_ip = true;
 172
 173	rcu_read_lock();
 174	nh = rcu_dereference(fdb->nh);
 175	if (nh) {
 176		nh_family = nexthop_get_family(nh);
 177		nh_id = nh->id;
 178	}
 179	rcu_read_unlock();
 180
 181	if (type == RTM_GETNEIGH) {
 182		if (rdst) {
 183			send_ip = !vxlan_addr_any(&rdst->remote_ip);
 184			ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
 185		} else if (nh) {
 186			ndm->ndm_family = nh_family;
 187		}
 188		send_eth = !is_zero_ether_addr(fdb->eth_addr);
 189	} else
 190		ndm->ndm_family	= AF_BRIDGE;
 191	ndm->ndm_state = fdb->state;
 192	ndm->ndm_ifindex = vxlan->dev->ifindex;
 193	ndm->ndm_flags = fdb->flags;
 194	if (rdst && rdst->offloaded)
 195		ndm->ndm_flags |= NTF_OFFLOADED;
 196	ndm->ndm_type = RTN_UNICAST;
 197
 198	if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
 199	    nla_put_s32(skb, NDA_LINK_NETNSID,
 200			peernet2id(dev_net(vxlan->dev), vxlan->net)))
 201		goto nla_put_failure;
 202
 203	if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
 204		goto nla_put_failure;
 205	if (nh) {
 206		if (nla_put_u32(skb, NDA_NH_ID, nh_id))
 207			goto nla_put_failure;
 208	} else if (rdst) {
 209		if (send_ip && vxlan_nla_put_addr(skb, NDA_DST,
 210						  &rdst->remote_ip))
 211			goto nla_put_failure;
 212
 213		if (rdst->remote_port &&
 214		    rdst->remote_port != vxlan->cfg.dst_port &&
 215		    nla_put_be16(skb, NDA_PORT, rdst->remote_port))
 216			goto nla_put_failure;
 217		if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
 218		    nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
 219			goto nla_put_failure;
 220		if (rdst->remote_ifindex &&
 221		    nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
 222			goto nla_put_failure;
 223	}
 224
 225	if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
 226	    nla_put_u32(skb, NDA_SRC_VNI,
 227			be32_to_cpu(fdb->vni)))
 228		goto nla_put_failure;
 229
 230	ci.ndm_used	 = jiffies_to_clock_t(now - fdb->used);
 231	ci.ndm_confirmed = 0;
 232	ci.ndm_updated	 = jiffies_to_clock_t(now - fdb->updated);
 233	ci.ndm_refcnt	 = 0;
 234
 235	if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
 236		goto nla_put_failure;
 237
 238	nlmsg_end(skb, nlh);
 239	return 0;
 240
 241nla_put_failure:
 242	nlmsg_cancel(skb, nlh);
 243	return -EMSGSIZE;
 244}
 245
 246static inline size_t vxlan_nlmsg_size(void)
 247{
 248	return NLMSG_ALIGN(sizeof(struct ndmsg))
 249		+ nla_total_size(ETH_ALEN) /* NDA_LLADDR */
 250		+ nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
 251		+ nla_total_size(sizeof(__be16)) /* NDA_PORT */
 252		+ nla_total_size(sizeof(__be32)) /* NDA_VNI */
 253		+ nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
 254		+ nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
 255		+ nla_total_size(sizeof(struct nda_cacheinfo));
 256}
 257
 258static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
 259			       struct vxlan_rdst *rd, int type)
 260{
 261	struct net *net = dev_net(vxlan->dev);
 262	struct sk_buff *skb;
 263	int err = -ENOBUFS;
 264
 265	skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
 266	if (skb == NULL)
 267		goto errout;
 268
 269	err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
 270	if (err < 0) {
 271		/* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
 272		WARN_ON(err == -EMSGSIZE);
 273		kfree_skb(skb);
 274		goto errout;
 275	}
 276
 277	rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
 278	return;
 279errout:
 280	if (err < 0)
 281		rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
 282}
 283
 284static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
 285			    const struct vxlan_fdb *fdb,
 286			    const struct vxlan_rdst *rd,
 287			    struct netlink_ext_ack *extack,
 288			    struct switchdev_notifier_vxlan_fdb_info *fdb_info)
 289{
 290	fdb_info->info.dev = vxlan->dev;
 291	fdb_info->info.extack = extack;
 292	fdb_info->remote_ip = rd->remote_ip;
 293	fdb_info->remote_port = rd->remote_port;
 294	fdb_info->remote_vni = rd->remote_vni;
 295	fdb_info->remote_ifindex = rd->remote_ifindex;
 296	memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
 297	fdb_info->vni = fdb->vni;
 298	fdb_info->offloaded = rd->offloaded;
 299	fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
 300}
 301
 302static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
 303					      struct vxlan_fdb *fdb,
 304					      struct vxlan_rdst *rd,
 305					      bool adding,
 306					      struct netlink_ext_ack *extack)
 307{
 308	struct switchdev_notifier_vxlan_fdb_info info;
 309	enum switchdev_notifier_type notifier_type;
 310	int ret;
 311
 312	if (WARN_ON(!rd))
 313		return 0;
 314
 315	notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
 316			       : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
 317	vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
 318	ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
 319				       &info.info, extack);
 320	return notifier_to_errno(ret);
 321}
 322
 323static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
 324			    struct vxlan_rdst *rd, int type, bool swdev_notify,
 325			    struct netlink_ext_ack *extack)
 326{
 327	int err;
 328
 329	if (swdev_notify && rd) {
 330		switch (type) {
 331		case RTM_NEWNEIGH:
 332			err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
 333								 true, extack);
 334			if (err)
 335				return err;
 336			break;
 337		case RTM_DELNEIGH:
 338			vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
 339							   false, extack);
 340			break;
 341		}
 342	}
 343
 344	__vxlan_fdb_notify(vxlan, fdb, rd, type);
 345	return 0;
 346}
 347
 348static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
 349{
 350	struct vxlan_dev *vxlan = netdev_priv(dev);
 351	struct vxlan_fdb f = {
 352		.state = NUD_STALE,
 353	};
 354	struct vxlan_rdst remote = {
 355		.remote_ip = *ipa, /* goes to NDA_DST */
 356		.remote_vni = cpu_to_be32(VXLAN_N_VID),
 357	};
 358
 359	vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
 360}
 361
 362static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
 363{
 364	struct vxlan_fdb f = {
 365		.state = NUD_STALE,
 366	};
 367	struct vxlan_rdst remote = { };
 368
 369	memcpy(f.eth_addr, eth_addr, ETH_ALEN);
 370
 371	vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
 372}
 373
 374/* Hash Ethernet address */
 375static u32 eth_hash(const unsigned char *addr)
 376{
 377	u64 value = get_unaligned((u64 *)addr);
 378
 379	/* only want 6 bytes */
 380#ifdef __BIG_ENDIAN
 381	value >>= 16;
 382#else
 383	value <<= 16;
 384#endif
 385	return hash_64(value, FDB_HASH_BITS);
 386}
 387
 388u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
 389{
 390	/* use 1 byte of OUI and 3 bytes of NIC */
 391	u32 key = get_unaligned((u32 *)(addr + 2));
 392
 393	return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
 394}
 395
 396u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
 397{
 398	if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
 399		return eth_vni_hash(mac, vni);
 400	else
 401		return eth_hash(mac);
 402}
 403
 404/* Hash chain to use given mac address */
 405static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
 406						const u8 *mac, __be32 vni)
 407{
 408	return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
 409}
 410
 411/* Look up Ethernet address in forwarding table */
 412static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
 413					  const u8 *mac, __be32 vni)
 414{
 415	struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
 416	struct vxlan_fdb *f;
 417
 418	hlist_for_each_entry_rcu(f, head, hlist) {
 419		if (ether_addr_equal(mac, f->eth_addr)) {
 420			if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
 421				if (vni == f->vni)
 422					return f;
 423			} else {
 424				return f;
 425			}
 426		}
 427	}
 428
 429	return NULL;
 430}
 431
 432static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
 433					const u8 *mac, __be32 vni)
 434{
 435	struct vxlan_fdb *f;
 436
 437	f = __vxlan_find_mac(vxlan, mac, vni);
 438	if (f && f->used != jiffies)
 439		f->used = jiffies;
 440
 441	return f;
 442}
 443
 444/* caller should hold vxlan->hash_lock */
 445static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
 446					      union vxlan_addr *ip, __be16 port,
 447					      __be32 vni, __u32 ifindex)
 448{
 449	struct vxlan_rdst *rd;
 450
 451	list_for_each_entry(rd, &f->remotes, list) {
 452		if (vxlan_addr_equal(&rd->remote_ip, ip) &&
 453		    rd->remote_port == port &&
 454		    rd->remote_vni == vni &&
 455		    rd->remote_ifindex == ifindex)
 456			return rd;
 457	}
 458
 459	return NULL;
 460}
 461
 462int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
 463		      struct switchdev_notifier_vxlan_fdb_info *fdb_info)
 464{
 465	struct vxlan_dev *vxlan = netdev_priv(dev);
 466	u8 eth_addr[ETH_ALEN + 2] = { 0 };
 467	struct vxlan_rdst *rdst;
 468	struct vxlan_fdb *f;
 469	int rc = 0;
 470
 471	if (is_multicast_ether_addr(mac) ||
 472	    is_zero_ether_addr(mac))
 473		return -EINVAL;
 474
 475	ether_addr_copy(eth_addr, mac);
 476
 477	rcu_read_lock();
 478
 479	f = __vxlan_find_mac(vxlan, eth_addr, vni);
 480	if (!f) {
 481		rc = -ENOENT;
 482		goto out;
 483	}
 484
 485	rdst = first_remote_rcu(f);
 486	vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
 487
 488out:
 489	rcu_read_unlock();
 490	return rc;
 491}
 492EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
 493
 494static int vxlan_fdb_notify_one(struct notifier_block *nb,
 495				const struct vxlan_dev *vxlan,
 496				const struct vxlan_fdb *f,
 497				const struct vxlan_rdst *rdst,
 498				struct netlink_ext_ack *extack)
 499{
 500	struct switchdev_notifier_vxlan_fdb_info fdb_info;
 501	int rc;
 502
 503	vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
 504	rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
 505			       &fdb_info);
 506	return notifier_to_errno(rc);
 507}
 508
 509int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
 510		     struct notifier_block *nb,
 511		     struct netlink_ext_ack *extack)
 512{
 513	struct vxlan_dev *vxlan;
 514	struct vxlan_rdst *rdst;
 515	struct vxlan_fdb *f;
 516	unsigned int h;
 517	int rc = 0;
 518
 519	if (!netif_is_vxlan(dev))
 520		return -EINVAL;
 521	vxlan = netdev_priv(dev);
 522
 523	for (h = 0; h < FDB_HASH_SIZE; ++h) {
 524		spin_lock_bh(&vxlan->hash_lock[h]);
 525		hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
 526			if (f->vni == vni) {
 527				list_for_each_entry(rdst, &f->remotes, list) {
 528					rc = vxlan_fdb_notify_one(nb, vxlan,
 529								  f, rdst,
 530								  extack);
 531					if (rc)
 532						goto unlock;
 533				}
 534			}
 535		}
 536		spin_unlock_bh(&vxlan->hash_lock[h]);
 537	}
 538	return 0;
 539
 540unlock:
 541	spin_unlock_bh(&vxlan->hash_lock[h]);
 542	return rc;
 543}
 544EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
 545
 546void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
 547{
 548	struct vxlan_dev *vxlan;
 549	struct vxlan_rdst *rdst;
 550	struct vxlan_fdb *f;
 551	unsigned int h;
 552
 553	if (!netif_is_vxlan(dev))
 554		return;
 555	vxlan = netdev_priv(dev);
 556
 557	for (h = 0; h < FDB_HASH_SIZE; ++h) {
 558		spin_lock_bh(&vxlan->hash_lock[h]);
 559		hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
 560			if (f->vni == vni)
 561				list_for_each_entry(rdst, &f->remotes, list)
 562					rdst->offloaded = false;
 563		spin_unlock_bh(&vxlan->hash_lock[h]);
 564	}
 565
 566}
 567EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
 568
 569/* Replace destination of unicast mac */
 570static int vxlan_fdb_replace(struct vxlan_fdb *f,
 571			     union vxlan_addr *ip, __be16 port, __be32 vni,
 572			     __u32 ifindex, struct vxlan_rdst *oldrd)
 573{
 574	struct vxlan_rdst *rd;
 575
 576	rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
 577	if (rd)
 578		return 0;
 579
 580	rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
 581	if (!rd)
 582		return 0;
 583
 584	*oldrd = *rd;
 585	dst_cache_reset(&rd->dst_cache);
 586	rd->remote_ip = *ip;
 587	rd->remote_port = port;
 588	rd->remote_vni = vni;
 589	rd->remote_ifindex = ifindex;
 590	rd->offloaded = false;
 591	return 1;
 592}
 593
 594/* Add/update destinations for multicast */
 595static int vxlan_fdb_append(struct vxlan_fdb *f,
 596			    union vxlan_addr *ip, __be16 port, __be32 vni,
 597			    __u32 ifindex, struct vxlan_rdst **rdp)
 598{
 599	struct vxlan_rdst *rd;
 600
 601	rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
 602	if (rd)
 603		return 0;
 604
 605	rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
 606	if (rd == NULL)
 607		return -ENOMEM;
 608
 609	if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
 610		kfree(rd);
 611		return -ENOMEM;
 612	}
 613
 614	rd->remote_ip = *ip;
 615	rd->remote_port = port;
 616	rd->offloaded = false;
 617	rd->remote_vni = vni;
 618	rd->remote_ifindex = ifindex;
 619
 620	list_add_tail_rcu(&rd->list, &f->remotes);
 621
 622	*rdp = rd;
 623	return 1;
 624}
 625
 626static bool vxlan_parse_gpe_proto(struct vxlanhdr *hdr, __be16 *protocol)
 627{
 628	struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)hdr;
 629
 630	/* Need to have Next Protocol set for interfaces in GPE mode. */
 631	if (!gpe->np_applied)
 632		return false;
 633	/* "The initial version is 0. If a receiver does not support the
 634	 * version indicated it MUST drop the packet.
 635	 */
 636	if (gpe->version != 0)
 637		return false;
 638	/* "When the O bit is set to 1, the packet is an OAM packet and OAM
 639	 * processing MUST occur." However, we don't implement OAM
 640	 * processing, thus drop the packet.
 641	 */
 642	if (gpe->oam_flag)
 643		return false;
 644
 645	*protocol = tun_p_to_eth_p(gpe->next_protocol);
 646	if (!*protocol)
 647		return false;
 648
 649	return true;
 650}
 651
 652static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
 653					  unsigned int off,
 654					  struct vxlanhdr *vh, size_t hdrlen,
 655					  __be32 vni_field,
 656					  struct gro_remcsum *grc,
 657					  bool nopartial)
 658{
 659	size_t start, offset;
 660
 661	if (skb->remcsum_offload)
 662		return vh;
 663
 664	if (!NAPI_GRO_CB(skb)->csum_valid)
 665		return NULL;
 666
 667	start = vxlan_rco_start(vni_field);
 668	offset = start + vxlan_rco_offset(vni_field);
 669
 670	vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
 671				     start, offset, grc, nopartial);
 672
 673	skb->remcsum_offload = 1;
 674
 675	return vh;
 676}
 677
 678static struct vxlanhdr *vxlan_gro_prepare_receive(struct sock *sk,
 679						  struct list_head *head,
 680						  struct sk_buff *skb,
 681						  struct gro_remcsum *grc)
 682{
 683	struct sk_buff *p;
 684	struct vxlanhdr *vh, *vh2;
 685	unsigned int hlen, off_vx;
 686	struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
 687	__be32 flags;
 688
 689	skb_gro_remcsum_init(grc);
 690
 691	off_vx = skb_gro_offset(skb);
 692	hlen = off_vx + sizeof(*vh);
 693	vh = skb_gro_header(skb, hlen, off_vx);
 694	if (unlikely(!vh))
 695		return NULL;
 696
 697	skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
 698
 699	flags = vh->vx_flags;
 700
 701	if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
 702		vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
 703				       vh->vx_vni, grc,
 704				       !!(vs->flags &
 705					  VXLAN_F_REMCSUM_NOPARTIAL));
 706
 707		if (!vh)
 708			return NULL;
 709	}
 710
 711	skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
 712
 713	list_for_each_entry(p, head, list) {
 714		if (!NAPI_GRO_CB(p)->same_flow)
 715			continue;
 716
 717		vh2 = (struct vxlanhdr *)(p->data + off_vx);
 718		if (vh->vx_flags != vh2->vx_flags ||
 719		    vh->vx_vni != vh2->vx_vni) {
 720			NAPI_GRO_CB(p)->same_flow = 0;
 721			continue;
 722		}
 723	}
 724
 725	return vh;
 726}
 727
 728static struct sk_buff *vxlan_gro_receive(struct sock *sk,
 729					 struct list_head *head,
 730					 struct sk_buff *skb)
 731{
 732	struct sk_buff *pp = NULL;
 733	struct gro_remcsum grc;
 734	int flush = 1;
 735
 736	if (vxlan_gro_prepare_receive(sk, head, skb, &grc)) {
 737		pp = call_gro_receive(eth_gro_receive, head, skb);
 738		flush = 0;
 739	}
 740	skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
 741	return pp;
 742}
 743
 744static struct sk_buff *vxlan_gpe_gro_receive(struct sock *sk,
 745					     struct list_head *head,
 746					     struct sk_buff *skb)
 747{
 748	const struct packet_offload *ptype;
 749	struct sk_buff *pp = NULL;
 750	struct gro_remcsum grc;
 751	struct vxlanhdr *vh;
 752	__be16 protocol;
 753	int flush = 1;
 754
 755	vh = vxlan_gro_prepare_receive(sk, head, skb, &grc);
 756	if (vh) {
 757		if (!vxlan_parse_gpe_proto(vh, &protocol))
 758			goto out;
 759		ptype = gro_find_receive_by_type(protocol);
 760		if (!ptype)
 761			goto out;
 762		pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb);
 763		flush = 0;
 764	}
 765out:
 766	skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
 767	return pp;
 768}
 769
 770static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
 771{
 772	/* Sets 'skb->inner_mac_header' since we are always called with
 773	 * 'skb->encapsulation' set.
 774	 */
 775	return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
 776}
 777
 778static int vxlan_gpe_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
 779{
 780	struct vxlanhdr *vh = (struct vxlanhdr *)(skb->data + nhoff);
 781	const struct packet_offload *ptype;
 782	int err = -ENOSYS;
 783	__be16 protocol;
 784
 785	if (!vxlan_parse_gpe_proto(vh, &protocol))
 786		return err;
 787	ptype = gro_find_complete_by_type(protocol);
 788	if (ptype)
 789		err = ptype->callbacks.gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
 790	return err;
 791}
 792
 793static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
 794					 __u16 state, __be32 src_vni,
 795					 __u16 ndm_flags)
 796{
 797	struct vxlan_fdb *f;
 798
 799	f = kmalloc(sizeof(*f), GFP_ATOMIC);
 800	if (!f)
 801		return NULL;
 802	f->state = state;
 803	f->flags = ndm_flags;
 804	f->updated = f->used = jiffies;
 805	f->vni = src_vni;
 806	f->nh = NULL;
 807	RCU_INIT_POINTER(f->vdev, vxlan);
 808	INIT_LIST_HEAD(&f->nh_list);
 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_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
 824			       u32 nhid, struct netlink_ext_ack *extack)
 825{
 826	struct nexthop *old_nh = rtnl_dereference(fdb->nh);
 827	struct nexthop *nh;
 828	int err = -EINVAL;
 829
 830	if (old_nh && old_nh->id == nhid)
 831		return 0;
 832
 833	nh = nexthop_find_by_id(vxlan->net, nhid);
 834	if (!nh) {
 835		NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
 836		goto err_inval;
 837	}
 838
 839	if (!nexthop_get(nh)) {
 840		NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
 841		nh = NULL;
 842		goto err_inval;
 843	}
 844	if (!nexthop_is_fdb(nh)) {
 845		NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
 846		goto err_inval;
 847	}
 848
 849	if (!nexthop_is_multipath(nh)) {
 850		NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
 851		goto err_inval;
 852	}
 853
 854	/* check nexthop group family */
 855	switch (vxlan->default_dst.remote_ip.sa.sa_family) {
 856	case AF_INET:
 857		if (!nexthop_has_v4(nh)) {
 858			err = -EAFNOSUPPORT;
 859			NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
 860			goto err_inval;
 861		}
 862		break;
 863	case AF_INET6:
 864		if (nexthop_has_v4(nh)) {
 865			err = -EAFNOSUPPORT;
 866			NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
 867			goto err_inval;
 868		}
 869	}
 870
 871	if (old_nh) {
 872		list_del_rcu(&fdb->nh_list);
 873		nexthop_put(old_nh);
 874	}
 875	rcu_assign_pointer(fdb->nh, nh);
 876	list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
 877	return 1;
 878
 879err_inval:
 880	if (nh)
 881		nexthop_put(nh);
 882	return err;
 883}
 884
 885int vxlan_fdb_create(struct vxlan_dev *vxlan,
 886		     const u8 *mac, union vxlan_addr *ip,
 887		     __u16 state, __be16 port, __be32 src_vni,
 888		     __be32 vni, __u32 ifindex, __u16 ndm_flags,
 889		     u32 nhid, struct vxlan_fdb **fdb,
 890		     struct netlink_ext_ack *extack)
 891{
 892	struct vxlan_rdst *rd = NULL;
 893	struct vxlan_fdb *f;
 894	int rc;
 895
 896	if (vxlan->cfg.addrmax &&
 897	    vxlan->addrcnt >= vxlan->cfg.addrmax)
 898		return -ENOSPC;
 899
 900	netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
 901	f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
 902	if (!f)
 903		return -ENOMEM;
 904
 905	if (nhid)
 906		rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
 907	else
 908		rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
 909	if (rc < 0)
 910		goto errout;
 911
 912	*fdb = f;
 913
 914	return 0;
 915
 916errout:
 917	kfree(f);
 918	return rc;
 919}
 920
 921static void __vxlan_fdb_free(struct vxlan_fdb *f)
 922{
 923	struct vxlan_rdst *rd, *nd;
 924	struct nexthop *nh;
 925
 926	nh = rcu_dereference_raw(f->nh);
 927	if (nh) {
 928		rcu_assign_pointer(f->nh, NULL);
 929		rcu_assign_pointer(f->vdev, NULL);
 930		nexthop_put(nh);
 931	}
 932
 933	list_for_each_entry_safe(rd, nd, &f->remotes, list) {
 934		dst_cache_destroy(&rd->dst_cache);
 935		kfree(rd);
 936	}
 937	kfree(f);
 938}
 939
 940static void vxlan_fdb_free(struct rcu_head *head)
 941{
 942	struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
 943
 944	__vxlan_fdb_free(f);
 945}
 946
 947static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
 948			      bool do_notify, bool swdev_notify)
 949{
 950	struct vxlan_rdst *rd;
 951
 952	netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
 953
 954	--vxlan->addrcnt;
 955	if (do_notify) {
 956		if (rcu_access_pointer(f->nh))
 957			vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
 958					 swdev_notify, NULL);
 959		else
 960			list_for_each_entry(rd, &f->remotes, list)
 961				vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
 962						 swdev_notify, NULL);
 963	}
 964
 965	hlist_del_rcu(&f->hlist);
 966	list_del_rcu(&f->nh_list);
 967	call_rcu(&f->rcu, vxlan_fdb_free);
 968}
 969
 970static void vxlan_dst_free(struct rcu_head *head)
 971{
 972	struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
 973
 974	dst_cache_destroy(&rd->dst_cache);
 975	kfree(rd);
 976}
 977
 978static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
 979				     union vxlan_addr *ip,
 980				     __u16 state, __u16 flags,
 981				     __be16 port, __be32 vni,
 982				     __u32 ifindex, __u16 ndm_flags,
 983				     struct vxlan_fdb *f, u32 nhid,
 984				     bool swdev_notify,
 985				     struct netlink_ext_ack *extack)
 986{
 987	__u16 fdb_flags = (ndm_flags & ~NTF_USE);
 988	struct vxlan_rdst *rd = NULL;
 989	struct vxlan_rdst oldrd;
 990	int notify = 0;
 991	int rc = 0;
 992	int err;
 993
 994	if (nhid && !rcu_access_pointer(f->nh)) {
 995		NL_SET_ERR_MSG(extack,
 996			       "Cannot replace an existing non nexthop fdb with a nexthop");
 997		return -EOPNOTSUPP;
 998	}
 999
1000	if (nhid && (flags & NLM_F_APPEND)) {
1001		NL_SET_ERR_MSG(extack,
1002			       "Cannot append to a nexthop fdb");
1003		return -EOPNOTSUPP;
1004	}
1005
1006	/* Do not allow an externally learned entry to take over an entry added
1007	 * by the user.
1008	 */
1009	if (!(fdb_flags & NTF_EXT_LEARNED) ||
1010	    !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1011		if (f->state != state) {
1012			f->state = state;
1013			f->updated = jiffies;
1014			notify = 1;
1015		}
1016		if (f->flags != fdb_flags) {
1017			f->flags = fdb_flags;
1018			f->updated = jiffies;
1019			notify = 1;
1020		}
1021	}
1022
1023	if ((flags & NLM_F_REPLACE)) {
1024		/* Only change unicasts */
1025		if (!(is_multicast_ether_addr(f->eth_addr) ||
1026		      is_zero_ether_addr(f->eth_addr))) {
1027			if (nhid) {
1028				rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1029				if (rc < 0)
1030					return rc;
1031			} else {
1032				rc = vxlan_fdb_replace(f, ip, port, vni,
1033						       ifindex, &oldrd);
1034			}
1035			notify |= rc;
1036		} else {
1037			NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
1038			return -EOPNOTSUPP;
1039		}
1040	}
1041	if ((flags & NLM_F_APPEND) &&
1042	    (is_multicast_ether_addr(f->eth_addr) ||
1043	     is_zero_ether_addr(f->eth_addr))) {
1044		rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1045
1046		if (rc < 0)
1047			return rc;
1048		notify |= rc;
1049	}
1050
1051	if (ndm_flags & NTF_USE)
1052		f->used = jiffies;
1053
1054	if (notify) {
1055		if (rd == NULL)
1056			rd = first_remote_rtnl(f);
1057
1058		err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
1059				       swdev_notify, extack);
1060		if (err)
1061			goto err_notify;
1062	}
1063
1064	return 0;
1065
1066err_notify:
1067	if (nhid)
1068		return err;
1069	if ((flags & NLM_F_REPLACE) && rc)
1070		*rd = oldrd;
1071	else if ((flags & NLM_F_APPEND) && rc) {
1072		list_del_rcu(&rd->list);
1073		call_rcu(&rd->rcu, vxlan_dst_free);
1074	}
1075	return err;
1076}
1077
1078static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
1079				   const u8 *mac, union vxlan_addr *ip,
1080				   __u16 state, __u16 flags,
1081				   __be16 port, __be32 src_vni, __be32 vni,
1082				   __u32 ifindex, __u16 ndm_flags, u32 nhid,
1083				   bool swdev_notify,
1084				   struct netlink_ext_ack *extack)
1085{
1086	__u16 fdb_flags = (ndm_flags & ~NTF_USE);
1087	struct vxlan_fdb *f;
1088	int rc;
1089
1090	/* Disallow replace to add a multicast entry */
1091	if ((flags & NLM_F_REPLACE) &&
1092	    (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1093		return -EOPNOTSUPP;
1094
1095	netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1096	rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1097			      vni, ifindex, fdb_flags, nhid, &f, extack);
1098	if (rc < 0)
1099		return rc;
1100
1101	vxlan_fdb_insert(vxlan, mac, src_vni, f);
1102	rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1103			      swdev_notify, extack);
1104	if (rc)
1105		goto err_notify;
1106
1107	return 0;
1108
1109err_notify:
1110	vxlan_fdb_destroy(vxlan, f, false, false);
1111	return rc;
1112}
1113
1114/* Add new entry to forwarding table -- assumes lock held */
1115int vxlan_fdb_update(struct vxlan_dev *vxlan,
1116		     const u8 *mac, union vxlan_addr *ip,
1117		     __u16 state, __u16 flags,
1118		     __be16 port, __be32 src_vni, __be32 vni,
1119		     __u32 ifindex, __u16 ndm_flags, u32 nhid,
1120		     bool swdev_notify,
1121		     struct netlink_ext_ack *extack)
1122{
1123	struct vxlan_fdb *f;
1124
1125	f = __vxlan_find_mac(vxlan, mac, src_vni);
1126	if (f) {
1127		if (flags & NLM_F_EXCL) {
1128			netdev_dbg(vxlan->dev,
1129				   "lost race to create %pM\n", mac);
1130			return -EEXIST;
1131		}
1132
1133		return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1134						 vni, ifindex, ndm_flags, f,
1135						 nhid, swdev_notify, extack);
1136	} else {
1137		if (!(flags & NLM_F_CREATE))
1138			return -ENOENT;
1139
1140		return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1141					       port, src_vni, vni, ifindex,
1142					       ndm_flags, nhid, swdev_notify,
1143					       extack);
1144	}
1145}
1146
1147static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1148				  struct vxlan_rdst *rd, bool swdev_notify)
1149{
1150	list_del_rcu(&rd->list);
1151	vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1152	call_rcu(&rd->rcu, vxlan_dst_free);
1153}
1154
1155static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1156			   union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1157			   __be32 *vni, u32 *ifindex, u32 *nhid,
1158			   struct netlink_ext_ack *extack)
1159{
1160	struct net *net = dev_net(vxlan->dev);
1161	int err;
1162
1163	if (tb[NDA_NH_ID] &&
1164	    (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) {
1165		NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID");
1166		return -EINVAL;
1167	}
1168
1169	if (tb[NDA_DST]) {
1170		err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1171		if (err) {
1172			NL_SET_ERR_MSG(extack, "Unsupported address family");
1173			return err;
1174		}
1175	} else {
1176		union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1177
1178		if (remote->sa.sa_family == AF_INET) {
1179			ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1180			ip->sa.sa_family = AF_INET;
1181#if IS_ENABLED(CONFIG_IPV6)
1182		} else {
1183			ip->sin6.sin6_addr = in6addr_any;
1184			ip->sa.sa_family = AF_INET6;
1185#endif
1186		}
1187	}
1188
1189	if (tb[NDA_PORT]) {
1190		if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) {
1191			NL_SET_ERR_MSG(extack, "Invalid vxlan port");
1192			return -EINVAL;
1193		}
1194		*port = nla_get_be16(tb[NDA_PORT]);
1195	} else {
1196		*port = vxlan->cfg.dst_port;
1197	}
1198
1199	if (tb[NDA_VNI]) {
1200		if (nla_len(tb[NDA_VNI]) != sizeof(u32)) {
1201			NL_SET_ERR_MSG(extack, "Invalid vni");
1202			return -EINVAL;
1203		}
1204		*vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1205	} else {
1206		*vni = vxlan->default_dst.remote_vni;
1207	}
1208
1209	if (tb[NDA_SRC_VNI]) {
1210		if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) {
1211			NL_SET_ERR_MSG(extack, "Invalid src vni");
1212			return -EINVAL;
1213		}
1214		*src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1215	} else {
1216		*src_vni = vxlan->default_dst.remote_vni;
1217	}
1218
1219	if (tb[NDA_IFINDEX]) {
1220		struct net_device *tdev;
1221
1222		if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) {
1223			NL_SET_ERR_MSG(extack, "Invalid ifindex");
1224			return -EINVAL;
1225		}
1226		*ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1227		tdev = __dev_get_by_index(net, *ifindex);
1228		if (!tdev) {
1229			NL_SET_ERR_MSG(extack, "Device not found");
1230			return -EADDRNOTAVAIL;
1231		}
1232	} else {
1233		*ifindex = 0;
1234	}
1235
1236	if (tb[NDA_NH_ID])
1237		*nhid = nla_get_u32(tb[NDA_NH_ID]);
1238	else
1239		*nhid = 0;
1240
1241	return 0;
1242}
1243
1244/* Add static entry (via netlink) */
1245static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1246			 struct net_device *dev,
1247			 const unsigned char *addr, u16 vid, u16 flags,
1248			 struct netlink_ext_ack *extack)
1249{
1250	struct vxlan_dev *vxlan = netdev_priv(dev);
1251	/* struct net *net = dev_net(vxlan->dev); */
1252	union vxlan_addr ip;
1253	__be16 port;
1254	__be32 src_vni, vni;
1255	u32 ifindex, nhid;
1256	u32 hash_index;
1257	int err;
1258
1259	if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1260		pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1261			ndm->ndm_state);
1262		return -EINVAL;
1263	}
1264
1265	if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
1266		return -EINVAL;
1267
1268	err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1269			      &nhid, extack);
1270	if (err)
1271		return err;
1272
1273	if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1274		return -EAFNOSUPPORT;
1275
1276	hash_index = fdb_head_index(vxlan, addr, src_vni);
1277	spin_lock_bh(&vxlan->hash_lock[hash_index]);
1278	err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1279			       port, src_vni, vni, ifindex,
1280			       ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1281			       nhid, true, extack);
1282	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1283
1284	return err;
1285}
1286
1287int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1288		       const unsigned char *addr, union vxlan_addr ip,
1289		       __be16 port, __be32 src_vni, __be32 vni,
1290		       u32 ifindex, bool swdev_notify)
1291{
1292	struct vxlan_rdst *rd = NULL;
1293	struct vxlan_fdb *f;
1294	int err = -ENOENT;
1295
1296	f = vxlan_find_mac(vxlan, addr, src_vni);
1297	if (!f)
1298		return err;
1299
1300	if (!vxlan_addr_any(&ip)) {
1301		rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1302		if (!rd)
1303			goto out;
1304	}
1305
1306	/* remove a destination if it's not the only one on the list,
1307	 * otherwise destroy the fdb entry
1308	 */
1309	if (rd && !list_is_singular(&f->remotes)) {
1310		vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1311		goto out;
1312	}
1313
1314	vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1315
1316out:
1317	return 0;
1318}
1319
1320/* Delete entry (via netlink) */
1321static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1322			    struct net_device *dev,
1323			    const unsigned char *addr, u16 vid,
1324			    struct netlink_ext_ack *extack)
1325{
1326	struct vxlan_dev *vxlan = netdev_priv(dev);
1327	union vxlan_addr ip;
1328	__be32 src_vni, vni;
1329	u32 ifindex, nhid;
1330	u32 hash_index;
1331	__be16 port;
1332	int err;
1333
1334	err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1335			      &nhid, extack);
1336	if (err)
1337		return err;
1338
1339	hash_index = fdb_head_index(vxlan, addr, src_vni);
1340	spin_lock_bh(&vxlan->hash_lock[hash_index]);
1341	err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1342				 true);
1343	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1344
1345	return err;
1346}
1347
1348/* Dump forwarding table */
1349static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1350			  struct net_device *dev,
1351			  struct net_device *filter_dev, int *idx)
1352{
1353	struct vxlan_dev *vxlan = netdev_priv(dev);
1354	unsigned int h;
1355	int err = 0;
1356
1357	for (h = 0; h < FDB_HASH_SIZE; ++h) {
1358		struct vxlan_fdb *f;
1359
1360		rcu_read_lock();
1361		hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1362			struct vxlan_rdst *rd;
1363
1364			if (rcu_access_pointer(f->nh)) {
1365				if (*idx < cb->args[2])
1366					goto skip_nh;
1367				err = vxlan_fdb_info(skb, vxlan, f,
1368						     NETLINK_CB(cb->skb).portid,
1369						     cb->nlh->nlmsg_seq,
1370						     RTM_NEWNEIGH,
1371						     NLM_F_MULTI, NULL);
1372				if (err < 0) {
1373					rcu_read_unlock();
1374					goto out;
1375				}
1376skip_nh:
1377				*idx += 1;
1378				continue;
1379			}
1380
1381			list_for_each_entry_rcu(rd, &f->remotes, list) {
1382				if (*idx < cb->args[2])
1383					goto skip;
1384
1385				err = vxlan_fdb_info(skb, vxlan, f,
1386						     NETLINK_CB(cb->skb).portid,
1387						     cb->nlh->nlmsg_seq,
1388						     RTM_NEWNEIGH,
1389						     NLM_F_MULTI, rd);
1390				if (err < 0) {
1391					rcu_read_unlock();
1392					goto out;
1393				}
1394skip:
1395				*idx += 1;
1396			}
1397		}
1398		rcu_read_unlock();
1399	}
1400out:
1401	return err;
1402}
1403
1404static int vxlan_fdb_get(struct sk_buff *skb,
1405			 struct nlattr *tb[],
1406			 struct net_device *dev,
1407			 const unsigned char *addr,
1408			 u16 vid, u32 portid, u32 seq,
1409			 struct netlink_ext_ack *extack)
1410{
1411	struct vxlan_dev *vxlan = netdev_priv(dev);
1412	struct vxlan_fdb *f;
1413	__be32 vni;
1414	int err;
1415
1416	if (tb[NDA_VNI])
1417		vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1418	else
1419		vni = vxlan->default_dst.remote_vni;
1420
1421	rcu_read_lock();
1422
1423	f = __vxlan_find_mac(vxlan, addr, vni);
1424	if (!f) {
1425		NL_SET_ERR_MSG(extack, "Fdb entry not found");
1426		err = -ENOENT;
1427		goto errout;
1428	}
1429
1430	err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1431			     RTM_NEWNEIGH, 0, first_remote_rcu(f));
1432errout:
1433	rcu_read_unlock();
1434	return err;
1435}
1436
1437/* Watch incoming packets to learn mapping between Ethernet address
1438 * and Tunnel endpoint.
1439 * Return true if packet is bogus and should be dropped.
1440 */
1441static bool vxlan_snoop(struct net_device *dev,
1442			union vxlan_addr *src_ip, const u8 *src_mac,
1443			u32 src_ifindex, __be32 vni)
1444{
1445	struct vxlan_dev *vxlan = netdev_priv(dev);
1446	struct vxlan_fdb *f;
1447	u32 ifindex = 0;
1448
1449#if IS_ENABLED(CONFIG_IPV6)
1450	if (src_ip->sa.sa_family == AF_INET6 &&
1451	    (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1452		ifindex = src_ifindex;
1453#endif
1454
1455	f = vxlan_find_mac(vxlan, src_mac, vni);
1456	if (likely(f)) {
1457		struct vxlan_rdst *rdst = first_remote_rcu(f);
1458
1459		if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1460			   rdst->remote_ifindex == ifindex))
1461			return false;
1462
1463		/* Don't migrate static entries, drop packets */
1464		if (f->state & (NUD_PERMANENT | NUD_NOARP))
1465			return true;
1466
1467		/* Don't override an fdb with nexthop with a learnt entry */
1468		if (rcu_access_pointer(f->nh))
1469			return true;
1470
1471		if (net_ratelimit())
1472			netdev_info(dev,
1473				    "%pM migrated from %pIS to %pIS\n",
1474				    src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1475
1476		rdst->remote_ip = *src_ip;
1477		f->updated = jiffies;
1478		vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1479	} else {
1480		u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1481
1482		/* learned new entry */
1483		spin_lock(&vxlan->hash_lock[hash_index]);
1484
1485		/* close off race between vxlan_flush and incoming packets */
1486		if (netif_running(dev))
1487			vxlan_fdb_update(vxlan, src_mac, src_ip,
1488					 NUD_REACHABLE,
1489					 NLM_F_EXCL|NLM_F_CREATE,
1490					 vxlan->cfg.dst_port,
1491					 vni,
1492					 vxlan->default_dst.remote_vni,
1493					 ifindex, NTF_SELF, 0, true, NULL);
1494		spin_unlock(&vxlan->hash_lock[hash_index]);
1495	}
1496
1497	return false;
1498}
1499
1500static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1501{
1502	struct vxlan_net *vn;
1503
1504	if (!vs)
1505		return false;
1506	if (!refcount_dec_and_test(&vs->refcnt))
1507		return false;
1508
1509	vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1510	spin_lock(&vn->sock_lock);
1511	hlist_del_rcu(&vs->hlist);
1512	udp_tunnel_notify_del_rx_port(vs->sock,
1513				      (vs->flags & VXLAN_F_GPE) ?
1514				      UDP_TUNNEL_TYPE_VXLAN_GPE :
1515				      UDP_TUNNEL_TYPE_VXLAN);
1516	spin_unlock(&vn->sock_lock);
1517
1518	return true;
1519}
1520
1521static void vxlan_sock_release(struct vxlan_dev *vxlan)
1522{
1523	struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1524#if IS_ENABLED(CONFIG_IPV6)
1525	struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1526
1527	RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1528#endif
1529
1530	RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1531	synchronize_net();
1532
1533	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
1534		vxlan_vs_del_vnigrp(vxlan);
1535	else
1536		vxlan_vs_del_dev(vxlan);
1537
1538	if (__vxlan_sock_release_prep(sock4)) {
1539		udp_tunnel_sock_release(sock4->sock);
1540		kfree(sock4);
1541	}
1542
1543#if IS_ENABLED(CONFIG_IPV6)
1544	if (__vxlan_sock_release_prep(sock6)) {
1545		udp_tunnel_sock_release(sock6->sock);
1546		kfree(sock6);
1547	}
1548#endif
1549}
1550
1551static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1552			  struct sk_buff *skb, u32 vxflags)
1553{
1554	size_t start, offset;
1555
1556	if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1557		goto out;
1558
1559	start = vxlan_rco_start(unparsed->vx_vni);
1560	offset = start + vxlan_rco_offset(unparsed->vx_vni);
1561
1562	if (!pskb_may_pull(skb, offset + sizeof(u16)))
1563		return false;
1564
1565	skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1566			    !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1567out:
1568	unparsed->vx_flags &= ~VXLAN_HF_RCO;
1569	unparsed->vx_vni &= VXLAN_VNI_MASK;
1570	return true;
1571}
1572
1573static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1574				struct sk_buff *skb, u32 vxflags,
1575				struct vxlan_metadata *md)
1576{
1577	struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1578	struct metadata_dst *tun_dst;
1579
1580	if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1581		goto out;
1582
1583	md->gbp = ntohs(gbp->policy_id);
1584
1585	tun_dst = (struct metadata_dst *)skb_dst(skb);
1586	if (tun_dst) {
1587		tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1588		tun_dst->u.tun_info.options_len = sizeof(*md);
1589	}
1590	if (gbp->dont_learn)
1591		md->gbp |= VXLAN_GBP_DONT_LEARN;
1592
1593	if (gbp->policy_applied)
1594		md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1595
1596	/* In flow-based mode, GBP is carried in dst_metadata */
1597	if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1598		skb->mark = md->gbp;
1599out:
1600	unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1601}
1602
1603static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1604			  struct vxlan_sock *vs,
1605			  struct sk_buff *skb, __be32 vni)
1606{
1607	union vxlan_addr saddr;
1608	u32 ifindex = skb->dev->ifindex;
1609
1610	skb_reset_mac_header(skb);
1611	skb->protocol = eth_type_trans(skb, vxlan->dev);
1612	skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1613
1614	/* Ignore packet loops (and multicast echo) */
1615	if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1616		return false;
1617
1618	/* Get address from the outer IP header */
1619	if (vxlan_get_sk_family(vs) == AF_INET) {
1620		saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1621		saddr.sa.sa_family = AF_INET;
1622#if IS_ENABLED(CONFIG_IPV6)
1623	} else {
1624		saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1625		saddr.sa.sa_family = AF_INET6;
1626#endif
1627	}
1628
1629	if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1630	    vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1631		return false;
1632
1633	return true;
1634}
1635
1636static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1637				  struct sk_buff *skb)
1638{
1639	int err = 0;
1640
1641	if (vxlan_get_sk_family(vs) == AF_INET)
1642		err = IP_ECN_decapsulate(oiph, skb);
1643#if IS_ENABLED(CONFIG_IPV6)
1644	else
1645		err = IP6_ECN_decapsulate(oiph, skb);
1646#endif
1647
1648	if (unlikely(err) && log_ecn_error) {
1649		if (vxlan_get_sk_family(vs) == AF_INET)
1650			net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1651					     &((struct iphdr *)oiph)->saddr,
1652					     ((struct iphdr *)oiph)->tos);
1653		else
1654			net_info_ratelimited("non-ECT from %pI6\n",
1655					     &((struct ipv6hdr *)oiph)->saddr);
1656	}
1657	return err <= 1;
1658}
1659
1660/* Callback from net/ipv4/udp.c to receive packets */
1661static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1662{
1663	struct vxlan_vni_node *vninode = NULL;
1664	struct vxlan_dev *vxlan;
1665	struct vxlan_sock *vs;
1666	struct vxlanhdr unparsed;
1667	struct vxlan_metadata _md;
1668	struct vxlan_metadata *md = &_md;
1669	__be16 protocol = htons(ETH_P_TEB);
1670	bool raw_proto = false;
1671	void *oiph;
1672	__be32 vni = 0;
1673
1674	/* Need UDP and VXLAN header to be present */
1675	if (!pskb_may_pull(skb, VXLAN_HLEN))
1676		goto drop;
1677
1678	unparsed = *vxlan_hdr(skb);
1679	/* VNI flag always required to be set */
1680	if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1681		netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1682			   ntohl(vxlan_hdr(skb)->vx_flags),
1683			   ntohl(vxlan_hdr(skb)->vx_vni));
1684		/* Return non vxlan pkt */
1685		goto drop;
1686	}
1687	unparsed.vx_flags &= ~VXLAN_HF_VNI;
1688	unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1689
1690	vs = rcu_dereference_sk_user_data(sk);
1691	if (!vs)
1692		goto drop;
1693
1694	vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1695
1696	vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode);
1697	if (!vxlan)
1698		goto drop;
1699
1700	/* For backwards compatibility, only allow reserved fields to be
1701	 * used by VXLAN extensions if explicitly requested.
1702	 */
1703	if (vs->flags & VXLAN_F_GPE) {
1704		if (!vxlan_parse_gpe_proto(&unparsed, &protocol))
1705			goto drop;
1706		unparsed.vx_flags &= ~VXLAN_GPE_USED_BITS;
1707		raw_proto = true;
1708	}
1709
1710	if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1711				   !net_eq(vxlan->net, dev_net(vxlan->dev))))
1712		goto drop;
1713
1714	if (vs->flags & VXLAN_F_REMCSUM_RX)
1715		if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags)))
1716			goto drop;
1717
1718	if (vxlan_collect_metadata(vs)) {
1719		struct metadata_dst *tun_dst;
1720
1721		tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1722					 key32_to_tunnel_id(vni), sizeof(*md));
1723
1724		if (!tun_dst)
1725			goto drop;
1726
1727		md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1728
1729		skb_dst_set(skb, (struct dst_entry *)tun_dst);
1730	} else {
1731		memset(md, 0, sizeof(*md));
1732	}
1733
1734	if (vs->flags & VXLAN_F_GBP)
1735		vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1736	/* Note that GBP and GPE can never be active together. This is
1737	 * ensured in vxlan_dev_configure.
1738	 */
1739
1740	if (unparsed.vx_flags || unparsed.vx_vni) {
1741		/* If there are any unprocessed flags remaining treat
1742		 * this as a malformed packet. This behavior diverges from
1743		 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1744		 * in reserved fields are to be ignored. The approach here
1745		 * maintains compatibility with previous stack code, and also
1746		 * is more robust and provides a little more security in
1747		 * adding extensions to VXLAN.
1748		 */
1749		goto drop;
1750	}
1751
1752	if (!raw_proto) {
1753		if (!vxlan_set_mac(vxlan, vs, skb, vni))
1754			goto drop;
1755	} else {
1756		skb_reset_mac_header(skb);
1757		skb->dev = vxlan->dev;
1758		skb->pkt_type = PACKET_HOST;
1759	}
1760
1761	oiph = skb_network_header(skb);
1762	skb_reset_network_header(skb);
1763
1764	if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1765		++vxlan->dev->stats.rx_frame_errors;
1766		++vxlan->dev->stats.rx_errors;
1767		vxlan_vnifilter_count(vxlan, vni, vninode,
1768				      VXLAN_VNI_STATS_RX_ERRORS, 0);
1769		goto drop;
1770	}
1771
1772	rcu_read_lock();
1773
1774	if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1775		rcu_read_unlock();
1776		dev_core_stats_rx_dropped_inc(vxlan->dev);
1777		vxlan_vnifilter_count(vxlan, vni, vninode,
1778				      VXLAN_VNI_STATS_RX_DROPS, 0);
1779		goto drop;
1780	}
1781
1782	dev_sw_netstats_rx_add(vxlan->dev, skb->len);
1783	vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len);
1784	gro_cells_receive(&vxlan->gro_cells, skb);
1785
1786	rcu_read_unlock();
1787
1788	return 0;
1789
1790drop:
1791	/* Consume bad packet */
1792	kfree_skb(skb);
1793	return 0;
1794}
1795
1796/* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1797static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1798{
1799	struct vxlan_dev *vxlan;
1800	struct vxlan_sock *vs;
1801	struct vxlanhdr *hdr;
1802	__be32 vni;
1803
1804	if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1805		return -EINVAL;
1806
1807	hdr = vxlan_hdr(skb);
1808
1809	if (!(hdr->vx_flags & VXLAN_HF_VNI))
1810		return -EINVAL;
1811
1812	vs = rcu_dereference_sk_user_data(sk);
1813	if (!vs)
1814		return -ENOENT;
1815
1816	vni = vxlan_vni(hdr->vx_vni);
1817	vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL);
1818	if (!vxlan)
1819		return -ENOENT;
1820
1821	return 0;
1822}
1823
1824static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1825{
1826	struct vxlan_dev *vxlan = netdev_priv(dev);
1827	struct arphdr *parp;
1828	u8 *arpptr, *sha;
1829	__be32 sip, tip;
1830	struct neighbour *n;
1831
1832	if (dev->flags & IFF_NOARP)
1833		goto out;
1834
1835	if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1836		dev->stats.tx_dropped++;
1837		goto out;
1838	}
1839	parp = arp_hdr(skb);
1840
1841	if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1842	     parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1843	    parp->ar_pro != htons(ETH_P_IP) ||
1844	    parp->ar_op != htons(ARPOP_REQUEST) ||
1845	    parp->ar_hln != dev->addr_len ||
1846	    parp->ar_pln != 4)
1847		goto out;
1848	arpptr = (u8 *)parp + sizeof(struct arphdr);
1849	sha = arpptr;
1850	arpptr += dev->addr_len;	/* sha */
1851	memcpy(&sip, arpptr, sizeof(sip));
1852	arpptr += sizeof(sip);
1853	arpptr += dev->addr_len;	/* tha */
1854	memcpy(&tip, arpptr, sizeof(tip));
1855
1856	if (ipv4_is_loopback(tip) ||
1857	    ipv4_is_multicast(tip))
1858		goto out;
1859
1860	n = neigh_lookup(&arp_tbl, &tip, dev);
1861
1862	if (n) {
1863		struct vxlan_fdb *f;
1864		struct sk_buff	*reply;
1865
1866		if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) {
1867			neigh_release(n);
1868			goto out;
1869		}
1870
1871		f = vxlan_find_mac(vxlan, n->ha, vni);
1872		if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1873			/* bridge-local neighbor */
1874			neigh_release(n);
1875			goto out;
1876		}
1877
1878		reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1879				n->ha, sha);
1880
1881		neigh_release(n);
1882
1883		if (reply == NULL)
1884			goto out;
1885
1886		skb_reset_mac_header(reply);
1887		__skb_pull(reply, skb_network_offset(reply));
1888		reply->ip_summed = CHECKSUM_UNNECESSARY;
1889		reply->pkt_type = PACKET_HOST;
1890
1891		if (netif_rx(reply) == NET_RX_DROP) {
1892			dev->stats.rx_dropped++;
1893			vxlan_vnifilter_count(vxlan, vni, NULL,
1894					      VXLAN_VNI_STATS_RX_DROPS, 0);
1895		}
1896
1897	} else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1898		union vxlan_addr ipa = {
1899			.sin.sin_addr.s_addr = tip,
1900			.sin.sin_family = AF_INET,
1901		};
1902
1903		vxlan_ip_miss(dev, &ipa);
1904	}
1905out:
1906	consume_skb(skb);
1907	return NETDEV_TX_OK;
1908}
1909
1910#if IS_ENABLED(CONFIG_IPV6)
1911static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1912	struct neighbour *n, bool isrouter)
1913{
1914	struct net_device *dev = request->dev;
1915	struct sk_buff *reply;
1916	struct nd_msg *ns, *na;
1917	struct ipv6hdr *pip6;
1918	u8 *daddr;
1919	int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1920	int ns_olen;
1921	int i, len;
1922
1923	if (dev == NULL || !pskb_may_pull(request, request->len))
1924		return NULL;
1925
1926	len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1927		sizeof(*na) + na_olen + dev->needed_tailroom;
1928	reply = alloc_skb(len, GFP_ATOMIC);
1929	if (reply == NULL)
1930		return NULL;
1931
1932	reply->protocol = htons(ETH_P_IPV6);
1933	reply->dev = dev;
1934	skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1935	skb_push(reply, sizeof(struct ethhdr));
1936	skb_reset_mac_header(reply);
1937
1938	ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
1939
1940	daddr = eth_hdr(request)->h_source;
1941	ns_olen = request->len - skb_network_offset(request) -
1942		sizeof(struct ipv6hdr) - sizeof(*ns);
1943	for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1944		if (!ns->opt[i + 1]) {
1945			kfree_skb(reply);
1946			return NULL;
1947		}
1948		if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1949			daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1950			break;
1951		}
1952	}
1953
1954	/* Ethernet header */
1955	ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1956	ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1957	eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1958	reply->protocol = htons(ETH_P_IPV6);
1959
1960	skb_pull(reply, sizeof(struct ethhdr));
1961	skb_reset_network_header(reply);
1962	skb_put(reply, sizeof(struct ipv6hdr));
1963
1964	/* IPv6 header */
1965
1966	pip6 = ipv6_hdr(reply);
1967	memset(pip6, 0, sizeof(struct ipv6hdr));
1968	pip6->version = 6;
1969	pip6->priority = ipv6_hdr(request)->priority;
1970	pip6->nexthdr = IPPROTO_ICMPV6;
1971	pip6->hop_limit = 255;
1972	pip6->daddr = ipv6_hdr(request)->saddr;
1973	pip6->saddr = *(struct in6_addr *)n->primary_key;
1974
1975	skb_pull(reply, sizeof(struct ipv6hdr));
1976	skb_reset_transport_header(reply);
1977
1978	/* Neighbor Advertisement */
1979	na = skb_put_zero(reply, sizeof(*na) + na_olen);
1980	na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1981	na->icmph.icmp6_router = isrouter;
1982	na->icmph.icmp6_override = 1;
1983	na->icmph.icmp6_solicited = 1;
1984	na->target = ns->target;
1985	ether_addr_copy(&na->opt[2], n->ha);
1986	na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1987	na->opt[1] = na_olen >> 3;
1988
1989	na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1990		&pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1991		csum_partial(na, sizeof(*na)+na_olen, 0));
1992
1993	pip6->payload_len = htons(sizeof(*na)+na_olen);
1994
1995	skb_push(reply, sizeof(struct ipv6hdr));
1996
1997	reply->ip_summed = CHECKSUM_UNNECESSARY;
1998
1999	return reply;
2000}
2001
2002static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2003{
2004	struct vxlan_dev *vxlan = netdev_priv(dev);
2005	const struct in6_addr *daddr;
2006	const struct ipv6hdr *iphdr;
2007	struct inet6_dev *in6_dev;
2008	struct neighbour *n;
2009	struct nd_msg *msg;
2010
2011	rcu_read_lock();
2012	in6_dev = __in6_dev_get(dev);
2013	if (!in6_dev)
2014		goto out;
2015
2016	iphdr = ipv6_hdr(skb);
2017	daddr = &iphdr->daddr;
2018	msg = (struct nd_msg *)(iphdr + 1);
2019
2020	if (ipv6_addr_loopback(daddr) ||
2021	    ipv6_addr_is_multicast(&msg->target))
2022		goto out;
2023
2024	n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2025
2026	if (n) {
2027		struct vxlan_fdb *f;
2028		struct sk_buff *reply;
2029
2030		if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) {
2031			neigh_release(n);
2032			goto out;
2033		}
2034
2035		f = vxlan_find_mac(vxlan, n->ha, vni);
2036		if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2037			/* bridge-local neighbor */
2038			neigh_release(n);
2039			goto out;
2040		}
2041
2042		reply = vxlan_na_create(skb, n,
2043					!!(f ? f->flags & NTF_ROUTER : 0));
2044
2045		neigh_release(n);
2046
2047		if (reply == NULL)
2048			goto out;
2049
2050		if (netif_rx(reply) == NET_RX_DROP) {
2051			dev->stats.rx_dropped++;
2052			vxlan_vnifilter_count(vxlan, vni, NULL,
2053					      VXLAN_VNI_STATS_RX_DROPS, 0);
2054		}
2055	} else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2056		union vxlan_addr ipa = {
2057			.sin6.sin6_addr = msg->target,
2058			.sin6.sin6_family = AF_INET6,
2059		};
2060
2061		vxlan_ip_miss(dev, &ipa);
2062	}
2063
2064out:
2065	rcu_read_unlock();
2066	consume_skb(skb);
2067	return NETDEV_TX_OK;
2068}
2069#endif
2070
2071static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2072{
2073	struct vxlan_dev *vxlan = netdev_priv(dev);
2074	struct neighbour *n;
2075
2076	if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2077		return false;
2078
2079	n = NULL;
2080	switch (ntohs(eth_hdr(skb)->h_proto)) {
2081	case ETH_P_IP:
2082	{
2083		struct iphdr *pip;
2084
2085		if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2086			return false;
2087		pip = ip_hdr(skb);
2088		n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2089		if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2090			union vxlan_addr ipa = {
2091				.sin.sin_addr.s_addr = pip->daddr,
2092				.sin.sin_family = AF_INET,
2093			};
2094
2095			vxlan_ip_miss(dev, &ipa);
2096			return false;
2097		}
2098
2099		break;
2100	}
2101#if IS_ENABLED(CONFIG_IPV6)
2102	case ETH_P_IPV6:
2103	{
2104		struct ipv6hdr *pip6;
2105
2106		if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2107			return false;
2108		pip6 = ipv6_hdr(skb);
2109		n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2110		if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2111			union vxlan_addr ipa = {
2112				.sin6.sin6_addr = pip6->daddr,
2113				.sin6.sin6_family = AF_INET6,
2114			};
2115
2116			vxlan_ip_miss(dev, &ipa);
2117			return false;
2118		}
2119
2120		break;
2121	}
2122#endif
2123	default:
2124		return false;
2125	}
2126
2127	if (n) {
2128		bool diff;
2129
2130		diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2131		if (diff) {
2132			memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2133				dev->addr_len);
2134			memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2135		}
2136		neigh_release(n);
2137		return diff;
2138	}
2139
2140	return false;
2141}
2142
2143static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol)
2144{
2145	struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2146
2147	gpe->np_applied = 1;
2148	gpe->next_protocol = tun_p_from_eth_p(protocol);
2149	if (!gpe->next_protocol)
2150		return -EPFNOSUPPORT;
2151	return 0;
2152}
2153
2154static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2155			   int iphdr_len, __be32 vni,
2156			   struct vxlan_metadata *md, u32 vxflags,
2157			   bool udp_sum)
2158{
2159	struct vxlanhdr *vxh;
2160	int min_headroom;
2161	int err;
2162	int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2163	__be16 inner_protocol = htons(ETH_P_TEB);
2164
2165	if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2166	    skb->ip_summed == CHECKSUM_PARTIAL) {
2167		int csum_start = skb_checksum_start_offset(skb);
2168
2169		if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2170		    !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2171		    (skb->csum_offset == offsetof(struct udphdr, check) ||
2172		     skb->csum_offset == offsetof(struct tcphdr, check)))
2173			type |= SKB_GSO_TUNNEL_REMCSUM;
2174	}
2175
2176	min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2177			+ VXLAN_HLEN + iphdr_len;
2178
2179	/* Need space for new headers (invalidates iph ptr) */
2180	err = skb_cow_head(skb, min_headroom);
2181	if (unlikely(err))
2182		return err;
2183
2184	err = iptunnel_handle_offloads(skb, type);
2185	if (err)
2186		return err;
2187
2188	vxh = __skb_push(skb, sizeof(*vxh));
2189	vxh->vx_flags = VXLAN_HF_VNI;
2190	vxh->vx_vni = vxlan_vni_field(vni);
2191
2192	if (type & SKB_GSO_TUNNEL_REMCSUM) {
2193		unsigned int start;
2194
2195		start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2196		vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2197		vxh->vx_flags |= VXLAN_HF_RCO;
2198
2199		if (!skb_is_gso(skb)) {
2200			skb->ip_summed = CHECKSUM_NONE;
2201			skb->encapsulation = 0;
2202		}
2203	}
2204
2205	if (vxflags & VXLAN_F_GBP)
2206		vxlan_build_gbp_hdr(vxh, md);
2207	if (vxflags & VXLAN_F_GPE) {
2208		err = vxlan_build_gpe_hdr(vxh, skb->protocol);
2209		if (err < 0)
2210			return err;
2211		inner_protocol = skb->protocol;
2212	}
2213
2214	skb_set_inner_protocol(skb, inner_protocol);
2215	return 0;
2216}
2217
2218/* Bypass encapsulation if the destination is local */
2219static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2220			       struct vxlan_dev *dst_vxlan, __be32 vni,
2221			       bool snoop)
2222{
2223	union vxlan_addr loopback;
2224	union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2225	struct net_device *dev;
2226	int len = skb->len;
2227
2228	skb->pkt_type = PACKET_HOST;
2229	skb->encapsulation = 0;
2230	skb->dev = dst_vxlan->dev;
2231	__skb_pull(skb, skb_network_offset(skb));
2232
2233	if (remote_ip->sa.sa_family == AF_INET) {
2234		loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2235		loopback.sa.sa_family =  AF_INET;
2236#if IS_ENABLED(CONFIG_IPV6)
2237	} else {
2238		loopback.sin6.sin6_addr = in6addr_loopback;
2239		loopback.sa.sa_family =  AF_INET6;
2240#endif
2241	}
2242
2243	rcu_read_lock();
2244	dev = skb->dev;
2245	if (unlikely(!(dev->flags & IFF_UP))) {
2246		kfree_skb(skb);
2247		goto drop;
2248	}
2249
2250	if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
2251		vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2252
2253	dev_sw_netstats_tx_add(src_vxlan->dev, 1, len);
2254	vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len);
2255
2256	if (__netif_rx(skb) == NET_RX_SUCCESS) {
2257		dev_sw_netstats_rx_add(dst_vxlan->dev, len);
2258		vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX,
2259				      len);
2260	} else {
2261drop:
2262		dev->stats.rx_dropped++;
2263		vxlan_vnifilter_count(dst_vxlan, vni, NULL,
2264				      VXLAN_VNI_STATS_RX_DROPS, 0);
2265	}
2266	rcu_read_unlock();
2267}
2268
2269static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2270				 struct vxlan_dev *vxlan,
2271				 int addr_family,
2272				 __be16 dst_port, int dst_ifindex, __be32 vni,
2273				 struct dst_entry *dst,
2274				 u32 rt_flags)
2275{
2276#if IS_ENABLED(CONFIG_IPV6)
2277	/* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2278	 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2279	 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2280	 */
2281	BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2282#endif
2283	/* Bypass encapsulation if the destination is local */
2284	if (rt_flags & RTCF_LOCAL &&
2285	    !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) &&
2286	    vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) {
2287		struct vxlan_dev *dst_vxlan;
2288
2289		dst_release(dst);
2290		dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2291					   addr_family, dst_port,
2292					   vxlan->cfg.flags);
2293		if (!dst_vxlan) {
2294			dev->stats.tx_errors++;
2295			vxlan_vnifilter_count(vxlan, vni, NULL,
2296					      VXLAN_VNI_STATS_TX_ERRORS, 0);
2297			kfree_skb(skb);
2298
2299			return -ENOENT;
2300		}
2301		vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
2302		return 1;
2303	}
2304
2305	return 0;
2306}
2307
2308void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2309		    __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc)
2310{
2311	struct dst_cache *dst_cache;
2312	struct ip_tunnel_info *info;
2313	struct ip_tunnel_key *pkey;
2314	struct ip_tunnel_key key;
2315	struct vxlan_dev *vxlan = netdev_priv(dev);
2316	const struct iphdr *old_iph = ip_hdr(skb);
2317	struct vxlan_metadata _md;
2318	struct vxlan_metadata *md = &_md;
2319	unsigned int pkt_len = skb->len;
2320	__be16 src_port = 0, dst_port;
2321	struct dst_entry *ndst = NULL;
2322	int addr_family;
2323	__u8 tos, ttl;
2324	int ifindex;
2325	int err;
2326	u32 flags = vxlan->cfg.flags;
2327	bool use_cache;
2328	bool udp_sum = false;
2329	bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2330	__be32 vni = 0;
2331
2332	info = skb_tunnel_info(skb);
2333	use_cache = ip_tunnel_dst_cache_usable(skb, info);
2334
2335	if (rdst) {
2336		memset(&key, 0, sizeof(key));
2337		pkey = &key;
2338
2339		if (vxlan_addr_any(&rdst->remote_ip)) {
2340			if (did_rsc) {
2341				/* short-circuited back to local bridge */
2342				vxlan_encap_bypass(skb, vxlan, vxlan,
2343						   default_vni, true);
2344				return;
2345			}
2346			goto drop;
2347		}
2348
2349		addr_family = vxlan->cfg.saddr.sa.sa_family;
2350		dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2351		vni = (rdst->remote_vni) ? : default_vni;
2352		ifindex = rdst->remote_ifindex;
2353
2354		if (addr_family == AF_INET) {
2355			key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr;
2356			key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr;
2357		} else {
2358			key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr;
2359			key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr;
2360		}
2361
2362		dst_cache = &rdst->dst_cache;
2363		md->gbp = skb->mark;
2364		if (flags & VXLAN_F_TTL_INHERIT) {
2365			ttl = ip_tunnel_get_ttl(old_iph, skb);
2366		} else {
2367			ttl = vxlan->cfg.ttl;
2368			if (!ttl && vxlan_addr_multicast(&rdst->remote_ip))
2369				ttl = 1;
2370		}
2371		tos = vxlan->cfg.tos;
2372		if (tos == 1)
2373			tos = ip_tunnel_get_dsfield(old_iph, skb);
2374		if (tos && !info)
2375			use_cache = false;
2376
2377		if (addr_family == AF_INET)
2378			udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2379		else
2380			udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2381#if IS_ENABLED(CONFIG_IPV6)
2382		switch (vxlan->cfg.label_policy) {
2383		case VXLAN_LABEL_FIXED:
2384			key.label = vxlan->cfg.label;
2385			break;
2386		case VXLAN_LABEL_INHERIT:
2387			key.label = ip_tunnel_get_flowlabel(old_iph, skb);
2388			break;
2389		default:
2390			DEBUG_NET_WARN_ON_ONCE(1);
2391			goto drop;
2392		}
2393#endif
2394	} else {
2395		if (!info) {
2396			WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2397				  dev->name);
2398			goto drop;
2399		}
2400		pkey = &info->key;
2401		addr_family = ip_tunnel_info_af(info);
2402		dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2403		vni = tunnel_id_to_key32(info->key.tun_id);
2404		ifindex = 0;
2405		dst_cache = &info->dst_cache;
2406		if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2407			if (info->options_len < sizeof(*md))
2408				goto drop;
2409			md = ip_tunnel_info_opts(info);
2410		}
2411		ttl = info->key.ttl;
2412		tos = info->key.tos;
2413		udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2414	}
2415	src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2416				     vxlan->cfg.port_max, true);
2417
2418	rcu_read_lock();
2419	if (addr_family == AF_INET) {
2420		struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2421		struct rtable *rt;
2422		__be16 df = 0;
2423		__be32 saddr;
2424
2425		if (!ifindex)
2426			ifindex = sock4->sock->sk->sk_bound_dev_if;
2427
2428		rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex,
2429					   &saddr, pkey, src_port, dst_port,
2430					   tos, use_cache ? dst_cache : NULL);
2431		if (IS_ERR(rt)) {
2432			err = PTR_ERR(rt);
2433			goto tx_error;
2434		}
2435
2436		if (!info) {
2437			/* Bypass encapsulation if the destination is local */
2438			err = encap_bypass_if_local(skb, dev, vxlan, AF_INET,
2439						    dst_port, ifindex, vni,
2440						    &rt->dst, rt->rt_flags);
2441			if (err)
2442				goto out_unlock;
2443
2444			if (vxlan->cfg.df == VXLAN_DF_SET) {
2445				df = htons(IP_DF);
2446			} else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2447				struct ethhdr *eth = eth_hdr(skb);
2448
2449				if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2450				    (ntohs(eth->h_proto) == ETH_P_IP &&
2451				     old_iph->frag_off & htons(IP_DF)))
2452					df = htons(IP_DF);
2453			}
2454		} else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2455			df = htons(IP_DF);
2456		}
2457
2458		ndst = &rt->dst;
2459		err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE),
2460					    netif_is_any_bridge_port(dev));
2461		if (err < 0) {
2462			goto tx_error;
2463		} else if (err) {
2464			if (info) {
2465				struct ip_tunnel_info *unclone;
2466
2467				unclone = skb_tunnel_info_unclone(skb);
2468				if (unlikely(!unclone))
2469					goto tx_error;
2470
2471				unclone->key.u.ipv4.src = pkey->u.ipv4.dst;
2472				unclone->key.u.ipv4.dst = saddr;
2473			}
2474			vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2475			dst_release(ndst);
2476			goto out_unlock;
2477		}
2478
2479		tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2480		ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2481		err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2482				      vni, md, flags, udp_sum);
2483		if (err < 0)
2484			goto tx_error;
2485
2486		udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, saddr,
2487				    pkey->u.ipv4.dst, tos, ttl, df,
2488				    src_port, dst_port, xnet, !udp_sum);
2489#if IS_ENABLED(CONFIG_IPV6)
2490	} else {
2491		struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2492		struct in6_addr saddr;
2493
2494		if (!ifindex)
2495			ifindex = sock6->sock->sk->sk_bound_dev_if;
2496
2497		ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock,
2498					      ifindex, &saddr, pkey,
2499					      src_port, dst_port, tos,
2500					      use_cache ? dst_cache : NULL);
2501		if (IS_ERR(ndst)) {
2502			err = PTR_ERR(ndst);
2503			ndst = NULL;
2504			goto tx_error;
2505		}
2506
2507		if (!info) {
2508			u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2509
2510			err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6,
2511						    dst_port, ifindex, vni,
2512						    ndst, rt6i_flags);
2513			if (err)
2514				goto out_unlock;
2515		}
2516
2517		err = skb_tunnel_check_pmtu(skb, ndst,
2518					    vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6),
2519					    netif_is_any_bridge_port(dev));
2520		if (err < 0) {
2521			goto tx_error;
2522		} else if (err) {
2523			if (info) {
2524				struct ip_tunnel_info *unclone;
2525
2526				unclone = skb_tunnel_info_unclone(skb);
2527				if (unlikely(!unclone))
2528					goto tx_error;
2529
2530				unclone->key.u.ipv6.src = pkey->u.ipv6.dst;
2531				unclone->key.u.ipv6.dst = saddr;
2532			}
2533
2534			vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2535			dst_release(ndst);
2536			goto out_unlock;
2537		}
2538
2539		tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2540		ttl = ttl ? : ip6_dst_hoplimit(ndst);
2541		skb_scrub_packet(skb, xnet);
2542		err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2543				      vni, md, flags, udp_sum);
2544		if (err < 0)
2545			goto tx_error;
2546
2547		udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2548				     &saddr, &pkey->u.ipv6.dst, tos, ttl,
2549				     pkey->label, src_port, dst_port, !udp_sum);
2550#endif
2551	}
2552	vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len);
2553out_unlock:
2554	rcu_read_unlock();
2555	return;
2556
2557drop:
2558	dev->stats.tx_dropped++;
2559	vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0);
2560	dev_kfree_skb(skb);
2561	return;
2562
2563tx_error:
2564	rcu_read_unlock();
2565	if (err == -ELOOP)
2566		dev->stats.collisions++;
2567	else if (err == -ENETUNREACH)
2568		dev->stats.tx_carrier_errors++;
2569	dst_release(ndst);
2570	dev->stats.tx_errors++;
2571	vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0);
2572	kfree_skb(skb);
2573}
2574
2575static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2576			  struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2577{
2578	struct vxlan_rdst nh_rdst;
2579	struct nexthop *nh;
2580	bool do_xmit;
2581	u32 hash;
2582
2583	memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2584	hash = skb_get_hash(skb);
2585
2586	rcu_read_lock();
2587	nh = rcu_dereference(f->nh);
2588	if (!nh) {
2589		rcu_read_unlock();
2590		goto drop;
2591	}
2592	do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2593	rcu_read_unlock();
2594
2595	if (likely(do_xmit))
2596		vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2597	else
2598		goto drop;
2599
2600	return;
2601
2602drop:
2603	dev->stats.tx_dropped++;
2604	vxlan_vnifilter_count(netdev_priv(dev), vni, NULL,
2605			      VXLAN_VNI_STATS_TX_DROPS, 0);
2606	dev_kfree_skb(skb);
2607}
2608
2609static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev,
2610				   u32 nhid, __be32 vni)
2611{
2612	struct vxlan_dev *vxlan = netdev_priv(dev);
2613	struct vxlan_rdst nh_rdst;
2614	struct nexthop *nh;
2615	bool do_xmit;
2616	u32 hash;
2617
2618	memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2619	hash = skb_get_hash(skb);
2620
2621	rcu_read_lock();
2622	nh = nexthop_find_by_id(dev_net(dev), nhid);
2623	if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) {
2624		rcu_read_unlock();
2625		goto drop;
2626	}
2627	do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2628	rcu_read_unlock();
2629
2630	if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family)
2631		goto drop;
2632
2633	if (likely(do_xmit))
2634		vxlan_xmit_one(skb, dev, vni, &nh_rdst, false);
2635	else
2636		goto drop;
2637
2638	return NETDEV_TX_OK;
2639
2640drop:
2641	dev->stats.tx_dropped++;
2642	vxlan_vnifilter_count(netdev_priv(dev), vni, NULL,
2643			      VXLAN_VNI_STATS_TX_DROPS, 0);
2644	dev_kfree_skb(skb);
2645	return NETDEV_TX_OK;
2646}
2647
2648/* Transmit local packets over Vxlan
2649 *
2650 * Outer IP header inherits ECN and DF from inner header.
2651 * Outer UDP destination is the VXLAN assigned port.
2652 *           source port is based on hash of flow
2653 */
2654static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2655{
2656	struct vxlan_dev *vxlan = netdev_priv(dev);
2657	struct vxlan_rdst *rdst, *fdst = NULL;
2658	const struct ip_tunnel_info *info;
2659	bool did_rsc = false;
2660	struct vxlan_fdb *f;
2661	struct ethhdr *eth;
2662	__be32 vni = 0;
2663	u32 nhid = 0;
2664
2665	info = skb_tunnel_info(skb);
2666
2667	skb_reset_mac_header(skb);
2668
2669	if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2670		if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2671		    info->mode & IP_TUNNEL_INFO_TX) {
2672			vni = tunnel_id_to_key32(info->key.tun_id);
2673			nhid = info->key.nhid;
2674		} else {
2675			if (info && info->mode & IP_TUNNEL_INFO_TX)
2676				vxlan_xmit_one(skb, dev, vni, NULL, false);
2677			else
2678				kfree_skb(skb);
2679			return NETDEV_TX_OK;
2680		}
2681	}
2682
2683	if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2684		eth = eth_hdr(skb);
2685		if (ntohs(eth->h_proto) == ETH_P_ARP)
2686			return arp_reduce(dev, skb, vni);
2687#if IS_ENABLED(CONFIG_IPV6)
2688		else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2689			 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2690					    sizeof(struct nd_msg)) &&
2691			 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2692			struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2693
2694			if (m->icmph.icmp6_code == 0 &&
2695			    m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2696				return neigh_reduce(dev, skb, vni);
2697		}
2698#endif
2699	}
2700
2701	if (nhid)
2702		return vxlan_xmit_nhid(skb, dev, nhid, vni);
2703
2704	if (vxlan->cfg.flags & VXLAN_F_MDB) {
2705		struct vxlan_mdb_entry *mdb_entry;
2706
2707		rcu_read_lock();
2708		mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni);
2709		if (mdb_entry) {
2710			netdev_tx_t ret;
2711
2712			ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb);
2713			rcu_read_unlock();
2714			return ret;
2715		}
2716		rcu_read_unlock();
2717	}
2718
2719	eth = eth_hdr(skb);
2720	f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2721	did_rsc = false;
2722
2723	if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2724	    (ntohs(eth->h_proto) == ETH_P_IP ||
2725	     ntohs(eth->h_proto) == ETH_P_IPV6)) {
2726		did_rsc = route_shortcircuit(dev, skb);
2727		if (did_rsc)
2728			f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2729	}
2730
2731	if (f == NULL) {
2732		f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2733		if (f == NULL) {
2734			if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2735			    !is_multicast_ether_addr(eth->h_dest))
2736				vxlan_fdb_miss(vxlan, eth->h_dest);
2737
2738			dev->stats.tx_dropped++;
2739			vxlan_vnifilter_count(vxlan, vni, NULL,
2740					      VXLAN_VNI_STATS_TX_DROPS, 0);
2741			kfree_skb(skb);
2742			return NETDEV_TX_OK;
2743		}
2744	}
2745
2746	if (rcu_access_pointer(f->nh)) {
2747		vxlan_xmit_nh(skb, dev, f,
2748			      (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2749	} else {
2750		list_for_each_entry_rcu(rdst, &f->remotes, list) {
2751			struct sk_buff *skb1;
2752
2753			if (!fdst) {
2754				fdst = rdst;
2755				continue;
2756			}
2757			skb1 = skb_clone(skb, GFP_ATOMIC);
2758			if (skb1)
2759				vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2760		}
2761		if (fdst)
2762			vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2763		else
2764			kfree_skb(skb);
2765	}
2766
2767	return NETDEV_TX_OK;
2768}
2769
2770/* Walk the forwarding table and purge stale entries */
2771static void vxlan_cleanup(struct timer_list *t)
2772{
2773	struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2774	unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2775	unsigned int h;
2776
2777	if (!netif_running(vxlan->dev))
2778		return;
2779
2780	for (h = 0; h < FDB_HASH_SIZE; ++h) {
2781		struct hlist_node *p, *n;
2782
2783		spin_lock(&vxlan->hash_lock[h]);
2784		hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2785			struct vxlan_fdb *f
2786				= container_of(p, struct vxlan_fdb, hlist);
2787			unsigned long timeout;
2788
2789			if (f->state & (NUD_PERMANENT | NUD_NOARP))
2790				continue;
2791
2792			if (f->flags & NTF_EXT_LEARNED)
2793				continue;
2794
2795			timeout = f->used + vxlan->cfg.age_interval * HZ;
2796			if (time_before_eq(timeout, jiffies)) {
2797				netdev_dbg(vxlan->dev,
2798					   "garbage collect %pM\n",
2799					   f->eth_addr);
2800				f->state = NUD_STALE;
2801				vxlan_fdb_destroy(vxlan, f, true, true);
2802			} else if (time_before(timeout, next_timer))
2803				next_timer = timeout;
2804		}
2805		spin_unlock(&vxlan->hash_lock[h]);
2806	}
2807
2808	mod_timer(&vxlan->age_timer, next_timer);
2809}
2810
2811static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2812{
2813	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2814
2815	spin_lock(&vn->sock_lock);
2816	hlist_del_init_rcu(&vxlan->hlist4.hlist);
2817#if IS_ENABLED(CONFIG_IPV6)
2818	hlist_del_init_rcu(&vxlan->hlist6.hlist);
2819#endif
2820	spin_unlock(&vn->sock_lock);
2821}
2822
2823static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2824			     struct vxlan_dev_node *node)
2825{
2826	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2827	__be32 vni = vxlan->default_dst.remote_vni;
2828
2829	node->vxlan = vxlan;
2830	spin_lock(&vn->sock_lock);
2831	hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2832	spin_unlock(&vn->sock_lock);
2833}
2834
2835/* Setup stats when device is created */
2836static int vxlan_init(struct net_device *dev)
2837{
2838	struct vxlan_dev *vxlan = netdev_priv(dev);
2839	int err;
2840
2841	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2842		vxlan_vnigroup_init(vxlan);
2843
2844	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2845	if (!dev->tstats) {
2846		err = -ENOMEM;
2847		goto err_vnigroup_uninit;
2848	}
2849
2850	err = gro_cells_init(&vxlan->gro_cells, dev);
2851	if (err)
2852		goto err_free_percpu;
2853
2854	err = vxlan_mdb_init(vxlan);
2855	if (err)
2856		goto err_gro_cells_destroy;
2857
2858	return 0;
2859
2860err_gro_cells_destroy:
2861	gro_cells_destroy(&vxlan->gro_cells);
2862err_free_percpu:
2863	free_percpu(dev->tstats);
2864err_vnigroup_uninit:
2865	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2866		vxlan_vnigroup_uninit(vxlan);
2867	return err;
2868}
2869
2870static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
2871{
2872	struct vxlan_fdb *f;
2873	u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
2874
2875	spin_lock_bh(&vxlan->hash_lock[hash_index]);
2876	f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
2877	if (f)
2878		vxlan_fdb_destroy(vxlan, f, true, true);
2879	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
2880}
2881
2882static void vxlan_uninit(struct net_device *dev)
2883{
2884	struct vxlan_dev *vxlan = netdev_priv(dev);
2885
2886	vxlan_mdb_fini(vxlan);
2887
2888	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2889		vxlan_vnigroup_uninit(vxlan);
2890
2891	gro_cells_destroy(&vxlan->gro_cells);
2892
2893	vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
2894
2895	free_percpu(dev->tstats);
2896}
2897
2898/* Start ageing timer and join group when device is brought up */
2899static int vxlan_open(struct net_device *dev)
2900{
2901	struct vxlan_dev *vxlan = netdev_priv(dev);
2902	int ret;
2903
2904	ret = vxlan_sock_add(vxlan);
2905	if (ret < 0)
2906		return ret;
2907
2908	ret = vxlan_multicast_join(vxlan);
2909	if (ret) {
2910		vxlan_sock_release(vxlan);
2911		return ret;
2912	}
2913
2914	if (vxlan->cfg.age_interval)
2915		mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2916
2917	return ret;
2918}
2919
2920struct vxlan_fdb_flush_desc {
2921	bool				ignore_default_entry;
2922	unsigned long                   state;
2923	unsigned long			state_mask;
2924	unsigned long                   flags;
2925	unsigned long			flags_mask;
2926	__be32				src_vni;
2927	u32				nhid;
2928	__be32				vni;
2929	__be16				port;
2930	union vxlan_addr		dst_ip;
2931};
2932
2933static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f,
2934				       const struct vxlan_dev *vxlan)
2935{
2936	return is_zero_ether_addr(f->eth_addr) && f->vni == vxlan->cfg.vni;
2937}
2938
2939static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid)
2940{
2941	struct nexthop *nh = rtnl_dereference(f->nh);
2942
2943	return nh && nh->id == nhid;
2944}
2945
2946static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f,
2947				    const struct vxlan_dev *vxlan,
2948				    const struct vxlan_fdb_flush_desc *desc)
2949{
2950	if (desc->state_mask && (f->state & desc->state_mask) != desc->state)
2951		return false;
2952
2953	if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags)
2954		return false;
2955
2956	if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan))
2957		return false;
2958
2959	if (desc->src_vni && f->vni != desc->src_vni)
2960		return false;
2961
2962	if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid))
2963		return false;
2964
2965	return true;
2966}
2967
2968static bool
2969vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc)
2970{
2971	return desc->vni || desc->port || desc->dst_ip.sa.sa_family;
2972}
2973
2974static bool
2975vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc,
2976			       const struct vxlan_rdst *rd)
2977{
2978	if (desc->vni && rd->remote_vni != desc->vni)
2979		return false;
2980
2981	if (desc->port && rd->remote_port != desc->port)
2982		return false;
2983
2984	if (desc->dst_ip.sa.sa_family &&
2985	    !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip))
2986		return false;
2987
2988	return true;
2989}
2990
2991static void
2992vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan,
2993			      const struct vxlan_fdb_flush_desc *desc,
2994			      bool *p_destroy_fdb)
2995{
2996	bool remotes_flushed = false;
2997	struct vxlan_rdst *rd, *tmp;
2998
2999	list_for_each_entry_safe(rd, tmp, &f->remotes, list) {
3000		if (!vxlan_fdb_flush_remote_matches(desc, rd))
3001			continue;
3002
3003		vxlan_fdb_dst_destroy(vxlan, f, rd, true);
3004		remotes_flushed = true;
3005	}
3006
3007	*p_destroy_fdb = remotes_flushed && list_empty(&f->remotes);
3008}
3009
3010/* Purge the forwarding table */
3011static void vxlan_flush(struct vxlan_dev *vxlan,
3012			const struct vxlan_fdb_flush_desc *desc)
3013{
3014	bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc);
3015	unsigned int h;
3016
3017	for (h = 0; h < FDB_HASH_SIZE; ++h) {
3018		struct hlist_node *p, *n;
3019
3020		spin_lock_bh(&vxlan->hash_lock[h]);
3021		hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3022			struct vxlan_fdb *f
3023				= container_of(p, struct vxlan_fdb, hlist);
3024
3025			if (!vxlan_fdb_flush_matches(f, vxlan, desc))
3026				continue;
3027
3028			if (match_remotes) {
3029				bool destroy_fdb = false;
3030
3031				vxlan_fdb_flush_match_remotes(f, vxlan, desc,
3032							      &destroy_fdb);
3033
3034				if (!destroy_fdb)
3035					continue;
3036			}
3037
3038			vxlan_fdb_destroy(vxlan, f, true, true);
3039		}
3040		spin_unlock_bh(&vxlan->hash_lock[h]);
3041	}
3042}
3043
3044static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = {
3045	[NDA_SRC_VNI]   = { .type = NLA_U32 },
3046	[NDA_NH_ID]	= { .type = NLA_U32 },
3047	[NDA_VNI]	= { .type = NLA_U32 },
3048	[NDA_PORT]	= { .type = NLA_U16 },
3049	[NDA_DST]	= NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr),
3050					   sizeof(struct in6_addr)),
3051	[NDA_NDM_STATE_MASK]	= { .type = NLA_U16 },
3052	[NDA_NDM_FLAGS_MASK]	= { .type = NLA_U8 },
3053};
3054
3055#define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF)
3056#define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP)
3057#define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \
3058					   NTF_ROUTER)
3059
3060static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev,
3061				 struct netlink_ext_ack *extack)
3062{
3063	struct vxlan_dev *vxlan = netdev_priv(dev);
3064	struct vxlan_fdb_flush_desc desc = {};
3065	struct ndmsg *ndm = nlmsg_data(nlh);
3066	struct nlattr *tb[NDA_MAX + 1];
3067	u8 ndm_flags;
3068	int err;
3069
3070	ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS;
3071
3072	err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy,
3073			  extack);
3074	if (err)
3075		return err;
3076
3077	if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) {
3078		NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set");
3079		return -EINVAL;
3080	}
3081	if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) {
3082		NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set");
3083		return -EINVAL;
3084	}
3085
3086	desc.state = ndm->ndm_state;
3087	desc.flags = ndm_flags;
3088
3089	if (tb[NDA_NDM_STATE_MASK])
3090		desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]);
3091
3092	if (tb[NDA_NDM_FLAGS_MASK])
3093		desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]);
3094
3095	if (tb[NDA_SRC_VNI])
3096		desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
3097
3098	if (tb[NDA_NH_ID])
3099		desc.nhid = nla_get_u32(tb[NDA_NH_ID]);
3100
3101	if (tb[NDA_VNI])
3102		desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
3103
3104	if (tb[NDA_PORT])
3105		desc.port = nla_get_be16(tb[NDA_PORT]);
3106
3107	if (tb[NDA_DST]) {
3108		union vxlan_addr ip;
3109
3110		err = vxlan_nla_get_addr(&ip, tb[NDA_DST]);
3111		if (err) {
3112			NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST],
3113					    "Unsupported address family");
3114			return err;
3115		}
3116		desc.dst_ip = ip;
3117	}
3118
3119	vxlan_flush(vxlan, &desc);
3120
3121	return 0;
3122}
3123
3124/* Cleanup timer and forwarding table on shutdown */
3125static int vxlan_stop(struct net_device *dev)
3126{
3127	struct vxlan_dev *vxlan = netdev_priv(dev);
3128	struct vxlan_fdb_flush_desc desc = {
3129		/* Default entry is deleted at vxlan_uninit. */
3130		.ignore_default_entry = true,
3131		.state = 0,
3132		.state_mask = NUD_PERMANENT | NUD_NOARP,
3133	};
3134
3135	vxlan_multicast_leave(vxlan);
3136
3137	del_timer_sync(&vxlan->age_timer);
3138
3139	vxlan_flush(vxlan, &desc);
3140	vxlan_sock_release(vxlan);
3141
3142	return 0;
3143}
3144
3145/* Stub, nothing needs to be done. */
3146static void vxlan_set_multicast_list(struct net_device *dev)
3147{
3148}
3149
3150static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3151{
3152	struct vxlan_dev *vxlan = netdev_priv(dev);
3153	struct vxlan_rdst *dst = &vxlan->default_dst;
3154	struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3155							 dst->remote_ifindex);
3156
3157	/* This check is different than dev->max_mtu, because it looks at
3158	 * the lowerdev->mtu, rather than the static dev->max_mtu
3159	 */
3160	if (lowerdev) {
3161		int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags);
3162		if (new_mtu > max_mtu)
3163			return -EINVAL;
3164	}
3165
3166	dev->mtu = new_mtu;
3167	return 0;
3168}
3169
3170static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3171{
3172	struct vxlan_dev *vxlan = netdev_priv(dev);
3173	struct ip_tunnel_info *info = skb_tunnel_info(skb);
3174	__be16 sport, dport;
3175
3176	sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3177				  vxlan->cfg.port_max, true);
3178	dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3179
3180	if (ip_tunnel_info_af(info) == AF_INET) {
3181		struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3182		struct rtable *rt;
3183
3184		if (!sock4)
3185			return -EIO;
3186
3187		rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0,
3188					   &info->key.u.ipv4.src,
3189					   &info->key,
3190					   sport, dport, info->key.tos,
3191					   &info->dst_cache);
3192		if (IS_ERR(rt))
3193			return PTR_ERR(rt);
3194		ip_rt_put(rt);
3195	} else {
3196#if IS_ENABLED(CONFIG_IPV6)
3197		struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3198		struct dst_entry *ndst;
3199
3200		if (!sock6)
3201			return -EIO;
3202
3203		ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock,
3204					      0, &info->key.u.ipv6.src,
3205					      &info->key,
3206					      sport, dport, info->key.tos,
3207					      &info->dst_cache);
3208		if (IS_ERR(ndst))
3209			return PTR_ERR(ndst);
3210		dst_release(ndst);
3211#else /* !CONFIG_IPV6 */
3212		return -EPFNOSUPPORT;
3213#endif
3214	}
3215	info->key.tp_src = sport;
3216	info->key.tp_dst = dport;
3217	return 0;
3218}
3219
3220static const struct net_device_ops vxlan_netdev_ether_ops = {
3221	.ndo_init		= vxlan_init,
3222	.ndo_uninit		= vxlan_uninit,
3223	.ndo_open		= vxlan_open,
3224	.ndo_stop		= vxlan_stop,
3225	.ndo_start_xmit		= vxlan_xmit,
3226	.ndo_get_stats64	= dev_get_tstats64,
3227	.ndo_set_rx_mode	= vxlan_set_multicast_list,
3228	.ndo_change_mtu		= vxlan_change_mtu,
3229	.ndo_validate_addr	= eth_validate_addr,
3230	.ndo_set_mac_address	= eth_mac_addr,
3231	.ndo_fdb_add		= vxlan_fdb_add,
3232	.ndo_fdb_del		= vxlan_fdb_delete,
3233	.ndo_fdb_del_bulk	= vxlan_fdb_delete_bulk,
3234	.ndo_fdb_dump		= vxlan_fdb_dump,
3235	.ndo_fdb_get		= vxlan_fdb_get,
3236	.ndo_mdb_add		= vxlan_mdb_add,
3237	.ndo_mdb_del		= vxlan_mdb_del,
3238	.ndo_mdb_del_bulk	= vxlan_mdb_del_bulk,
3239	.ndo_mdb_dump		= vxlan_mdb_dump,
3240	.ndo_mdb_get		= vxlan_mdb_get,
3241	.ndo_fill_metadata_dst	= vxlan_fill_metadata_dst,
3242};
3243
3244static const struct net_device_ops vxlan_netdev_raw_ops = {
3245	.ndo_init		= vxlan_init,
3246	.ndo_uninit		= vxlan_uninit,
3247	.ndo_open		= vxlan_open,
3248	.ndo_stop		= vxlan_stop,
3249	.ndo_start_xmit		= vxlan_xmit,
3250	.ndo_get_stats64	= dev_get_tstats64,
3251	.ndo_change_mtu		= vxlan_change_mtu,
3252	.ndo_fill_metadata_dst	= vxlan_fill_metadata_dst,
3253};
3254
3255/* Info for udev, that this is a virtual tunnel endpoint */
3256static struct device_type vxlan_type = {
3257	.name = "vxlan",
3258};
3259
3260/* Calls the ndo_udp_tunnel_add of the caller in order to
3261 * supply the listening VXLAN udp ports. Callers are expected
3262 * to implement the ndo_udp_tunnel_add.
3263 */
3264static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3265{
3266	struct vxlan_sock *vs;
3267	struct net *net = dev_net(dev);
3268	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3269	unsigned int i;
3270
3271	spin_lock(&vn->sock_lock);
3272	for (i = 0; i < PORT_HASH_SIZE; ++i) {
3273		hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3274			unsigned short type;
3275
3276			if (vs->flags & VXLAN_F_GPE)
3277				type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3278			else
3279				type = UDP_TUNNEL_TYPE_VXLAN;
3280
3281			if (push)
3282				udp_tunnel_push_rx_port(dev, vs->sock, type);
3283			else
3284				udp_tunnel_drop_rx_port(dev, vs->sock, type);
3285		}
3286	}
3287	spin_unlock(&vn->sock_lock);
3288}
3289
3290/* Initialize the device structure. */
3291static void vxlan_setup(struct net_device *dev)
3292{
3293	struct vxlan_dev *vxlan = netdev_priv(dev);
3294	unsigned int h;
3295
3296	eth_hw_addr_random(dev);
3297	ether_setup(dev);
3298
3299	dev->needs_free_netdev = true;
3300	SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3301
3302	dev->features	|= NETIF_F_LLTX;
3303	dev->features	|= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3304	dev->features   |= NETIF_F_RXCSUM;
3305	dev->features   |= NETIF_F_GSO_SOFTWARE;
3306
3307	dev->vlan_features = dev->features;
3308	dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3309	dev->hw_features |= NETIF_F_RXCSUM;
3310	dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3311	netif_keep_dst(dev);
3312	dev->priv_flags |= IFF_NO_QUEUE | IFF_CHANGE_PROTO_DOWN;
3313
3314	/* MTU range: 68 - 65535 */
3315	dev->min_mtu = ETH_MIN_MTU;
3316	dev->max_mtu = ETH_MAX_MTU;
3317
3318	INIT_LIST_HEAD(&vxlan->next);
3319
3320	timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3321
3322	vxlan->dev = dev;
3323
3324	for (h = 0; h < FDB_HASH_SIZE; ++h) {
3325		spin_lock_init(&vxlan->hash_lock[h]);
3326		INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3327	}
3328}
3329
3330static void vxlan_ether_setup(struct net_device *dev)
3331{
3332	dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3333	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3334	dev->netdev_ops = &vxlan_netdev_ether_ops;
3335}
3336
3337static void vxlan_raw_setup(struct net_device *dev)
3338{
3339	dev->header_ops = NULL;
3340	dev->type = ARPHRD_NONE;
3341	dev->hard_header_len = 0;
3342	dev->addr_len = 0;
3343	dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3344	dev->netdev_ops = &vxlan_netdev_raw_ops;
3345}
3346
3347static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3348	[IFLA_VXLAN_UNSPEC]     = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS },
3349	[IFLA_VXLAN_ID]		= { .type = NLA_U32 },
3350	[IFLA_VXLAN_GROUP]	= { .len = sizeof_field(struct iphdr, daddr) },
3351	[IFLA_VXLAN_GROUP6]	= { .len = sizeof(struct in6_addr) },
3352	[IFLA_VXLAN_LINK]	= { .type = NLA_U32 },
3353	[IFLA_VXLAN_LOCAL]	= { .len = sizeof_field(struct iphdr, saddr) },
3354	[IFLA_VXLAN_LOCAL6]	= { .len = sizeof(struct in6_addr) },
3355	[IFLA_VXLAN_TOS]	= { .type = NLA_U8 },
3356	[IFLA_VXLAN_TTL]	= { .type = NLA_U8 },
3357	[IFLA_VXLAN_LABEL]	= { .type = NLA_U32 },
3358	[IFLA_VXLAN_LEARNING]	= { .type = NLA_U8 },
3359	[IFLA_VXLAN_AGEING]	= { .type = NLA_U32 },
3360	[IFLA_VXLAN_LIMIT]	= { .type = NLA_U32 },
3361	[IFLA_VXLAN_PORT_RANGE] = { .len  = sizeof(struct ifla_vxlan_port_range) },
3362	[IFLA_VXLAN_PROXY]	= { .type = NLA_U8 },
3363	[IFLA_VXLAN_RSC]	= { .type = NLA_U8 },
3364	[IFLA_VXLAN_L2MISS]	= { .type = NLA_U8 },
3365	[IFLA_VXLAN_L3MISS]	= { .type = NLA_U8 },
3366	[IFLA_VXLAN_COLLECT_METADATA]	= { .type = NLA_U8 },
3367	[IFLA_VXLAN_PORT]	= { .type = NLA_U16 },
3368	[IFLA_VXLAN_UDP_CSUM]	= { .type = NLA_U8 },
3369	[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]	= { .type = NLA_U8 },
3370	[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]	= { .type = NLA_U8 },
3371	[IFLA_VXLAN_REMCSUM_TX]	= { .type = NLA_U8 },
3372	[IFLA_VXLAN_REMCSUM_RX]	= { .type = NLA_U8 },
3373	[IFLA_VXLAN_GBP]	= { .type = NLA_FLAG, },
3374	[IFLA_VXLAN_GPE]	= { .type = NLA_FLAG, },
3375	[IFLA_VXLAN_REMCSUM_NOPARTIAL]	= { .type = NLA_FLAG },
3376	[IFLA_VXLAN_TTL_INHERIT]	= { .type = NLA_FLAG },
3377	[IFLA_VXLAN_DF]		= { .type = NLA_U8 },
3378	[IFLA_VXLAN_VNIFILTER]	= { .type = NLA_U8 },
3379	[IFLA_VXLAN_LOCALBYPASS]	= NLA_POLICY_MAX(NLA_U8, 1),
3380	[IFLA_VXLAN_LABEL_POLICY]       = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX),
3381};
3382
3383static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3384			  struct netlink_ext_ack *extack)
3385{
3386	if (tb[IFLA_ADDRESS]) {
3387		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3388			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3389					    "Provided link layer address is not Ethernet");
3390			return -EINVAL;
3391		}
3392
3393		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3394			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3395					    "Provided Ethernet address is not unicast");
3396			return -EADDRNOTAVAIL;
3397		}
3398	}
3399
3400	if (tb[IFLA_MTU]) {
3401		u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3402
3403		if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3404			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3405					    "MTU must be between 68 and 65535");
3406			return -EINVAL;
3407		}
3408	}
3409
3410	if (!data) {
3411		NL_SET_ERR_MSG(extack,
3412			       "Required attributes not provided to perform the operation");
3413		return -EINVAL;
3414	}
3415
3416	if (data[IFLA_VXLAN_ID]) {
3417		u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3418
3419		if (id >= VXLAN_N_VID) {
3420			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3421					    "VXLAN ID must be lower than 16777216");
3422			return -ERANGE;
3423		}
3424	}
3425
3426	if (data[IFLA_VXLAN_PORT_RANGE]) {
3427		const struct ifla_vxlan_port_range *p
3428			= nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3429
3430		if (ntohs(p->high) < ntohs(p->low)) {
3431			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3432					    "Invalid source port range");
3433			return -EINVAL;
3434		}
3435	}
3436
3437	if (data[IFLA_VXLAN_DF]) {
3438		enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3439
3440		if (df < 0 || df > VXLAN_DF_MAX) {
3441			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3442					    "Invalid DF attribute");
3443			return -EINVAL;
3444		}
3445	}
3446
3447	return 0;
3448}
3449
3450static void vxlan_get_drvinfo(struct net_device *netdev,
3451			      struct ethtool_drvinfo *drvinfo)
3452{
3453	strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3454	strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3455}
3456
3457static int vxlan_get_link_ksettings(struct net_device *dev,
3458				    struct ethtool_link_ksettings *cmd)
3459{
3460	struct vxlan_dev *vxlan = netdev_priv(dev);
3461	struct vxlan_rdst *dst = &vxlan->default_dst;
3462	struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3463							 dst->remote_ifindex);
3464
3465	if (!lowerdev) {
3466		cmd->base.duplex = DUPLEX_UNKNOWN;
3467		cmd->base.port = PORT_OTHER;
3468		cmd->base.speed = SPEED_UNKNOWN;
3469
3470		return 0;
3471	}
3472
3473	return __ethtool_get_link_ksettings(lowerdev, cmd);
3474}
3475
3476static const struct ethtool_ops vxlan_ethtool_ops = {
3477	.get_drvinfo		= vxlan_get_drvinfo,
3478	.get_link		= ethtool_op_get_link,
3479	.get_link_ksettings	= vxlan_get_link_ksettings,
3480};
3481
3482static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3483					__be16 port, u32 flags, int ifindex)
3484{
3485	struct socket *sock;
3486	struct udp_port_cfg udp_conf;
3487	int err;
3488
3489	memset(&udp_conf, 0, sizeof(udp_conf));
3490
3491	if (ipv6) {
3492		udp_conf.family = AF_INET6;
3493		udp_conf.use_udp6_rx_checksums =
3494		    !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3495		udp_conf.ipv6_v6only = 1;
3496	} else {
3497		udp_conf.family = AF_INET;
3498	}
3499
3500	udp_conf.local_udp_port = port;
3501	udp_conf.bind_ifindex = ifindex;
3502
3503	/* Open UDP socket */
3504	err = udp_sock_create(net, &udp_conf, &sock);
3505	if (err < 0)
3506		return ERR_PTR(err);
3507
3508	udp_allow_gso(sock->sk);
3509	return sock;
3510}
3511
3512/* Create new listen socket if needed */
3513static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3514					      __be16 port, u32 flags,
3515					      int ifindex)
3516{
3517	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3518	struct vxlan_sock *vs;
3519	struct socket *sock;
3520	unsigned int h;
3521	struct udp_tunnel_sock_cfg tunnel_cfg;
3522
3523	vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3524	if (!vs)
3525		return ERR_PTR(-ENOMEM);
3526
3527	for (h = 0; h < VNI_HASH_SIZE; ++h)
3528		INIT_HLIST_HEAD(&vs->vni_list[h]);
3529
3530	sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3531	if (IS_ERR(sock)) {
3532		kfree(vs);
3533		return ERR_CAST(sock);
3534	}
3535
3536	vs->sock = sock;
3537	refcount_set(&vs->refcnt, 1);
3538	vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3539
3540	spin_lock(&vn->sock_lock);
3541	hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3542	udp_tunnel_notify_add_rx_port(sock,
3543				      (vs->flags & VXLAN_F_GPE) ?
3544				      UDP_TUNNEL_TYPE_VXLAN_GPE :
3545				      UDP_TUNNEL_TYPE_VXLAN);
3546	spin_unlock(&vn->sock_lock);
3547
3548	/* Mark socket as an encapsulation socket. */
3549	memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3550	tunnel_cfg.sk_user_data = vs;
3551	tunnel_cfg.encap_type = 1;
3552	tunnel_cfg.encap_rcv = vxlan_rcv;
3553	tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3554	tunnel_cfg.encap_destroy = NULL;
3555	if (vs->flags & VXLAN_F_GPE) {
3556		tunnel_cfg.gro_receive = vxlan_gpe_gro_receive;
3557		tunnel_cfg.gro_complete = vxlan_gpe_gro_complete;
3558	} else {
3559		tunnel_cfg.gro_receive = vxlan_gro_receive;
3560		tunnel_cfg.gro_complete = vxlan_gro_complete;
3561	}
3562
3563	setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3564
3565	return vs;
3566}
3567
3568static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3569{
3570	struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3571	bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3572	struct vxlan_sock *vs = NULL;
3573	struct vxlan_dev_node *node;
3574	int l3mdev_index = 0;
3575
3576	if (vxlan->cfg.remote_ifindex)
3577		l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3578			vxlan->net, vxlan->cfg.remote_ifindex);
3579
3580	if (!vxlan->cfg.no_share) {
3581		spin_lock(&vn->sock_lock);
3582		vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3583				     vxlan->cfg.dst_port, vxlan->cfg.flags,
3584				     l3mdev_index);
3585		if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3586			spin_unlock(&vn->sock_lock);
3587			return -EBUSY;
3588		}
3589		spin_unlock(&vn->sock_lock);
3590	}
3591	if (!vs)
3592		vs = vxlan_socket_create(vxlan->net, ipv6,
3593					 vxlan->cfg.dst_port, vxlan->cfg.flags,
3594					 l3mdev_index);
3595	if (IS_ERR(vs))
3596		return PTR_ERR(vs);
3597#if IS_ENABLED(CONFIG_IPV6)
3598	if (ipv6) {
3599		rcu_assign_pointer(vxlan->vn6_sock, vs);
3600		node = &vxlan->hlist6;
3601	} else
3602#endif
3603	{
3604		rcu_assign_pointer(vxlan->vn4_sock, vs);
3605		node = &vxlan->hlist4;
3606	}
3607
3608	if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER))
3609		vxlan_vs_add_vnigrp(vxlan, vs, ipv6);
3610	else
3611		vxlan_vs_add_dev(vs, vxlan, node);
3612
3613	return 0;
3614}
3615
3616static int vxlan_sock_add(struct vxlan_dev *vxlan)
3617{
3618	bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3619	bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3620	bool ipv4 = !ipv6 || metadata;
3621	int ret = 0;
3622
3623	RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3624#if IS_ENABLED(CONFIG_IPV6)
3625	RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3626	if (ipv6) {
3627		ret = __vxlan_sock_add(vxlan, true);
3628		if (ret < 0 && ret != -EAFNOSUPPORT)
3629			ipv4 = false;
3630	}
3631#endif
3632	if (ipv4)
3633		ret = __vxlan_sock_add(vxlan, false);
3634	if (ret < 0)
3635		vxlan_sock_release(vxlan);
3636	return ret;
3637}
3638
3639int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan,
3640		     struct vxlan_config *conf, __be32 vni)
3641{
3642	struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3643	struct vxlan_dev *tmp;
3644
3645	list_for_each_entry(tmp, &vn->vxlan_list, next) {
3646		if (tmp == vxlan)
3647			continue;
3648		if (tmp->cfg.flags & VXLAN_F_VNIFILTER) {
3649			if (!vxlan_vnifilter_lookup(tmp, vni))
3650				continue;
3651		} else if (tmp->cfg.vni != vni) {
3652			continue;
3653		}
3654		if (tmp->cfg.dst_port != conf->dst_port)
3655			continue;
3656		if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3657		    (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3658			continue;
3659
3660		if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3661		    tmp->cfg.remote_ifindex != conf->remote_ifindex)
3662			continue;
3663
3664		return -EEXIST;
3665	}
3666
3667	return 0;
3668}
3669
3670static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3671				 struct net_device **lower,
3672				 struct vxlan_dev *old,
3673				 struct netlink_ext_ack *extack)
3674{
3675	bool use_ipv6 = false;
3676
3677	if (conf->flags & VXLAN_F_GPE) {
3678		/* For now, allow GPE only together with
3679		 * COLLECT_METADATA. This can be relaxed later; in such
3680		 * case, the other side of the PtP link will have to be
3681		 * provided.
3682		 */
3683		if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3684		    !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3685			NL_SET_ERR_MSG(extack,
3686				       "VXLAN GPE does not support this combination of attributes");
3687			return -EINVAL;
3688		}
3689	}
3690
3691	if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3692		/* Unless IPv6 is explicitly requested, assume IPv4 */
3693		conf->remote_ip.sa.sa_family = AF_INET;
3694		conf->saddr.sa.sa_family = AF_INET;
3695	} else if (!conf->remote_ip.sa.sa_family) {
3696		conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3697	} else if (!conf->saddr.sa.sa_family) {
3698		conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3699	}
3700
3701	if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3702		NL_SET_ERR_MSG(extack,
3703			       "Local and remote address must be from the same family");
3704		return -EINVAL;
3705	}
3706
3707	if (vxlan_addr_multicast(&conf->saddr)) {
3708		NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3709		return -EINVAL;
3710	}
3711
3712	if (conf->saddr.sa.sa_family == AF_INET6) {
3713		if (!IS_ENABLED(CONFIG_IPV6)) {
3714			NL_SET_ERR_MSG(extack,
3715				       "IPv6 support not enabled in the kernel");
3716			return -EPFNOSUPPORT;
3717		}
3718		use_ipv6 = true;
3719		conf->flags |= VXLAN_F_IPV6;
3720
3721		if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3722			int local_type =
3723				ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3724			int remote_type =
3725				ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3726
3727			if (local_type & IPV6_ADDR_LINKLOCAL) {
3728				if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3729				    (remote_type != IPV6_ADDR_ANY)) {
3730					NL_SET_ERR_MSG(extack,
3731						       "Invalid combination of local and remote address scopes");
3732					return -EINVAL;
3733				}
3734
3735				conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3736			} else {
3737				if (remote_type ==
3738				    (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3739					NL_SET_ERR_MSG(extack,
3740						       "Invalid combination of local and remote address scopes");
3741					return -EINVAL;
3742				}
3743
3744				conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3745			}
3746		}
3747	}
3748
3749	if (conf->label && !use_ipv6) {
3750		NL_SET_ERR_MSG(extack,
3751			       "Label attribute only applies to IPv6 VXLAN devices");
3752		return -EINVAL;
3753	}
3754
3755	if (conf->label_policy && !use_ipv6) {
3756		NL_SET_ERR_MSG(extack,
3757			       "Label policy only applies to IPv6 VXLAN devices");
3758		return -EINVAL;
3759	}
3760
3761	if (conf->remote_ifindex) {
3762		struct net_device *lowerdev;
3763
3764		lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3765		if (!lowerdev) {
3766			NL_SET_ERR_MSG(extack,
3767				       "Invalid local interface, device not found");
3768			return -ENODEV;
3769		}
3770
3771#if IS_ENABLED(CONFIG_IPV6)
3772		if (use_ipv6) {
3773			struct inet6_dev *idev = __in6_dev_get(lowerdev);
3774
3775			if (idev && idev->cnf.disable_ipv6) {
3776				NL_SET_ERR_MSG(extack,
3777					       "IPv6 support disabled by administrator");
3778				return -EPERM;
3779			}
3780		}
3781#endif
3782
3783		*lower = lowerdev;
3784	} else {
3785		if (vxlan_addr_multicast(&conf->remote_ip)) {
3786			NL_SET_ERR_MSG(extack,
3787				       "Local interface required for multicast remote destination");
3788
3789			return -EINVAL;
3790		}
3791
3792#if IS_ENABLED(CONFIG_IPV6)
3793		if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3794			NL_SET_ERR_MSG(extack,
3795				       "Local interface required for link-local local/remote addresses");
3796			return -EINVAL;
3797		}
3798#endif
3799
3800		*lower = NULL;
3801	}
3802
3803	if (!conf->dst_port) {
3804		if (conf->flags & VXLAN_F_GPE)
3805			conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT);
3806		else
3807			conf->dst_port = htons(vxlan_port);
3808	}
3809
3810	if (!conf->age_interval)
3811		conf->age_interval = FDB_AGE_DEFAULT;
3812
3813	if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) {
3814		NL_SET_ERR_MSG(extack,
3815			       "A VXLAN device with the specified VNI already exists");
3816		return -EEXIST;
3817	}
3818
3819	return 0;
3820}
3821
3822static void vxlan_config_apply(struct net_device *dev,
3823			       struct vxlan_config *conf,
3824			       struct net_device *lowerdev,
3825			       struct net *src_net,
3826			       bool changelink)
3827{
3828	struct vxlan_dev *vxlan = netdev_priv(dev);
3829	struct vxlan_rdst *dst = &vxlan->default_dst;
3830	unsigned short needed_headroom = ETH_HLEN;
3831	int max_mtu = ETH_MAX_MTU;
3832	u32 flags = conf->flags;
3833
3834	if (!changelink) {
3835		if (flags & VXLAN_F_GPE)
3836			vxlan_raw_setup(dev);
3837		else
3838			vxlan_ether_setup(dev);
3839
3840		if (conf->mtu)
3841			dev->mtu = conf->mtu;
3842
3843		vxlan->net = src_net;
3844	}
3845
3846	dst->remote_vni = conf->vni;
3847
3848	memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3849
3850	if (lowerdev) {
3851		dst->remote_ifindex = conf->remote_ifindex;
3852
3853		netif_inherit_tso_max(dev, lowerdev);
3854
3855		needed_headroom = lowerdev->hard_header_len;
3856		needed_headroom += lowerdev->needed_headroom;
3857
3858		dev->needed_tailroom = lowerdev->needed_tailroom;
3859
3860		max_mtu = lowerdev->mtu - vxlan_headroom(flags);
3861		if (max_mtu < ETH_MIN_MTU)
3862			max_mtu = ETH_MIN_MTU;
3863
3864		if (!changelink && !conf->mtu)
3865			dev->mtu = max_mtu;
3866	}
3867
3868	if (dev->mtu > max_mtu)
3869		dev->mtu = max_mtu;
3870
3871	if (flags & VXLAN_F_COLLECT_METADATA)
3872		flags |= VXLAN_F_IPV6;
3873	needed_headroom += vxlan_headroom(flags);
3874	dev->needed_headroom = needed_headroom;
3875
3876	memcpy(&vxlan->cfg, conf, sizeof(*conf));
3877}
3878
3879static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3880			       struct vxlan_config *conf, bool changelink,
3881			       struct netlink_ext_ack *extack)
3882{
3883	struct vxlan_dev *vxlan = netdev_priv(dev);
3884	struct net_device *lowerdev;
3885	int ret;
3886
3887	ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3888	if (ret)
3889		return ret;
3890
3891	vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3892
3893	return 0;
3894}
3895
3896static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3897			      struct vxlan_config *conf,
3898			      struct netlink_ext_ack *extack)
3899{
3900	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3901	struct vxlan_dev *vxlan = netdev_priv(dev);
3902	struct net_device *remote_dev = NULL;
3903	struct vxlan_fdb *f = NULL;
3904	bool unregister = false;
3905	struct vxlan_rdst *dst;
3906	int err;
3907
3908	dst = &vxlan->default_dst;
3909	err = vxlan_dev_configure(net, dev, conf, false, extack);
3910	if (err)
3911		return err;
3912
3913	dev->ethtool_ops = &vxlan_ethtool_ops;
3914
3915	/* create an fdb entry for a valid default destination */
3916	if (!vxlan_addr_any(&dst->remote_ip)) {
3917		err = vxlan_fdb_create(vxlan, all_zeros_mac,
3918				       &dst->remote_ip,
3919				       NUD_REACHABLE | NUD_PERMANENT,
3920				       vxlan->cfg.dst_port,
3921				       dst->remote_vni,
3922				       dst->remote_vni,
3923				       dst->remote_ifindex,
3924				       NTF_SELF, 0, &f, extack);
3925		if (err)
3926			return err;
3927	}
3928
3929	err = register_netdevice(dev);
3930	if (err)
3931		goto errout;
3932	unregister = true;
3933
3934	if (dst->remote_ifindex) {
3935		remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3936		if (!remote_dev) {
3937			err = -ENODEV;
3938			goto errout;
3939		}
3940
3941		err = netdev_upper_dev_link(remote_dev, dev, extack);
3942		if (err)
3943			goto errout;
3944	}
3945
3946	err = rtnl_configure_link(dev, NULL, 0, NULL);
3947	if (err < 0)
3948		goto unlink;
3949
3950	if (f) {
3951		vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
3952
3953		/* notify default fdb entry */
3954		err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
3955				       RTM_NEWNEIGH, true, extack);
3956		if (err) {
3957			vxlan_fdb_destroy(vxlan, f, false, false);
3958			if (remote_dev)
3959				netdev_upper_dev_unlink(remote_dev, dev);
3960			goto unregister;
3961		}
3962	}
3963
3964	list_add(&vxlan->next, &vn->vxlan_list);
3965	if (remote_dev)
3966		dst->remote_dev = remote_dev;
3967	return 0;
3968unlink:
3969	if (remote_dev)
3970		netdev_upper_dev_unlink(remote_dev, dev);
3971errout:
3972	/* unregister_netdevice() destroys the default FDB entry with deletion
3973	 * notification. But the addition notification was not sent yet, so
3974	 * destroy the entry by hand here.
3975	 */
3976	if (f)
3977		__vxlan_fdb_free(f);
3978unregister:
3979	if (unregister)
3980		unregister_netdevice(dev);
3981	return err;
3982}
3983
3984/* Set/clear flags based on attribute */
3985static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3986			  int attrtype, unsigned long mask, bool changelink,
3987			  bool changelink_supported,
3988			  struct netlink_ext_ack *extack)
3989{
3990	unsigned long flags;
3991
3992	if (!tb[attrtype])
3993		return 0;
3994
3995	if (changelink && !changelink_supported) {
3996		vxlan_flag_attr_error(attrtype, extack);
3997		return -EOPNOTSUPP;
3998	}
3999
4000	if (vxlan_policy[attrtype].type == NLA_FLAG)
4001		flags = conf->flags | mask;
4002	else if (nla_get_u8(tb[attrtype]))
4003		flags = conf->flags | mask;
4004	else
4005		flags = conf->flags & ~mask;
4006
4007	conf->flags = flags;
4008
4009	return 0;
4010}
4011
4012static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
4013			 struct net_device *dev, struct vxlan_config *conf,
4014			 bool changelink, struct netlink_ext_ack *extack)
4015{
4016	struct vxlan_dev *vxlan = netdev_priv(dev);
4017	int err = 0;
4018
4019	memset(conf, 0, sizeof(*conf));
4020
4021	/* if changelink operation, start with old existing cfg */
4022	if (changelink)
4023		memcpy(conf, &vxlan->cfg, sizeof(*conf));
4024
4025	if (data[IFLA_VXLAN_ID]) {
4026		__be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
4027
4028		if (changelink && (vni != conf->vni)) {
4029			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
4030			return -EOPNOTSUPP;
4031		}
4032		conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
4033	}
4034
4035	if (data[IFLA_VXLAN_GROUP]) {
4036		if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
4037			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
4038			return -EOPNOTSUPP;
4039		}
4040
4041		conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
4042		conf->remote_ip.sa.sa_family = AF_INET;
4043	} else if (data[IFLA_VXLAN_GROUP6]) {
4044		if (!IS_ENABLED(CONFIG_IPV6)) {
4045			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
4046			return -EPFNOSUPPORT;
4047		}
4048
4049		if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
4050			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
4051			return -EOPNOTSUPP;
4052		}
4053
4054		conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
4055		conf->remote_ip.sa.sa_family = AF_INET6;
4056	}
4057
4058	if (data[IFLA_VXLAN_LOCAL]) {
4059		if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
4060			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
4061			return -EOPNOTSUPP;
4062		}
4063
4064		conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
4065		conf->saddr.sa.sa_family = AF_INET;
4066	} else if (data[IFLA_VXLAN_LOCAL6]) {
4067		if (!IS_ENABLED(CONFIG_IPV6)) {
4068			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
4069			return -EPFNOSUPPORT;
4070		}
4071
4072		if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
4073			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
4074			return -EOPNOTSUPP;
4075		}
4076
4077		/* TODO: respect scope id */
4078		conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
4079		conf->saddr.sa.sa_family = AF_INET6;
4080	}
4081
4082	if (data[IFLA_VXLAN_LINK])
4083		conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
4084
4085	if (data[IFLA_VXLAN_TOS])
4086		conf->tos  = nla_get_u8(data[IFLA_VXLAN_TOS]);
4087
4088	if (data[IFLA_VXLAN_TTL])
4089		conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
4090
4091	if (data[IFLA_VXLAN_TTL_INHERIT]) {
4092		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4093				    VXLAN_F_TTL_INHERIT, changelink, false,
4094				    extack);
4095		if (err)
4096			return err;
4097
4098	}
4099
4100	if (data[IFLA_VXLAN_LABEL])
4101		conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
4102			     IPV6_FLOWLABEL_MASK;
4103	if (data[IFLA_VXLAN_LABEL_POLICY])
4104		conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]);
4105
4106	if (data[IFLA_VXLAN_LEARNING]) {
4107		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4108				    VXLAN_F_LEARN, changelink, true,
4109				    extack);
4110		if (err)
4111			return err;
4112	} else if (!changelink) {
4113		/* default to learn on a new device */
4114		conf->flags |= VXLAN_F_LEARN;
4115	}
4116
4117	if (data[IFLA_VXLAN_AGEING])
4118		conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
4119
4120	if (data[IFLA_VXLAN_PROXY]) {
4121		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4122				    VXLAN_F_PROXY, changelink, false,
4123				    extack);
4124		if (err)
4125			return err;
4126	}
4127
4128	if (data[IFLA_VXLAN_RSC]) {
4129		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4130				    VXLAN_F_RSC, changelink, false,
4131				    extack);
4132		if (err)
4133			return err;
4134	}
4135
4136	if (data[IFLA_VXLAN_L2MISS]) {
4137		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4138				    VXLAN_F_L2MISS, changelink, false,
4139				    extack);
4140		if (err)
4141			return err;
4142	}
4143
4144	if (data[IFLA_VXLAN_L3MISS]) {
4145		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4146				    VXLAN_F_L3MISS, changelink, false,
4147				    extack);
4148		if (err)
4149			return err;
4150	}
4151
4152	if (data[IFLA_VXLAN_LIMIT]) {
4153		if (changelink) {
4154			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4155					    "Cannot change limit");
4156			return -EOPNOTSUPP;
4157		}
4158		conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4159	}
4160
4161	if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4162		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4163				    VXLAN_F_COLLECT_METADATA, changelink, false,
4164				    extack);
4165		if (err)
4166			return err;
4167	}
4168
4169	if (data[IFLA_VXLAN_PORT_RANGE]) {
4170		if (!changelink) {
4171			const struct ifla_vxlan_port_range *p
4172				= nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4173			conf->port_min = ntohs(p->low);
4174			conf->port_max = ntohs(p->high);
4175		} else {
4176			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4177					    "Cannot change port range");
4178			return -EOPNOTSUPP;
4179		}
4180	}
4181
4182	if (data[IFLA_VXLAN_PORT]) {
4183		if (changelink) {
4184			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4185					    "Cannot change port");
4186			return -EOPNOTSUPP;
4187		}
4188		conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4189	}
4190
4191	if (data[IFLA_VXLAN_UDP_CSUM]) {
4192		if (changelink) {
4193			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4194					    "Cannot change UDP_CSUM flag");
4195			return -EOPNOTSUPP;
4196		}
4197		if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4198			conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4199	}
4200
4201	if (data[IFLA_VXLAN_LOCALBYPASS]) {
4202		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS,
4203				    VXLAN_F_LOCALBYPASS, changelink,
4204				    true, extack);
4205		if (err)
4206			return err;
4207	} else if (!changelink) {
4208		/* default to local bypass on a new device */
4209		conf->flags |= VXLAN_F_LOCALBYPASS;
4210	}
4211
4212	if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4213		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4214				    VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4215				    false, extack);
4216		if (err)
4217			return err;
4218	}
4219
4220	if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4221		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4222				    VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4223				    false, extack);
4224		if (err)
4225			return err;
4226	}
4227
4228	if (data[IFLA_VXLAN_REMCSUM_TX]) {
4229		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4230				    VXLAN_F_REMCSUM_TX, changelink, false,
4231				    extack);
4232		if (err)
4233			return err;
4234	}
4235
4236	if (data[IFLA_VXLAN_REMCSUM_RX]) {
4237		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4238				    VXLAN_F_REMCSUM_RX, changelink, false,
4239				    extack);
4240		if (err)
4241			return err;
4242	}
4243
4244	if (data[IFLA_VXLAN_GBP]) {
4245		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4246				    VXLAN_F_GBP, changelink, false, extack);
4247		if (err)
4248			return err;
4249	}
4250
4251	if (data[IFLA_VXLAN_GPE]) {
4252		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4253				    VXLAN_F_GPE, changelink, false,
4254				    extack);
4255		if (err)
4256			return err;
4257	}
4258
4259	if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4260		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4261				    VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4262				    false, extack);
4263		if (err)
4264			return err;
4265	}
4266
4267	if (tb[IFLA_MTU]) {
4268		if (changelink) {
4269			NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4270					    "Cannot change mtu");
4271			return -EOPNOTSUPP;
4272		}
4273		conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4274	}
4275
4276	if (data[IFLA_VXLAN_DF])
4277		conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4278
4279	if (data[IFLA_VXLAN_VNIFILTER]) {
4280		err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER,
4281				    VXLAN_F_VNIFILTER, changelink, false,
4282				    extack);
4283		if (err)
4284			return err;
4285
4286		if ((conf->flags & VXLAN_F_VNIFILTER) &&
4287		    !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
4288			NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER],
4289					    "vxlan vnifilter only valid in collect metadata mode");
4290			return -EINVAL;
4291		}
4292	}
4293
4294	return 0;
4295}
4296
4297static int vxlan_newlink(struct net *src_net, struct net_device *dev,
4298			 struct nlattr *tb[], struct nlattr *data[],
4299			 struct netlink_ext_ack *extack)
4300{
4301	struct vxlan_config conf;
4302	int err;
4303
4304	err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4305	if (err)
4306		return err;
4307
4308	return __vxlan_dev_create(src_net, dev, &conf, extack);
4309}
4310
4311static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4312			    struct nlattr *data[],
4313			    struct netlink_ext_ack *extack)
4314{
4315	struct vxlan_dev *vxlan = netdev_priv(dev);
4316	struct net_device *lowerdev;
4317	struct vxlan_config conf;
4318	struct vxlan_rdst *dst;
4319	int err;
4320
4321	dst = &vxlan->default_dst;
4322	err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4323	if (err)
4324		return err;
4325
4326	err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4327				    vxlan, extack);
4328	if (err)
4329		return err;
4330
4331	if (dst->remote_dev == lowerdev)
4332		lowerdev = NULL;
4333
4334	err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4335					     extack);
4336	if (err)
4337		return err;
4338
4339	/* handle default dst entry */
4340	if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
4341		u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4342
4343		spin_lock_bh(&vxlan->hash_lock[hash_index]);
4344		if (!vxlan_addr_any(&conf.remote_ip)) {
4345			err = vxlan_fdb_update(vxlan, all_zeros_mac,
4346					       &conf.remote_ip,
4347					       NUD_REACHABLE | NUD_PERMANENT,
4348					       NLM_F_APPEND | NLM_F_CREATE,
4349					       vxlan->cfg.dst_port,
4350					       conf.vni, conf.vni,
4351					       conf.remote_ifindex,
4352					       NTF_SELF, 0, true, extack);
4353			if (err) {
4354				spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4355				netdev_adjacent_change_abort(dst->remote_dev,
4356							     lowerdev, dev);
4357				return err;
4358			}
4359		}
4360		if (!vxlan_addr_any(&dst->remote_ip))
4361			__vxlan_fdb_delete(vxlan, all_zeros_mac,
4362					   dst->remote_ip,
4363					   vxlan->cfg.dst_port,
4364					   dst->remote_vni,
4365					   dst->remote_vni,
4366					   dst->remote_ifindex,
4367					   true);
4368		spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4369
4370		/* If vni filtering device, also update fdb entries of
4371		 * all vnis that were using default remote ip
4372		 */
4373		if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
4374			err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip,
4375							 &conf.remote_ip, extack);
4376			if (err) {
4377				netdev_adjacent_change_abort(dst->remote_dev,
4378							     lowerdev, dev);
4379				return err;
4380			}
4381		}
4382	}
4383
4384	if (conf.age_interval != vxlan->cfg.age_interval)
4385		mod_timer(&vxlan->age_timer, jiffies);
4386
4387	netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4388	if (lowerdev && lowerdev != dst->remote_dev)
4389		dst->remote_dev = lowerdev;
4390	vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4391	return 0;
4392}
4393
4394static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4395{
4396	struct vxlan_dev *vxlan = netdev_priv(dev);
4397	struct vxlan_fdb_flush_desc desc = {
4398		/* Default entry is deleted at vxlan_uninit. */
4399		.ignore_default_entry = true,
4400	};
4401
4402	vxlan_flush(vxlan, &desc);
4403
4404	list_del(&vxlan->next);
4405	unregister_netdevice_queue(dev, head);
4406	if (vxlan->default_dst.remote_dev)
4407		netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4408}
4409
4410static size_t vxlan_get_size(const struct net_device *dev)
4411{
4412	return nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_ID */
4413		nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4414		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_LINK */
4415		nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4416		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TTL */
4417		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TTL_INHERIT */
4418		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_TOS */
4419		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_DF */
4420		nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4421		nla_total_size(sizeof(__u32)) +  /* IFLA_VXLAN_LABEL_POLICY */
4422		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_LEARNING */
4423		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_PROXY */
4424		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_RSC */
4425		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_L2MISS */
4426		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_L3MISS */
4427		nla_total_size(sizeof(__u8)) +	/* IFLA_VXLAN_COLLECT_METADATA */
4428		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_AGEING */
4429		nla_total_size(sizeof(__u32)) +	/* IFLA_VXLAN_LIMIT */
4430		nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4431		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4432		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4433		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4434		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4435		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4436		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */
4437		/* IFLA_VXLAN_PORT_RANGE */
4438		nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4439		nla_total_size(0) + /* IFLA_VXLAN_GBP */
4440		nla_total_size(0) + /* IFLA_VXLAN_GPE */
4441		nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */
4442		nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */
4443		0;
4444}
4445
4446static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4447{
4448	const struct vxlan_dev *vxlan = netdev_priv(dev);
4449	const struct vxlan_rdst *dst = &vxlan->default_dst;
4450	struct ifla_vxlan_port_range ports = {
4451		.low =  htons(vxlan->cfg.port_min),
4452		.high = htons(vxlan->cfg.port_max),
4453	};
4454
4455	if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4456		goto nla_put_failure;
4457
4458	if (!vxlan_addr_any(&dst->remote_ip)) {
4459		if (dst->remote_ip.sa.sa_family == AF_INET) {
4460			if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4461					    dst->remote_ip.sin.sin_addr.s_addr))
4462				goto nla_put_failure;
4463#if IS_ENABLED(CONFIG_IPV6)
4464		} else {
4465			if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4466					     &dst->remote_ip.sin6.sin6_addr))
4467				goto nla_put_failure;
4468#endif
4469		}
4470	}
4471
4472	if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4473		goto nla_put_failure;
4474
4475	if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4476		if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4477			if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4478					    vxlan->cfg.saddr.sin.sin_addr.s_addr))
4479				goto nla_put_failure;
4480#if IS_ENABLED(CONFIG_IPV6)
4481		} else {
4482			if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4483					     &vxlan->cfg.saddr.sin6.sin6_addr))
4484				goto nla_put_failure;
4485#endif
4486		}
4487	}
4488
4489	if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4490	    nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4491		       !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4492	    nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4493	    nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4494	    nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4495	    nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) ||
4496	    nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4497		       !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4498	    nla_put_u8(skb, IFLA_VXLAN_PROXY,
4499		       !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4500	    nla_put_u8(skb, IFLA_VXLAN_RSC,
4501		       !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4502	    nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4503		       !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4504	    nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4505		       !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4506	    nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4507		       !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4508	    nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4509	    nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4510	    nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4511	    nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4512		       !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4513	    nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4514		       !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4515	    nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4516		       !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4517	    nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4518		       !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4519	    nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4520		       !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) ||
4521	    nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS,
4522		       !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS)))
4523		goto nla_put_failure;
4524
4525	if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4526		goto nla_put_failure;
4527
4528	if (vxlan->cfg.flags & VXLAN_F_GBP &&
4529	    nla_put_flag(skb, IFLA_VXLAN_GBP))
4530		goto nla_put_failure;
4531
4532	if (vxlan->cfg.flags & VXLAN_F_GPE &&
4533	    nla_put_flag(skb, IFLA_VXLAN_GPE))
4534		goto nla_put_failure;
4535
4536	if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4537	    nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4538		goto nla_put_failure;
4539
4540	if (vxlan->cfg.flags & VXLAN_F_VNIFILTER &&
4541	    nla_put_u8(skb, IFLA_VXLAN_VNIFILTER,
4542		       !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER)))
4543		goto nla_put_failure;
4544
4545	return 0;
4546
4547nla_put_failure:
4548	return -EMSGSIZE;
4549}
4550
4551static struct net *vxlan_get_link_net(const struct net_device *dev)
4552{
4553	struct vxlan_dev *vxlan = netdev_priv(dev);
4554
4555	return vxlan->net;
4556}
4557
4558static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4559	.kind		= "vxlan",
4560	.maxtype	= IFLA_VXLAN_MAX,
4561	.policy		= vxlan_policy,
4562	.priv_size	= sizeof(struct vxlan_dev),
4563	.setup		= vxlan_setup,
4564	.validate	= vxlan_validate,
4565	.newlink	= vxlan_newlink,
4566	.changelink	= vxlan_changelink,
4567	.dellink	= vxlan_dellink,
4568	.get_size	= vxlan_get_size,
4569	.fill_info	= vxlan_fill_info,
4570	.get_link_net	= vxlan_get_link_net,
4571};
4572
4573struct net_device *vxlan_dev_create(struct net *net, const char *name,
4574				    u8 name_assign_type,
4575				    struct vxlan_config *conf)
4576{
4577	struct nlattr *tb[IFLA_MAX + 1];
4578	struct net_device *dev;
4579	int err;
4580
4581	memset(&tb, 0, sizeof(tb));
4582
4583	dev = rtnl_create_link(net, name, name_assign_type,
4584			       &vxlan_link_ops, tb, NULL);
4585	if (IS_ERR(dev))
4586		return dev;
4587
4588	err = __vxlan_dev_create(net, dev, conf, NULL);
4589	if (err < 0) {
4590		free_netdev(dev);
4591		return ERR_PTR(err);
4592	}
4593
4594	err = rtnl_configure_link(dev, NULL, 0, NULL);
4595	if (err < 0) {
4596		LIST_HEAD(list_kill);
4597
4598		vxlan_dellink(dev, &list_kill);
4599		unregister_netdevice_many(&list_kill);
4600		return ERR_PTR(err);
4601	}
4602
4603	return dev;
4604}
4605EXPORT_SYMBOL_GPL(vxlan_dev_create);
4606
4607static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4608					     struct net_device *dev)
4609{
4610	struct vxlan_dev *vxlan, *next;
4611	LIST_HEAD(list_kill);
4612
4613	list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4614		struct vxlan_rdst *dst = &vxlan->default_dst;
4615
4616		/* In case we created vxlan device with carrier
4617		 * and we loose the carrier due to module unload
4618		 * we also need to remove vxlan device. In other
4619		 * cases, it's not necessary and remote_ifindex
4620		 * is 0 here, so no matches.
4621		 */
4622		if (dst->remote_ifindex == dev->ifindex)
4623			vxlan_dellink(vxlan->dev, &list_kill);
4624	}
4625
4626	unregister_netdevice_many(&list_kill);
4627}
4628
4629static int vxlan_netdevice_event(struct notifier_block *unused,
4630				 unsigned long event, void *ptr)
4631{
4632	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4633	struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4634
4635	if (event == NETDEV_UNREGISTER)
4636		vxlan_handle_lowerdev_unregister(vn, dev);
4637	else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
4638		vxlan_offload_rx_ports(dev, true);
4639	else if (event == NETDEV_UDP_TUNNEL_DROP_INFO)
4640		vxlan_offload_rx_ports(dev, false);
4641
4642	return NOTIFY_DONE;
4643}
4644
4645static struct notifier_block vxlan_notifier_block __read_mostly = {
4646	.notifier_call = vxlan_netdevice_event,
4647};
4648
4649static void
4650vxlan_fdb_offloaded_set(struct net_device *dev,
4651			struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4652{
4653	struct vxlan_dev *vxlan = netdev_priv(dev);
4654	struct vxlan_rdst *rdst;
4655	struct vxlan_fdb *f;
4656	u32 hash_index;
4657
4658	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4659
4660	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4661
4662	f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4663	if (!f)
4664		goto out;
4665
4666	rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4667				   fdb_info->remote_port,
4668				   fdb_info->remote_vni,
4669				   fdb_info->remote_ifindex);
4670	if (!rdst)
4671		goto out;
4672
4673	rdst->offloaded = fdb_info->offloaded;
4674
4675out:
4676	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4677}
4678
4679static int
4680vxlan_fdb_external_learn_add(struct net_device *dev,
4681			     struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4682{
4683	struct vxlan_dev *vxlan = netdev_priv(dev);
4684	struct netlink_ext_ack *extack;
4685	u32 hash_index;
4686	int err;
4687
4688	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4689	extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4690
4691	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4692	err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4693			       NUD_REACHABLE,
4694			       NLM_F_CREATE | NLM_F_REPLACE,
4695			       fdb_info->remote_port,
4696			       fdb_info->vni,
4697			       fdb_info->remote_vni,
4698			       fdb_info->remote_ifindex,
4699			       NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4700			       0, false, extack);
4701	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4702
4703	return err;
4704}
4705
4706static int
4707vxlan_fdb_external_learn_del(struct net_device *dev,
4708			     struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4709{
4710	struct vxlan_dev *vxlan = netdev_priv(dev);
4711	struct vxlan_fdb *f;
4712	u32 hash_index;
4713	int err = 0;
4714
4715	hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4716	spin_lock_bh(&vxlan->hash_lock[hash_index]);
4717
4718	f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4719	if (!f)
4720		err = -ENOENT;
4721	else if (f->flags & NTF_EXT_LEARNED)
4722		err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4723					 fdb_info->remote_ip,
4724					 fdb_info->remote_port,
4725					 fdb_info->vni,
4726					 fdb_info->remote_vni,
4727					 fdb_info->remote_ifindex,
4728					 false);
4729
4730	spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4731
4732	return err;
4733}
4734
4735static int vxlan_switchdev_event(struct notifier_block *unused,
4736				 unsigned long event, void *ptr)
4737{
4738	struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4739	struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4740	int err = 0;
4741
4742	switch (event) {
4743	case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4744		vxlan_fdb_offloaded_set(dev, ptr);
4745		break;
4746	case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4747		fdb_info = ptr;
4748		err = vxlan_fdb_external_learn_add(dev, fdb_info);
4749		if (err) {
4750			err = notifier_from_errno(err);
4751			break;
4752		}
4753		fdb_info->offloaded = true;
4754		vxlan_fdb_offloaded_set(dev, fdb_info);
4755		break;
4756	case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4757		fdb_info = ptr;
4758		err = vxlan_fdb_external_learn_del(dev, fdb_info);
4759		if (err) {
4760			err = notifier_from_errno(err);
4761			break;
4762		}
4763		fdb_info->offloaded = false;
4764		vxlan_fdb_offloaded_set(dev, fdb_info);
4765		break;
4766	}
4767
4768	return err;
4769}
4770
4771static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4772	.notifier_call = vxlan_switchdev_event,
4773};
4774
4775static void vxlan_fdb_nh_flush(struct nexthop *nh)
4776{
4777	struct vxlan_fdb *fdb;
4778	struct vxlan_dev *vxlan;
4779	u32 hash_index;
4780
4781	rcu_read_lock();
4782	list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4783		vxlan = rcu_dereference(fdb->vdev);
4784		WARN_ON(!vxlan);
4785		hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4786					    vxlan->default_dst.remote_vni);
4787		spin_lock_bh(&vxlan->hash_lock[hash_index]);
4788		if (!hlist_unhashed(&fdb->hlist))
4789			vxlan_fdb_destroy(vxlan, fdb, false, false);
4790		spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4791	}
4792	rcu_read_unlock();
4793}
4794
4795static int vxlan_nexthop_event(struct notifier_block *nb,
4796			       unsigned long event, void *ptr)
4797{
4798	struct nh_notifier_info *info = ptr;
4799	struct nexthop *nh;
4800
4801	if (event != NEXTHOP_EVENT_DEL)
4802		return NOTIFY_DONE;
4803
4804	nh = nexthop_find_by_id(info->net, info->id);
4805	if (!nh)
4806		return NOTIFY_DONE;
4807
4808	vxlan_fdb_nh_flush(nh);
4809
4810	return NOTIFY_DONE;
4811}
4812
4813static __net_init int vxlan_init_net(struct net *net)
4814{
4815	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4816	unsigned int h;
4817
4818	INIT_LIST_HEAD(&vn->vxlan_list);
4819	spin_lock_init(&vn->sock_lock);
4820	vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event;
4821
4822	for (h = 0; h < PORT_HASH_SIZE; ++h)
4823		INIT_HLIST_HEAD(&vn->sock_list[h]);
4824
4825	return register_nexthop_notifier(net, &vn->nexthop_notifier_block,
4826					 NULL);
4827}
4828
4829static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
4830{
4831	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4832	struct vxlan_dev *vxlan, *next;
4833	struct net_device *dev, *aux;
4834
4835	for_each_netdev_safe(net, dev, aux)
4836		if (dev->rtnl_link_ops == &vxlan_link_ops)
4837			unregister_netdevice_queue(dev, head);
4838
4839	list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4840		/* If vxlan->dev is in the same netns, it has already been added
4841		 * to the list by the previous loop.
4842		 */
4843		if (!net_eq(dev_net(vxlan->dev), net))
4844			unregister_netdevice_queue(vxlan->dev, head);
4845	}
4846
4847}
4848
4849static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4850{
4851	struct net *net;
4852	LIST_HEAD(list);
4853	unsigned int h;
4854
4855	list_for_each_entry(net, net_list, exit_list) {
4856		struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4857
4858		unregister_nexthop_notifier(net, &vn->nexthop_notifier_block);
4859	}
4860	rtnl_lock();
4861	list_for_each_entry(net, net_list, exit_list)
4862		vxlan_destroy_tunnels(net, &list);
4863
4864	unregister_netdevice_many(&list);
4865	rtnl_unlock();
4866
4867	list_for_each_entry(net, net_list, exit_list) {
4868		struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4869
4870		for (h = 0; h < PORT_HASH_SIZE; ++h)
4871			WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4872	}
4873}
4874
4875static struct pernet_operations vxlan_net_ops = {
4876	.init = vxlan_init_net,
4877	.exit_batch = vxlan_exit_batch_net,
4878	.id   = &vxlan_net_id,
4879	.size = sizeof(struct vxlan_net),
4880};
4881
4882static int __init vxlan_init_module(void)
4883{
4884	int rc;
4885
4886	get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4887
4888	rc = register_pernet_subsys(&vxlan_net_ops);
4889	if (rc)
4890		goto out1;
4891
4892	rc = register_netdevice_notifier(&vxlan_notifier_block);
4893	if (rc)
4894		goto out2;
4895
4896	rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4897	if (rc)
4898		goto out3;
4899
4900	rc = rtnl_link_register(&vxlan_link_ops);
4901	if (rc)
4902		goto out4;
4903
4904	vxlan_vnifilter_init();
4905
4906	return 0;
4907out4:
4908	unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4909out3:
4910	unregister_netdevice_notifier(&vxlan_notifier_block);
4911out2:
4912	unregister_pernet_subsys(&vxlan_net_ops);
4913out1:
4914	return rc;
4915}
4916late_initcall(vxlan_init_module);
4917
4918static void __exit vxlan_cleanup_module(void)
4919{
4920	vxlan_vnifilter_uninit();
4921	rtnl_link_unregister(&vxlan_link_ops);
4922	unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4923	unregister_netdevice_notifier(&vxlan_notifier_block);
4924	unregister_pernet_subsys(&vxlan_net_ops);
4925	/* rcu_barrier() is called by netns */
4926}
4927module_exit(vxlan_cleanup_module);
4928
4929MODULE_LICENSE("GPL");
4930MODULE_VERSION(VXLAN_VERSION);
4931MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4932MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4933MODULE_ALIAS_RTNL_LINK("vxlan");