drivers/net/vxlan.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * VXLAN: Virtual eXtensible Local Area Network
   4  *
   5  * Copyright (c) 2012-2013 Vyatta Inc.
   6  */
   7
   8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
   9
  10 #include <linux/kernel.h>
  11 #include <linux/module.h>
  12 #include <linux/errno.h>
  13 #include <linux/slab.h>
  14 #include <linux/udp.h>
  15 #include <linux/igmp.h>
  16 #include <linux/if_ether.h>
  17 #include <linux/ethtool.h>
  18 #include <net/arp.h>
  19 #include <net/ndisc.h>
  20 #include <net/ipv6_stubs.h>
  21 #include <net/ip.h>
  22 #include <net/icmp.h>
  23 #include <net/rtnetlink.h>
  24 #include <net/inet_ecn.h>
  25 #include <net/net_namespace.h>
  26 #include <net/netns/generic.h>
  27 #include <net/tun_proto.h>
  28 #include <net/vxlan.h>
  29 #include <net/nexthop.h>
  30
  31 #if IS_ENABLED(CONFIG_IPV6)
  32 #include <net/ip6_tunnel.h>
  33 #include <net/ip6_checksum.h>
  34 #endif
  35
  36 #define VXLAN_VERSION   "0.1"
  37
  38 #define PORT_HASH_BITS  8
  39 #define PORT_HASH_SIZE  (1<<PORT_HASH_BITS)
  40 #define FDB_AGE_DEFAULT 300 /* 5 min */
  41 #define FDB_AGE_INTERVAL (10 * HZ)      /* rescan interval */
  42
  43 /* UDP port for VXLAN traffic.
  44  * The IANA assigned port is 4789, but the Linux default is 8472
  45  * for compatibility with early adopters.
  46  */
  47 static unsigned short vxlan_port __read_mostly = 8472;
  48 module_param_named(udp_port, vxlan_port, ushort, 0444);
  49 MODULE_PARM_DESC(udp_port, "Destination UDP port");
  50
  51 static bool log_ecn_error = true;
  52 module_param(log_ecn_error, bool, 0644);
  53 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  54
  55 static unsigned int vxlan_net_id;
  56 static struct rtnl_link_ops vxlan_link_ops;
  57
  58 static const u8 all_zeros_mac[ETH_ALEN + 2];
  59
  60 static int vxlan_sock_add(struct vxlan_dev *vxlan);
  61
  62 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
  63
  64 /* per-network namespace private data for this module */
  65 struct vxlan_net {
  66         struct list_head  vxlan_list;
  67         struct hlist_head sock_list[PORT_HASH_SIZE];
  68         spinlock_t        sock_lock;
  69         struct notifier_block nexthop_notifier_block;
  70 };
  71
  72 /* Forwarding table entry */
  73 struct vxlan_fdb {
  74         struct hlist_node hlist;        /* linked list of entries */
  75         struct rcu_head   rcu;
  76         unsigned long     updated;      /* jiffies */
  77         unsigned long     used;
  78         struct list_head  remotes;
  79         u8                eth_addr[ETH_ALEN];
  80         u16               state;        /* see ndm_state */
  81         __be32            vni;
  82         u16               flags;        /* see ndm_flags and below */
  83         struct list_head  nh_list;
  84         struct nexthop __rcu *nh;
  85         struct vxlan_dev  __rcu *vdev;
  86 };
  87
  88 #define NTF_VXLAN_ADDED_BY_USER 0x100
  89
  90 /* salt for hash table */
  91 static u32 vxlan_salt __read_mostly;
  92
  93 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
  94 {
  95         return vs->flags & VXLAN_F_COLLECT_METADATA ||
  96                ip_tunnel_collect_metadata();
  97 }
  98
  99 #if IS_ENABLED(CONFIG_IPV6)
 100 static inline
 101 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
 102 {
 103         if (a->sa.sa_family != b->sa.sa_family)
 104                 return false;
 105         if (a->sa.sa_family == AF_INET6)
 106                 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
 107         else
 108                 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
 109 }
 110
 111 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
 112 {
 113         if (nla_len(nla) >= sizeof(struct in6_addr)) {
 114                 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
 115                 ip->sa.sa_family = AF_INET6;
 116                 return 0;
 117         } else if (nla_len(nla) >= sizeof(__be32)) {
 118                 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
 119                 ip->sa.sa_family = AF_INET;
 120                 return 0;
 121         } else {
 122                 return -EAFNOSUPPORT;
 123         }
 124 }
 125
 126 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
 127                               const union vxlan_addr *ip)
 128 {
 129         if (ip->sa.sa_family == AF_INET6)
 130                 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
 131         else
 132                 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
 133 }
 134
 135 #else /* !CONFIG_IPV6 */
 136
 137 static inline
 138 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
 139 {
 140         return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
 141 }
 142
 143 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
 144 {
 145         if (nla_len(nla) >= sizeof(struct in6_addr)) {
 146                 return -EAFNOSUPPORT;
 147         } else if (nla_len(nla) >= sizeof(__be32)) {
 148                 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
 149                 ip->sa.sa_family = AF_INET;
 150                 return 0;
 151         } else {
 152                 return -EAFNOSUPPORT;
 153         }
 154 }
 155
 156 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
 157                               const union vxlan_addr *ip)
 158 {
 159         return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
 160 }
 161 #endif
 162
 163 /* Virtual Network hash table head */
 164 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
 165 {
 166         return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
 167 }
 168
 169 /* Socket hash table head */
 170 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
 171 {
 172         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
 173
 174         return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
 175 }
 176
 177 /* First remote destination for a forwarding entry.
 178  * Guaranteed to be non-NULL because remotes are never deleted.
 179  */
 180 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
 181 {
 182         if (rcu_access_pointer(fdb->nh))
 183                 return NULL;
 184         return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
 185 }
 186
 187 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
 188 {
 189         if (rcu_access_pointer(fdb->nh))
 190                 return NULL;
 191         return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
 192 }
 193
 194 /* Find VXLAN socket based on network namespace, address family, UDP port,
 195  * enabled unshareable flags and socket device binding (see l3mdev with
 196  * non-default VRF).
 197  */
 198 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
 199                                           __be16 port, u32 flags, int ifindex)
 200 {
 201         struct vxlan_sock *vs;
 202
 203         flags &= VXLAN_F_RCV_FLAGS;
 204
 205         hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
 206                 if (inet_sk(vs->sock->sk)->inet_sport == port &&
 207                     vxlan_get_sk_family(vs) == family &&
 208                     vs->flags == flags &&
 209                     vs->sock->sk->sk_bound_dev_if == ifindex)
 210                         return vs;
 211         }
 212         return NULL;
 213 }
 214
 215 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
 216                                            __be32 vni)
 217 {
 218         struct vxlan_dev_node *node;
 219
 220         /* For flow based devices, map all packets to VNI 0 */
 221         if (vs->flags & VXLAN_F_COLLECT_METADATA)
 222                 vni = 0;
 223
 224         hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
 225                 if (node->vxlan->default_dst.remote_vni != vni)
 226                         continue;
 227
 228                 if (IS_ENABLED(CONFIG_IPV6)) {
 229                         const struct vxlan_config *cfg = &node->vxlan->cfg;
 230
 231                         if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
 232                             cfg->remote_ifindex != ifindex)
 233                                 continue;
 234                 }
 235
 236                 return node->vxlan;
 237         }
 238
 239         return NULL;
 240 }
 241
 242 /* Look up VNI in a per net namespace table */
 243 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
 244                                         __be32 vni, sa_family_t family,
 245                                         __be16 port, u32 flags)
 246 {
 247         struct vxlan_sock *vs;
 248
 249         vs = vxlan_find_sock(net, family, port, flags, ifindex);
 250         if (!vs)
 251                 return NULL;
 252
 253         return vxlan_vs_find_vni(vs, ifindex, vni);
 254 }
 255
 256 /* Fill in neighbour message in skbuff. */
 257 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
 258                           const struct vxlan_fdb *fdb,
 259                           u32 portid, u32 seq, int type, unsigned int flags,
 260                           const struct vxlan_rdst *rdst)
 261 {
 262         unsigned long now = jiffies;
 263         struct nda_cacheinfo ci;
 264         bool send_ip, send_eth;
 265         struct nlmsghdr *nlh;
 266         struct nexthop *nh;
 267         struct ndmsg *ndm;
 268         int nh_family;
 269         u32 nh_id;
 270
 271         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
 272         if (nlh == NULL)
 273                 return -EMSGSIZE;
 274
 275         ndm = nlmsg_data(nlh);
 276         memset(ndm, 0, sizeof(*ndm));
 277
 278         send_eth = send_ip = true;
 279
 280         rcu_read_lock();
 281         nh = rcu_dereference(fdb->nh);
 282         if (nh) {
 283                 nh_family = nexthop_get_family(nh);
 284                 nh_id = nh->id;
 285         }
 286         rcu_read_unlock();
 287
 288         if (type == RTM_GETNEIGH) {
 289                 if (rdst) {
 290                         send_ip = !vxlan_addr_any(&rdst->remote_ip);
 291                         ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
 292                 } else if (nh) {
 293                         ndm->ndm_family = nh_family;
 294                 }
 295                 send_eth = !is_zero_ether_addr(fdb->eth_addr);
 296         } else
 297                 ndm->ndm_family = AF_BRIDGE;
 298         ndm->ndm_state = fdb->state;
 299         ndm->ndm_ifindex = vxlan->dev->ifindex;
 300         ndm->ndm_flags = fdb->flags;
 301         if (rdst && rdst->offloaded)
 302                 ndm->ndm_flags |= NTF_OFFLOADED;
 303         ndm->ndm_type = RTN_UNICAST;
 304
 305         if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
 306             nla_put_s32(skb, NDA_LINK_NETNSID,
 307                         peernet2id(dev_net(vxlan->dev), vxlan->net)))
 308                 goto nla_put_failure;
 309
 310         if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
 311                 goto nla_put_failure;
 312         if (nh) {
 313                 if (nla_put_u32(skb, NDA_NH_ID, nh_id))
 314                         goto nla_put_failure;
 315         } else if (rdst) {
 316                 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST,
 317                                                   &rdst->remote_ip))
 318                         goto nla_put_failure;
 319
 320                 if (rdst->remote_port &&
 321                     rdst->remote_port != vxlan->cfg.dst_port &&
 322                     nla_put_be16(skb, NDA_PORT, rdst->remote_port))
 323                         goto nla_put_failure;
 324                 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
 325                     nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
 326                         goto nla_put_failure;
 327                 if (rdst->remote_ifindex &&
 328                     nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
 329                         goto nla_put_failure;
 330         }
 331
 332         if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
 333             nla_put_u32(skb, NDA_SRC_VNI,
 334                         be32_to_cpu(fdb->vni)))
 335                 goto nla_put_failure;
 336
 337         ci.ndm_used      = jiffies_to_clock_t(now - fdb->used);
 338         ci.ndm_confirmed = 0;
 339         ci.ndm_updated   = jiffies_to_clock_t(now - fdb->updated);
 340         ci.ndm_refcnt    = 0;
 341
 342         if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
 343                 goto nla_put_failure;
 344
 345         nlmsg_end(skb, nlh);
 346         return 0;
 347
 348 nla_put_failure:
 349         nlmsg_cancel(skb, nlh);
 350         return -EMSGSIZE;
 351 }
 352
 353 static inline size_t vxlan_nlmsg_size(void)
 354 {
 355         return NLMSG_ALIGN(sizeof(struct ndmsg))
 356                 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
 357                 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
 358                 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
 359                 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
 360                 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
 361                 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
 362                 + nla_total_size(sizeof(struct nda_cacheinfo));
 363 }
 364
 365 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
 366                                struct vxlan_rdst *rd, int type)
 367 {
 368         struct net *net = dev_net(vxlan->dev);
 369         struct sk_buff *skb;
 370         int err = -ENOBUFS;
 371
 372         skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
 373         if (skb == NULL)
 374                 goto errout;
 375
 376         err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
 377         if (err < 0) {
 378                 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
 379                 WARN_ON(err == -EMSGSIZE);
 380                 kfree_skb(skb);
 381                 goto errout;
 382         }
 383
 384         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
 385         return;
 386 errout:
 387         if (err < 0)
 388                 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
 389 }
 390
 391 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
 392                             const struct vxlan_fdb *fdb,
 393                             const struct vxlan_rdst *rd,
 394                             struct netlink_ext_ack *extack,
 395                             struct switchdev_notifier_vxlan_fdb_info *fdb_info)
 396 {
 397         fdb_info->info.dev = vxlan->dev;
 398         fdb_info->info.extack = extack;
 399         fdb_info->remote_ip = rd->remote_ip;
 400         fdb_info->remote_port = rd->remote_port;
 401         fdb_info->remote_vni = rd->remote_vni;
 402         fdb_info->remote_ifindex = rd->remote_ifindex;
 403         memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
 404         fdb_info->vni = fdb->vni;
 405         fdb_info->offloaded = rd->offloaded;
 406         fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
 407 }
 408
 409 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
 410                                               struct vxlan_fdb *fdb,
 411                                               struct vxlan_rdst *rd,
 412                                               bool adding,
 413                                               struct netlink_ext_ack *extack)
 414 {
 415         struct switchdev_notifier_vxlan_fdb_info info;
 416         enum switchdev_notifier_type notifier_type;
 417         int ret;
 418
 419         if (WARN_ON(!rd))
 420                 return 0;
 421
 422         notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
 423                                : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
 424         vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
 425         ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
 426                                        &info.info, extack);
 427         return notifier_to_errno(ret);
 428 }
 429
 430 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
 431                             struct vxlan_rdst *rd, int type, bool swdev_notify,
 432                             struct netlink_ext_ack *extack)
 433 {
 434         int err;
 435
 436         if (swdev_notify && rd) {
 437                 switch (type) {
 438                 case RTM_NEWNEIGH:
 439                         err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
 440                                                                  true, extack);
 441                         if (err)
 442                                 return err;
 443                         break;
 444                 case RTM_DELNEIGH:
 445                         vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
 446                                                            false, extack);
 447                         break;
 448                 }
 449         }
 450
 451         __vxlan_fdb_notify(vxlan, fdb, rd, type);
 452         return 0;
 453 }
 454
 455 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
 456 {
 457         struct vxlan_dev *vxlan = netdev_priv(dev);
 458         struct vxlan_fdb f = {
 459                 .state = NUD_STALE,
 460         };
 461         struct vxlan_rdst remote = {
 462                 .remote_ip = *ipa, /* goes to NDA_DST */
 463                 .remote_vni = cpu_to_be32(VXLAN_N_VID),
 464         };
 465
 466         vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
 467 }
 468
 469 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
 470 {
 471         struct vxlan_fdb f = {
 472                 .state = NUD_STALE,
 473         };
 474         struct vxlan_rdst remote = { };
 475
 476         memcpy(f.eth_addr, eth_addr, ETH_ALEN);
 477
 478         vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
 479 }
 480
 481 /* Hash Ethernet address */
 482 static u32 eth_hash(const unsigned char *addr)
 483 {
 484         u64 value = get_unaligned((u64 *)addr);
 485
 486         /* only want 6 bytes */
 487 #ifdef __BIG_ENDIAN
 488         value >>= 16;
 489 #else
 490         value <<= 16;
 491 #endif
 492         return hash_64(value, FDB_HASH_BITS);
 493 }
 494
 495 static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
 496 {
 497         /* use 1 byte of OUI and 3 bytes of NIC */
 498         u32 key = get_unaligned((u32 *)(addr + 2));
 499
 500         return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
 501 }
 502
 503 static u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
 504 {
 505         if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
 506                 return eth_vni_hash(mac, vni);
 507         else
 508                 return eth_hash(mac);
 509 }
 510
 511 /* Hash chain to use given mac address */
 512 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
 513                                                 const u8 *mac, __be32 vni)
 514 {
 515         return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
 516 }
 517
 518 /* Look up Ethernet address in forwarding table */
 519 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
 520                                           const u8 *mac, __be32 vni)
 521 {
 522         struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
 523         struct vxlan_fdb *f;
 524
 525         hlist_for_each_entry_rcu(f, head, hlist) {
 526                 if (ether_addr_equal(mac, f->eth_addr)) {
 527                         if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
 528                                 if (vni == f->vni)
 529                                         return f;
 530                         } else {
 531                                 return f;
 532                         }
 533                 }
 534         }
 535
 536         return NULL;
 537 }
 538
 539 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
 540                                         const u8 *mac, __be32 vni)
 541 {
 542         struct vxlan_fdb *f;
 543
 544         f = __vxlan_find_mac(vxlan, mac, vni);
 545         if (f && f->used != jiffies)
 546                 f->used = jiffies;
 547
 548         return f;
 549 }
 550
 551 /* caller should hold vxlan->hash_lock */
 552 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
 553                                               union vxlan_addr *ip, __be16 port,
 554                                               __be32 vni, __u32 ifindex)
 555 {
 556         struct vxlan_rdst *rd;
 557
 558         list_for_each_entry(rd, &f->remotes, list) {
 559                 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
 560                     rd->remote_port == port &&
 561                     rd->remote_vni == vni &&
 562                     rd->remote_ifindex == ifindex)
 563                         return rd;
 564         }
 565
 566         return NULL;
 567 }
 568
 569 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
 570                       struct switchdev_notifier_vxlan_fdb_info *fdb_info)
 571 {
 572         struct vxlan_dev *vxlan = netdev_priv(dev);
 573         u8 eth_addr[ETH_ALEN + 2] = { 0 };
 574         struct vxlan_rdst *rdst;
 575         struct vxlan_fdb *f;
 576         int rc = 0;
 577
 578         if (is_multicast_ether_addr(mac) ||
 579             is_zero_ether_addr(mac))
 580                 return -EINVAL;
 581
 582         ether_addr_copy(eth_addr, mac);
 583
 584         rcu_read_lock();
 585
 586         f = __vxlan_find_mac(vxlan, eth_addr, vni);
 587         if (!f) {
 588                 rc = -ENOENT;
 589                 goto out;
 590         }
 591
 592         rdst = first_remote_rcu(f);
 593         vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
 594
 595 out:
 596         rcu_read_unlock();
 597         return rc;
 598 }
 599 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
 600
 601 static int vxlan_fdb_notify_one(struct notifier_block *nb,
 602                                 const struct vxlan_dev *vxlan,
 603                                 const struct vxlan_fdb *f,
 604                                 const struct vxlan_rdst *rdst,
 605                                 struct netlink_ext_ack *extack)
 606 {
 607         struct switchdev_notifier_vxlan_fdb_info fdb_info;
 608         int rc;
 609
 610         vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
 611         rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
 612                                &fdb_info);
 613         return notifier_to_errno(rc);
 614 }
 615
 616 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
 617                      struct notifier_block *nb,
 618                      struct netlink_ext_ack *extack)
 619 {
 620         struct vxlan_dev *vxlan;
 621         struct vxlan_rdst *rdst;
 622         struct vxlan_fdb *f;
 623         unsigned int h;
 624         int rc = 0;
 625
 626         if (!netif_is_vxlan(dev))
 627                 return -EINVAL;
 628         vxlan = netdev_priv(dev);
 629
 630         for (h = 0; h < FDB_HASH_SIZE; ++h) {
 631                 spin_lock_bh(&vxlan->hash_lock[h]);
 632                 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
 633                         if (f->vni == vni) {
 634                                 list_for_each_entry(rdst, &f->remotes, list) {
 635                                         rc = vxlan_fdb_notify_one(nb, vxlan,
 636                                                                   f, rdst,
 637                                                                   extack);
 638                                         if (rc)
 639                                                 goto unlock;
 640                                 }
 641                         }
 642                 }
 643                 spin_unlock_bh(&vxlan->hash_lock[h]);
 644         }
 645         return 0;
 646
 647 unlock:
 648         spin_unlock_bh(&vxlan->hash_lock[h]);
 649         return rc;
 650 }
 651 EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
 652
 653 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
 654 {
 655         struct vxlan_dev *vxlan;
 656         struct vxlan_rdst *rdst;
 657         struct vxlan_fdb *f;
 658         unsigned int h;
 659
 660         if (!netif_is_vxlan(dev))
 661                 return;
 662         vxlan = netdev_priv(dev);
 663
 664         for (h = 0; h < FDB_HASH_SIZE; ++h) {
 665                 spin_lock_bh(&vxlan->hash_lock[h]);
 666                 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
 667                         if (f->vni == vni)
 668                                 list_for_each_entry(rdst, &f->remotes, list)
 669                                         rdst->offloaded = false;
 670                 spin_unlock_bh(&vxlan->hash_lock[h]);
 671         }
 672
 673 }
 674 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
 675
 676 /* Replace destination of unicast mac */
 677 static int vxlan_fdb_replace(struct vxlan_fdb *f,
 678                              union vxlan_addr *ip, __be16 port, __be32 vni,
 679                              __u32 ifindex, struct vxlan_rdst *oldrd)
 680 {
 681         struct vxlan_rdst *rd;
 682
 683         rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
 684         if (rd)
 685                 return 0;
 686
 687         rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
 688         if (!rd)
 689                 return 0;
 690
 691         *oldrd = *rd;
 692         dst_cache_reset(&rd->dst_cache);
 693         rd->remote_ip = *ip;
 694         rd->remote_port = port;
 695         rd->remote_vni = vni;
 696         rd->remote_ifindex = ifindex;
 697         rd->offloaded = false;
 698         return 1;
 699 }
 700
 701 /* Add/update destinations for multicast */
 702 static int vxlan_fdb_append(struct vxlan_fdb *f,
 703                             union vxlan_addr *ip, __be16 port, __be32 vni,
 704                             __u32 ifindex, struct vxlan_rdst **rdp)
 705 {
 706         struct vxlan_rdst *rd;
 707
 708         rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
 709         if (rd)
 710                 return 0;
 711
 712         rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
 713         if (rd == NULL)
 714                 return -ENOBUFS;
 715
 716         if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
 717                 kfree(rd);
 718                 return -ENOBUFS;
 719         }
 720
 721         rd->remote_ip = *ip;
 722         rd->remote_port = port;
 723         rd->offloaded = false;
 724         rd->remote_vni = vni;
 725         rd->remote_ifindex = ifindex;
 726
 727         list_add_tail_rcu(&rd->list, &f->remotes);
 728
 729         *rdp = rd;
 730         return 1;
 731 }
 732
 733 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
 734                                           unsigned int off,
 735                                           struct vxlanhdr *vh, size_t hdrlen,
 736                                           __be32 vni_field,
 737                                           struct gro_remcsum *grc,
 738                                           bool nopartial)
 739 {
 740         size_t start, offset;
 741
 742         if (skb->remcsum_offload)
 743                 return vh;
 744
 745         if (!NAPI_GRO_CB(skb)->csum_valid)
 746                 return NULL;
 747
 748         start = vxlan_rco_start(vni_field);
 749         offset = start + vxlan_rco_offset(vni_field);
 750
 751         vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
 752                                      start, offset, grc, nopartial);
 753
 754         skb->remcsum_offload = 1;
 755
 756         return vh;
 757 }
 758
 759 static struct sk_buff *vxlan_gro_receive(struct sock *sk,
 760                                          struct list_head *head,
 761                                          struct sk_buff *skb)
 762 {
 763         struct sk_buff *pp = NULL;
 764         struct sk_buff *p;
 765         struct vxlanhdr *vh, *vh2;
 766         unsigned int hlen, off_vx;
 767         int flush = 1;
 768         struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
 769         __be32 flags;
 770         struct gro_remcsum grc;
 771
 772         skb_gro_remcsum_init(&grc);
 773
 774         off_vx = skb_gro_offset(skb);
 775         hlen = off_vx + sizeof(*vh);
 776         vh   = skb_gro_header_fast(skb, off_vx);
 777         if (skb_gro_header_hard(skb, hlen)) {
 778                 vh = skb_gro_header_slow(skb, hlen, off_vx);
 779                 if (unlikely(!vh))
 780                         goto out;
 781         }
 782
 783         skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
 784
 785         flags = vh->vx_flags;
 786
 787         if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
 788                 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
 789                                        vh->vx_vni, &grc,
 790                                        !!(vs->flags &
 791                                           VXLAN_F_REMCSUM_NOPARTIAL));
 792
 793                 if (!vh)
 794                         goto out;
 795         }
 796
 797         skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
 798
 799         list_for_each_entry(p, head, list) {
 800                 if (!NAPI_GRO_CB(p)->same_flow)
 801                         continue;
 802
 803                 vh2 = (struct vxlanhdr *)(p->data + off_vx);
 804                 if (vh->vx_flags != vh2->vx_flags ||
 805                     vh->vx_vni != vh2->vx_vni) {
 806                         NAPI_GRO_CB(p)->same_flow = 0;
 807                         continue;
 808                 }
 809         }
 810
 811         pp = call_gro_receive(eth_gro_receive, head, skb);
 812         flush = 0;
 813
 814 out:
 815         skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
 816
 817         return pp;
 818 }
 819
 820 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
 821 {
 822         /* Sets 'skb->inner_mac_header' since we are always called with
 823          * 'skb->encapsulation' set.
 824          */
 825         return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
 826 }
 827
 828 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
 829                                          __u16 state, __be32 src_vni,
 830                                          __u16 ndm_flags)
 831 {
 832         struct vxlan_fdb *f;
 833
 834         f = kmalloc(sizeof(*f), GFP_ATOMIC);
 835         if (!f)
 836                 return NULL;
 837         f->state = state;
 838         f->flags = ndm_flags;
 839         f->updated = f->used = jiffies;
 840         f->vni = src_vni;
 841         f->nh = NULL;
 842         RCU_INIT_POINTER(f->vdev, vxlan);
 843         INIT_LIST_HEAD(&f->nh_list);
 844         INIT_LIST_HEAD(&f->remotes);
 845         memcpy(f->eth_addr, mac, ETH_ALEN);
 846
 847         return f;
 848 }
 849
 850 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
 851                              __be32 src_vni, struct vxlan_fdb *f)
 852 {
 853         ++vxlan->addrcnt;
 854         hlist_add_head_rcu(&f->hlist,
 855                            vxlan_fdb_head(vxlan, mac, src_vni));
 856 }
 857
 858 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
 859                                u32 nhid, struct netlink_ext_ack *extack)
 860 {
 861         struct nexthop *old_nh = rtnl_dereference(fdb->nh);
 862         struct nexthop *nh;
 863         int err = -EINVAL;
 864
 865         if (old_nh && old_nh->id == nhid)
 866                 return 0;
 867
 868         nh = nexthop_find_by_id(vxlan->net, nhid);
 869         if (!nh) {
 870                 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
 871                 goto err_inval;
 872         }
 873
 874         if (nh) {
 875                 if (!nexthop_get(nh)) {
 876                         NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
 877                         nh = NULL;
 878                         goto err_inval;
 879                 }
 880                 if (!nexthop_is_fdb(nh)) {
 881                         NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
 882                         goto err_inval;
 883                 }
 884
 885                 if (!nexthop_is_multipath(nh)) {
 886                         NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
 887                         goto err_inval;
 888                 }
 889
 890                 /* check nexthop group family */
 891                 switch (vxlan->default_dst.remote_ip.sa.sa_family) {
 892                 case AF_INET:
 893                         if (!nexthop_has_v4(nh)) {
 894                                 err = -EAFNOSUPPORT;
 895                                 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
 896                                 goto err_inval;
 897                         }
 898                         break;
 899                 case AF_INET6:
 900                         if (nexthop_has_v4(nh)) {
 901                                 err = -EAFNOSUPPORT;
 902                                 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
 903                                 goto err_inval;
 904                         }
 905                 }
 906         }
 907
 908         if (old_nh) {
 909                 list_del_rcu(&fdb->nh_list);
 910                 nexthop_put(old_nh);
 911         }
 912         rcu_assign_pointer(fdb->nh, nh);
 913         list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
 914         return 1;
 915
 916 err_inval:
 917         if (nh)
 918                 nexthop_put(nh);
 919         return err;
 920 }
 921
 922 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
 923                             const u8 *mac, union vxlan_addr *ip,
 924                             __u16 state, __be16 port, __be32 src_vni,
 925                             __be32 vni, __u32 ifindex, __u16 ndm_flags,
 926                             u32 nhid, struct vxlan_fdb **fdb,
 927                             struct netlink_ext_ack *extack)
 928 {
 929         struct vxlan_rdst *rd = NULL;
 930         struct vxlan_fdb *f;
 931         int rc;
 932
 933         if (vxlan->cfg.addrmax &&
 934             vxlan->addrcnt >= vxlan->cfg.addrmax)
 935                 return -ENOSPC;
 936
 937         netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
 938         f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
 939         if (!f)
 940                 return -ENOMEM;
 941
 942         if (nhid)
 943                 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
 944         else
 945                 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
 946         if (rc < 0)
 947                 goto errout;
 948
 949         *fdb = f;
 950
 951         return 0;
 952
 953 errout:
 954         kfree(f);
 955         return rc;
 956 }
 957
 958 static void __vxlan_fdb_free(struct vxlan_fdb *f)
 959 {
 960         struct vxlan_rdst *rd, *nd;
 961         struct nexthop *nh;
 962
 963         nh = rcu_dereference_raw(f->nh);
 964         if (nh) {
 965                 rcu_assign_pointer(f->nh, NULL);
 966                 rcu_assign_pointer(f->vdev, NULL);
 967                 nexthop_put(nh);
 968         }
 969
 970         list_for_each_entry_safe(rd, nd, &f->remotes, list) {
 971                 dst_cache_destroy(&rd->dst_cache);
 972                 kfree(rd);
 973         }
 974         kfree(f);
 975 }
 976
 977 static void vxlan_fdb_free(struct rcu_head *head)
 978 {
 979         struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
 980
 981         __vxlan_fdb_free(f);
 982 }
 983
 984 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
 985                               bool do_notify, bool swdev_notify)
 986 {
 987         struct vxlan_rdst *rd;
 988
 989         netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
 990
 991         --vxlan->addrcnt;
 992         if (do_notify) {
 993                 if (rcu_access_pointer(f->nh))
 994                         vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
 995                                          swdev_notify, NULL);
 996                 else
 997                         list_for_each_entry(rd, &f->remotes, list)
 998                                 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
 999                                                  swdev_notify, NULL);
1000         }
1001
1002         hlist_del_rcu(&f->hlist);
1003         list_del_rcu(&f->nh_list);
1004         call_rcu(&f->rcu, vxlan_fdb_free);
1005 }
1006
1007 static void vxlan_dst_free(struct rcu_head *head)
1008 {
1009         struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
1010
1011         dst_cache_destroy(&rd->dst_cache);
1012         kfree(rd);
1013 }
1014
1015 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
1016                                      union vxlan_addr *ip,
1017                                      __u16 state, __u16 flags,
1018                                      __be16 port, __be32 vni,
1019                                      __u32 ifindex, __u16 ndm_flags,
1020                                      struct vxlan_fdb *f, u32 nhid,
1021                                      bool swdev_notify,
1022                                      struct netlink_ext_ack *extack)
1023 {
1024         __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1025         struct vxlan_rdst *rd = NULL;
1026         struct vxlan_rdst oldrd;
1027         int notify = 0;
1028         int rc = 0;
1029         int err;
1030
1031         if (nhid && !rcu_access_pointer(f->nh)) {
1032                 NL_SET_ERR_MSG(extack,
1033                                "Cannot replace an existing non nexthop fdb with a nexthop");
1034                 return -EOPNOTSUPP;
1035         }
1036
1037         if (nhid && (flags & NLM_F_APPEND)) {
1038                 NL_SET_ERR_MSG(extack,
1039                                "Cannot append to a nexthop fdb");
1040                 return -EOPNOTSUPP;
1041         }
1042
1043         /* Do not allow an externally learned entry to take over an entry added
1044          * by the user.
1045          */
1046         if (!(fdb_flags & NTF_EXT_LEARNED) ||
1047             !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1048                 if (f->state != state) {
1049                         f->state = state;
1050                         f->updated = jiffies;
1051                         notify = 1;
1052                 }
1053                 if (f->flags != fdb_flags) {
1054                         f->flags = fdb_flags;
1055                         f->updated = jiffies;
1056                         notify = 1;
1057                 }
1058         }
1059
1060         if ((flags & NLM_F_REPLACE)) {
1061                 /* Only change unicasts */
1062                 if (!(is_multicast_ether_addr(f->eth_addr) ||
1063                       is_zero_ether_addr(f->eth_addr))) {
1064                         if (nhid) {
1065                                 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1066                                 if (rc < 0)
1067                                         return rc;
1068                         } else {
1069                                 rc = vxlan_fdb_replace(f, ip, port, vni,
1070                                                        ifindex, &oldrd);
1071                         }
1072                         notify |= rc;
1073                 } else {
1074                         NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
1075                         return -EOPNOTSUPP;
1076                 }
1077         }
1078         if ((flags & NLM_F_APPEND) &&
1079             (is_multicast_ether_addr(f->eth_addr) ||
1080              is_zero_ether_addr(f->eth_addr))) {
1081                 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1082
1083                 if (rc < 0)
1084                         return rc;
1085                 notify |= rc;
1086         }
1087
1088         if (ndm_flags & NTF_USE)
1089                 f->used = jiffies;
1090
1091         if (notify) {
1092                 if (rd == NULL)
1093                         rd = first_remote_rtnl(f);
1094
1095                 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
1096                                        swdev_notify, extack);
1097                 if (err)
1098                         goto err_notify;
1099         }
1100
1101         return 0;
1102
1103 err_notify:
1104         if (nhid)
1105                 return err;
1106         if ((flags & NLM_F_REPLACE) && rc)
1107                 *rd = oldrd;
1108         else if ((flags & NLM_F_APPEND) && rc) {
1109                 list_del_rcu(&rd->list);
1110                 call_rcu(&rd->rcu, vxlan_dst_free);
1111         }
1112         return err;
1113 }
1114
1115 static int vxlan_fdb_update_create(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         __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1124         struct vxlan_fdb *f;
1125         int rc;
1126
1127         /* Disallow replace to add a multicast entry */
1128         if ((flags & NLM_F_REPLACE) &&
1129             (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1130                 return -EOPNOTSUPP;
1131
1132         netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1133         rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1134                               vni, ifindex, fdb_flags, nhid, &f, extack);
1135         if (rc < 0)
1136                 return rc;
1137
1138         vxlan_fdb_insert(vxlan, mac, src_vni, f);
1139         rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1140                               swdev_notify, extack);
1141         if (rc)
1142                 goto err_notify;
1143
1144         return 0;
1145
1146 err_notify:
1147         vxlan_fdb_destroy(vxlan, f, false, false);
1148         return rc;
1149 }
1150
1151 /* Add new entry to forwarding table -- assumes lock held */
1152 static int vxlan_fdb_update(struct vxlan_dev *vxlan,
1153                             const u8 *mac, union vxlan_addr *ip,
1154                             __u16 state, __u16 flags,
1155                             __be16 port, __be32 src_vni, __be32 vni,
1156                             __u32 ifindex, __u16 ndm_flags, u32 nhid,
1157                             bool swdev_notify,
1158                             struct netlink_ext_ack *extack)
1159 {
1160         struct vxlan_fdb *f;
1161
1162         f = __vxlan_find_mac(vxlan, mac, src_vni);
1163         if (f) {
1164                 if (flags & NLM_F_EXCL) {
1165                         netdev_dbg(vxlan->dev,
1166                                    "lost race to create %pM\n", mac);
1167                         return -EEXIST;
1168                 }
1169
1170                 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1171                                                  vni, ifindex, ndm_flags, f,
1172                                                  nhid, swdev_notify, extack);
1173         } else {
1174                 if (!(flags & NLM_F_CREATE))
1175                         return -ENOENT;
1176
1177                 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1178                                                port, src_vni, vni, ifindex,
1179                                                ndm_flags, nhid, swdev_notify,
1180                                                extack);
1181         }
1182 }
1183
1184 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1185                                   struct vxlan_rdst *rd, bool swdev_notify)
1186 {
1187         list_del_rcu(&rd->list);
1188         vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1189         call_rcu(&rd->rcu, vxlan_dst_free);
1190 }
1191
1192 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1193                            union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1194                            __be32 *vni, u32 *ifindex, u32 *nhid)
1195 {
1196         struct net *net = dev_net(vxlan->dev);
1197         int err;
1198
1199         if (tb[NDA_NH_ID] && (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] ||
1200             tb[NDA_PORT]))
1201                 return -EINVAL;
1202
1203         if (tb[NDA_DST]) {
1204                 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1205                 if (err)
1206                         return err;
1207         } else {
1208                 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1209
1210                 if (remote->sa.sa_family == AF_INET) {
1211                         ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1212                         ip->sa.sa_family = AF_INET;
1213 #if IS_ENABLED(CONFIG_IPV6)
1214                 } else {
1215                         ip->sin6.sin6_addr = in6addr_any;
1216                         ip->sa.sa_family = AF_INET6;
1217 #endif
1218                 }
1219         }
1220
1221         if (tb[NDA_PORT]) {
1222                 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
1223                         return -EINVAL;
1224                 *port = nla_get_be16(tb[NDA_PORT]);
1225         } else {
1226                 *port = vxlan->cfg.dst_port;
1227         }
1228
1229         if (tb[NDA_VNI]) {
1230                 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
1231                         return -EINVAL;
1232                 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1233         } else {
1234                 *vni = vxlan->default_dst.remote_vni;
1235         }
1236
1237         if (tb[NDA_SRC_VNI]) {
1238                 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
1239                         return -EINVAL;
1240                 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1241         } else {
1242                 *src_vni = vxlan->default_dst.remote_vni;
1243         }
1244
1245         if (tb[NDA_IFINDEX]) {
1246                 struct net_device *tdev;
1247
1248                 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
1249                         return -EINVAL;
1250                 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1251                 tdev = __dev_get_by_index(net, *ifindex);
1252                 if (!tdev)
1253                         return -EADDRNOTAVAIL;
1254         } else {
1255                 *ifindex = 0;
1256         }
1257
1258         if (tb[NDA_NH_ID])
1259                 *nhid = nla_get_u32(tb[NDA_NH_ID]);
1260         else
1261                 *nhid = 0;
1262
1263         return 0;
1264 }
1265
1266 /* Add static entry (via netlink) */
1267 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1268                          struct net_device *dev,
1269                          const unsigned char *addr, u16 vid, u16 flags,
1270                          struct netlink_ext_ack *extack)
1271 {
1272         struct vxlan_dev *vxlan = netdev_priv(dev);
1273         /* struct net *net = dev_net(vxlan->dev); */
1274         union vxlan_addr ip;
1275         __be16 port;
1276         __be32 src_vni, vni;
1277         u32 ifindex, nhid;
1278         u32 hash_index;
1279         int err;
1280
1281         if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1282                 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1283                         ndm->ndm_state);
1284                 return -EINVAL;
1285         }
1286
1287         if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
1288                 return -EINVAL;
1289
1290         err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1291                               &nhid);
1292         if (err)
1293                 return err;
1294
1295         if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1296                 return -EAFNOSUPPORT;
1297
1298         hash_index = fdb_head_index(vxlan, addr, src_vni);
1299         spin_lock_bh(&vxlan->hash_lock[hash_index]);
1300         err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1301                                port, src_vni, vni, ifindex,
1302                                ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1303                                nhid, true, extack);
1304         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1305
1306         return err;
1307 }
1308
1309 static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1310                               const unsigned char *addr, union vxlan_addr ip,
1311                               __be16 port, __be32 src_vni, __be32 vni,
1312                               u32 ifindex, bool swdev_notify)
1313 {
1314         struct vxlan_rdst *rd = NULL;
1315         struct vxlan_fdb *f;
1316         int err = -ENOENT;
1317
1318         f = vxlan_find_mac(vxlan, addr, src_vni);
1319         if (!f)
1320                 return err;
1321
1322         if (!vxlan_addr_any(&ip)) {
1323                 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1324                 if (!rd)
1325                         goto out;
1326         }
1327
1328         /* remove a destination if it's not the only one on the list,
1329          * otherwise destroy the fdb entry
1330          */
1331         if (rd && !list_is_singular(&f->remotes)) {
1332                 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1333                 goto out;
1334         }
1335
1336         vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1337
1338 out:
1339         return 0;
1340 }
1341
1342 /* Delete entry (via netlink) */
1343 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1344                             struct net_device *dev,
1345                             const unsigned char *addr, u16 vid)
1346 {
1347         struct vxlan_dev *vxlan = netdev_priv(dev);
1348         union vxlan_addr ip;
1349         __be32 src_vni, vni;
1350         u32 ifindex, nhid;
1351         u32 hash_index;
1352         __be16 port;
1353         int err;
1354
1355         err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1356                               &nhid);
1357         if (err)
1358                 return err;
1359
1360         hash_index = fdb_head_index(vxlan, addr, src_vni);
1361         spin_lock_bh(&vxlan->hash_lock[hash_index]);
1362         err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1363                                  true);
1364         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1365
1366         return err;
1367 }
1368
1369 /* Dump forwarding table */
1370 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1371                           struct net_device *dev,
1372                           struct net_device *filter_dev, int *idx)
1373 {
1374         struct vxlan_dev *vxlan = netdev_priv(dev);
1375         unsigned int h;
1376         int err = 0;
1377
1378         for (h = 0; h < FDB_HASH_SIZE; ++h) {
1379                 struct vxlan_fdb *f;
1380
1381                 rcu_read_lock();
1382                 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1383                         struct vxlan_rdst *rd;
1384
1385                         if (rcu_access_pointer(f->nh)) {
1386                                 if (*idx < cb->args[2])
1387                                         goto skip_nh;
1388                                 err = vxlan_fdb_info(skb, vxlan, f,
1389                                                      NETLINK_CB(cb->skb).portid,
1390                                                      cb->nlh->nlmsg_seq,
1391                                                      RTM_NEWNEIGH,
1392                                                      NLM_F_MULTI, NULL);
1393                                 if (err < 0) {
1394                                         rcu_read_unlock();
1395                                         goto out;
1396                                 }
1397 skip_nh:
1398                                 *idx += 1;
1399                                 continue;
1400                         }
1401
1402                         list_for_each_entry_rcu(rd, &f->remotes, list) {
1403                                 if (*idx < cb->args[2])
1404                                         goto skip;
1405
1406                                 err = vxlan_fdb_info(skb, vxlan, f,
1407                                                      NETLINK_CB(cb->skb).portid,
1408                                                      cb->nlh->nlmsg_seq,
1409                                                      RTM_NEWNEIGH,
1410                                                      NLM_F_MULTI, rd);
1411                                 if (err < 0) {
1412                                         rcu_read_unlock();
1413                                         goto out;
1414                                 }
1415 skip:
1416                                 *idx += 1;
1417                         }
1418                 }
1419                 rcu_read_unlock();
1420         }
1421 out:
1422         return err;
1423 }
1424
1425 static int vxlan_fdb_get(struct sk_buff *skb,
1426                          struct nlattr *tb[],
1427                          struct net_device *dev,
1428                          const unsigned char *addr,
1429                          u16 vid, u32 portid, u32 seq,
1430                          struct netlink_ext_ack *extack)
1431 {
1432         struct vxlan_dev *vxlan = netdev_priv(dev);
1433         struct vxlan_fdb *f;
1434         __be32 vni;
1435         int err;
1436
1437         if (tb[NDA_VNI])
1438                 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1439         else
1440                 vni = vxlan->default_dst.remote_vni;
1441
1442         rcu_read_lock();
1443
1444         f = __vxlan_find_mac(vxlan, addr, vni);
1445         if (!f) {
1446                 NL_SET_ERR_MSG(extack, "Fdb entry not found");
1447                 err = -ENOENT;
1448                 goto errout;
1449         }
1450
1451         err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1452                              RTM_NEWNEIGH, 0, first_remote_rcu(f));
1453 errout:
1454         rcu_read_unlock();
1455         return err;
1456 }
1457
1458 /* Watch incoming packets to learn mapping between Ethernet address
1459  * and Tunnel endpoint.
1460  * Return true if packet is bogus and should be dropped.
1461  */
1462 static bool vxlan_snoop(struct net_device *dev,
1463                         union vxlan_addr *src_ip, const u8 *src_mac,
1464                         u32 src_ifindex, __be32 vni)
1465 {
1466         struct vxlan_dev *vxlan = netdev_priv(dev);
1467         struct vxlan_fdb *f;
1468         u32 ifindex = 0;
1469
1470 #if IS_ENABLED(CONFIG_IPV6)
1471         if (src_ip->sa.sa_family == AF_INET6 &&
1472             (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1473                 ifindex = src_ifindex;
1474 #endif
1475
1476         f = vxlan_find_mac(vxlan, src_mac, vni);
1477         if (likely(f)) {
1478                 struct vxlan_rdst *rdst = first_remote_rcu(f);
1479
1480                 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1481                            rdst->remote_ifindex == ifindex))
1482                         return false;
1483
1484                 /* Don't migrate static entries, drop packets */
1485                 if (f->state & (NUD_PERMANENT | NUD_NOARP))
1486                         return true;
1487
1488                 /* Don't override an fdb with nexthop with a learnt entry */
1489                 if (rcu_access_pointer(f->nh))
1490                         return true;
1491
1492                 if (net_ratelimit())
1493                         netdev_info(dev,
1494                                     "%pM migrated from %pIS to %pIS\n",
1495                                     src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1496
1497                 rdst->remote_ip = *src_ip;
1498                 f->updated = jiffies;
1499                 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1500         } else {
1501                 u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1502
1503                 /* learned new entry */
1504                 spin_lock(&vxlan->hash_lock[hash_index]);
1505
1506                 /* close off race between vxlan_flush and incoming packets */
1507                 if (netif_running(dev))
1508                         vxlan_fdb_update(vxlan, src_mac, src_ip,
1509                                          NUD_REACHABLE,
1510                                          NLM_F_EXCL|NLM_F_CREATE,
1511                                          vxlan->cfg.dst_port,
1512                                          vni,
1513                                          vxlan->default_dst.remote_vni,
1514                                          ifindex, NTF_SELF, 0, true, NULL);
1515                 spin_unlock(&vxlan->hash_lock[hash_index]);
1516         }
1517
1518         return false;
1519 }
1520
1521 /* See if multicast group is already in use by other ID */
1522 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1523 {
1524         struct vxlan_dev *vxlan;
1525         struct vxlan_sock *sock4;
1526 #if IS_ENABLED(CONFIG_IPV6)
1527         struct vxlan_sock *sock6;
1528 #endif
1529         unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1530
1531         sock4 = rtnl_dereference(dev->vn4_sock);
1532
1533         /* The vxlan_sock is only used by dev, leaving group has
1534          * no effect on other vxlan devices.
1535          */
1536         if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
1537                 return false;
1538 #if IS_ENABLED(CONFIG_IPV6)
1539         sock6 = rtnl_dereference(dev->vn6_sock);
1540         if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
1541                 return false;
1542 #endif
1543
1544         list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1545                 if (!netif_running(vxlan->dev) || vxlan == dev)
1546                         continue;
1547
1548                 if (family == AF_INET &&
1549                     rtnl_dereference(vxlan->vn4_sock) != sock4)
1550                         continue;
1551 #if IS_ENABLED(CONFIG_IPV6)
1552                 if (family == AF_INET6 &&
1553                     rtnl_dereference(vxlan->vn6_sock) != sock6)
1554                         continue;
1555 #endif
1556
1557                 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1558                                       &dev->default_dst.remote_ip))
1559                         continue;
1560
1561                 if (vxlan->default_dst.remote_ifindex !=
1562                     dev->default_dst.remote_ifindex)
1563                         continue;
1564
1565                 return true;
1566         }
1567
1568         return false;
1569 }
1570
1571 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1572 {
1573         struct vxlan_net *vn;
1574
1575         if (!vs)
1576                 return false;
1577         if (!refcount_dec_and_test(&vs->refcnt))
1578                 return false;
1579
1580         vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1581         spin_lock(&vn->sock_lock);
1582         hlist_del_rcu(&vs->hlist);
1583         udp_tunnel_notify_del_rx_port(vs->sock,
1584                                       (vs->flags & VXLAN_F_GPE) ?
1585                                       UDP_TUNNEL_TYPE_VXLAN_GPE :
1586                                       UDP_TUNNEL_TYPE_VXLAN);
1587         spin_unlock(&vn->sock_lock);
1588
1589         return true;
1590 }
1591
1592 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1593 {
1594         struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1595 #if IS_ENABLED(CONFIG_IPV6)
1596         struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1597
1598         RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1599 #endif
1600
1601         RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1602         synchronize_net();
1603
1604         vxlan_vs_del_dev(vxlan);
1605
1606         if (__vxlan_sock_release_prep(sock4)) {
1607                 udp_tunnel_sock_release(sock4->sock);
1608                 kfree(sock4);
1609         }
1610
1611 #if IS_ENABLED(CONFIG_IPV6)
1612         if (__vxlan_sock_release_prep(sock6)) {
1613                 udp_tunnel_sock_release(sock6->sock);
1614                 kfree(sock6);
1615         }
1616 #endif
1617 }
1618
1619 /* Update multicast group membership when first VNI on
1620  * multicast address is brought up
1621  */
1622 static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1623 {
1624         struct sock *sk;
1625         union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1626         int ifindex = vxlan->default_dst.remote_ifindex;
1627         int ret = -EINVAL;
1628
1629         if (ip->sa.sa_family == AF_INET) {
1630                 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1631                 struct ip_mreqn mreq = {
1632                         .imr_multiaddr.s_addr   = ip->sin.sin_addr.s_addr,
1633                         .imr_ifindex            = ifindex,
1634                 };
1635
1636                 sk = sock4->sock->sk;
1637                 lock_sock(sk);
1638                 ret = ip_mc_join_group(sk, &mreq);
1639                 release_sock(sk);
1640 #if IS_ENABLED(CONFIG_IPV6)
1641         } else {
1642                 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1643
1644                 sk = sock6->sock->sk;
1645                 lock_sock(sk);
1646                 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1647                                                    &ip->sin6.sin6_addr);
1648                 release_sock(sk);
1649 #endif
1650         }
1651
1652         return ret;
1653 }
1654
1655 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1656 static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1657 {
1658         struct sock *sk;
1659         union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1660         int ifindex = vxlan->default_dst.remote_ifindex;
1661         int ret = -EINVAL;
1662
1663         if (ip->sa.sa_family == AF_INET) {
1664                 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1665                 struct ip_mreqn mreq = {
1666                         .imr_multiaddr.s_addr   = ip->sin.sin_addr.s_addr,
1667                         .imr_ifindex            = ifindex,
1668                 };
1669
1670                 sk = sock4->sock->sk;
1671                 lock_sock(sk);
1672                 ret = ip_mc_leave_group(sk, &mreq);
1673                 release_sock(sk);
1674 #if IS_ENABLED(CONFIG_IPV6)
1675         } else {
1676                 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1677
1678                 sk = sock6->sock->sk;
1679                 lock_sock(sk);
1680                 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1681                                                    &ip->sin6.sin6_addr);
1682                 release_sock(sk);
1683 #endif
1684         }
1685
1686         return ret;
1687 }
1688
1689 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1690                           struct sk_buff *skb, u32 vxflags)
1691 {
1692         size_t start, offset;
1693
1694         if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1695                 goto out;
1696
1697         start = vxlan_rco_start(unparsed->vx_vni);
1698         offset = start + vxlan_rco_offset(unparsed->vx_vni);
1699
1700         if (!pskb_may_pull(skb, offset + sizeof(u16)))
1701                 return false;
1702
1703         skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1704                             !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1705 out:
1706         unparsed->vx_flags &= ~VXLAN_HF_RCO;
1707         unparsed->vx_vni &= VXLAN_VNI_MASK;
1708         return true;
1709 }
1710
1711 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1712                                 struct sk_buff *skb, u32 vxflags,
1713                                 struct vxlan_metadata *md)
1714 {
1715         struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1716         struct metadata_dst *tun_dst;
1717
1718         if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1719                 goto out;
1720
1721         md->gbp = ntohs(gbp->policy_id);
1722
1723         tun_dst = (struct metadata_dst *)skb_dst(skb);
1724         if (tun_dst) {
1725                 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1726                 tun_dst->u.tun_info.options_len = sizeof(*md);
1727         }
1728         if (gbp->dont_learn)
1729                 md->gbp |= VXLAN_GBP_DONT_LEARN;
1730
1731         if (gbp->policy_applied)
1732                 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1733
1734         /* In flow-based mode, GBP is carried in dst_metadata */
1735         if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1736                 skb->mark = md->gbp;
1737 out:
1738         unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1739 }
1740
1741 static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1742                                 __be16 *protocol,
1743                                 struct sk_buff *skb, u32 vxflags)
1744 {
1745         struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1746
1747         /* Need to have Next Protocol set for interfaces in GPE mode. */
1748         if (!gpe->np_applied)
1749                 return false;
1750         /* "The initial version is 0. If a receiver does not support the
1751          * version indicated it MUST drop the packet.
1752          */
1753         if (gpe->version != 0)
1754                 return false;
1755         /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1756          * processing MUST occur." However, we don't implement OAM
1757          * processing, thus drop the packet.
1758          */
1759         if (gpe->oam_flag)
1760                 return false;
1761
1762         *protocol = tun_p_to_eth_p(gpe->next_protocol);
1763         if (!*protocol)
1764                 return false;
1765
1766         unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1767         return true;
1768 }
1769
1770 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1771                           struct vxlan_sock *vs,
1772                           struct sk_buff *skb, __be32 vni)
1773 {
1774         union vxlan_addr saddr;
1775         u32 ifindex = skb->dev->ifindex;
1776
1777         skb_reset_mac_header(skb);
1778         skb->protocol = eth_type_trans(skb, vxlan->dev);
1779         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1780
1781         /* Ignore packet loops (and multicast echo) */
1782         if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1783                 return false;
1784
1785         /* Get address from the outer IP header */
1786         if (vxlan_get_sk_family(vs) == AF_INET) {
1787                 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1788                 saddr.sa.sa_family = AF_INET;
1789 #if IS_ENABLED(CONFIG_IPV6)
1790         } else {
1791                 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1792                 saddr.sa.sa_family = AF_INET6;
1793 #endif
1794         }
1795
1796         if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1797             vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1798                 return false;
1799
1800         return true;
1801 }
1802
1803 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1804                                   struct sk_buff *skb)
1805 {
1806         int err = 0;
1807
1808         if (vxlan_get_sk_family(vs) == AF_INET)
1809                 err = IP_ECN_decapsulate(oiph, skb);
1810 #if IS_ENABLED(CONFIG_IPV6)
1811         else
1812                 err = IP6_ECN_decapsulate(oiph, skb);
1813 #endif
1814
1815         if (unlikely(err) && log_ecn_error) {
1816                 if (vxlan_get_sk_family(vs) == AF_INET)
1817                         net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1818                                              &((struct iphdr *)oiph)->saddr,
1819                                              ((struct iphdr *)oiph)->tos);
1820                 else
1821                         net_info_ratelimited("non-ECT from %pI6\n",
1822                                              &((struct ipv6hdr *)oiph)->saddr);
1823         }
1824         return err <= 1;
1825 }
1826
1827 /* Callback from net/ipv4/udp.c to receive packets */
1828 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1829 {
1830         struct vxlan_dev *vxlan;
1831         struct vxlan_sock *vs;
1832         struct vxlanhdr unparsed;
1833         struct vxlan_metadata _md;
1834         struct vxlan_metadata *md = &_md;
1835         __be16 protocol = htons(ETH_P_TEB);
1836         bool raw_proto = false;
1837         void *oiph;
1838         __be32 vni = 0;
1839
1840         /* Need UDP and VXLAN header to be present */
1841         if (!pskb_may_pull(skb, VXLAN_HLEN))
1842                 goto drop;
1843
1844         unparsed = *vxlan_hdr(skb);
1845         /* VNI flag always required to be set */
1846         if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1847                 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1848                            ntohl(vxlan_hdr(skb)->vx_flags),
1849                            ntohl(vxlan_hdr(skb)->vx_vni));
1850                 /* Return non vxlan pkt */
1851                 goto drop;
1852         }
1853         unparsed.vx_flags &= ~VXLAN_HF_VNI;
1854         unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1855
1856         vs = rcu_dereference_sk_user_data(sk);
1857         if (!vs)
1858                 goto drop;
1859
1860         vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1861
1862         vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1863         if (!vxlan)
1864                 goto drop;
1865
1866         /* For backwards compatibility, only allow reserved fields to be
1867          * used by VXLAN extensions if explicitly requested.
1868          */
1869         if (vs->flags & VXLAN_F_GPE) {
1870                 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1871                         goto drop;
1872                 raw_proto = true;
1873         }
1874
1875         if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1876                                    !net_eq(vxlan->net, dev_net(vxlan->dev))))
1877                 goto drop;
1878
1879         if (vs->flags & VXLAN_F_REMCSUM_RX)
1880                 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags)))
1881                         goto drop;
1882
1883         if (vxlan_collect_metadata(vs)) {
1884                 struct metadata_dst *tun_dst;
1885
1886                 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1887                                          key32_to_tunnel_id(vni), sizeof(*md));
1888
1889                 if (!tun_dst)
1890                         goto drop;
1891
1892                 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1893
1894                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1895         } else {
1896                 memset(md, 0, sizeof(*md));
1897         }
1898
1899         if (vs->flags & VXLAN_F_GBP)
1900                 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1901         /* Note that GBP and GPE can never be active together. This is
1902          * ensured in vxlan_dev_configure.
1903          */
1904
1905         if (unparsed.vx_flags || unparsed.vx_vni) {
1906                 /* If there are any unprocessed flags remaining treat
1907                  * this as a malformed packet. This behavior diverges from
1908                  * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1909                  * in reserved fields are to be ignored. The approach here
1910                  * maintains compatibility with previous stack code, and also
1911                  * is more robust and provides a little more security in
1912                  * adding extensions to VXLAN.
1913                  */
1914                 goto drop;
1915         }
1916
1917         if (!raw_proto) {
1918                 if (!vxlan_set_mac(vxlan, vs, skb, vni))
1919                         goto drop;
1920         } else {
1921                 skb_reset_mac_header(skb);
1922                 skb->dev = vxlan->dev;
1923                 skb->pkt_type = PACKET_HOST;
1924         }
1925
1926         oiph = skb_network_header(skb);
1927         skb_reset_network_header(skb);
1928
1929         if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1930                 ++vxlan->dev->stats.rx_frame_errors;
1931                 ++vxlan->dev->stats.rx_errors;
1932                 goto drop;
1933         }
1934
1935         rcu_read_lock();
1936
1937         if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1938                 rcu_read_unlock();
1939                 atomic_long_inc(&vxlan->dev->rx_dropped);
1940                 goto drop;
1941         }
1942
1943         dev_sw_netstats_rx_add(vxlan->dev, skb->len);
1944         gro_cells_receive(&vxlan->gro_cells, skb);
1945
1946         rcu_read_unlock();
1947
1948         return 0;
1949
1950 drop:
1951         /* Consume bad packet */
1952         kfree_skb(skb);
1953         return 0;
1954 }
1955
1956 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1957 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1958 {
1959         struct vxlan_dev *vxlan;
1960         struct vxlan_sock *vs;
1961         struct vxlanhdr *hdr;
1962         __be32 vni;
1963
1964         if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1965                 return -EINVAL;
1966
1967         hdr = vxlan_hdr(skb);
1968
1969         if (!(hdr->vx_flags & VXLAN_HF_VNI))
1970                 return -EINVAL;
1971
1972         vs = rcu_dereference_sk_user_data(sk);
1973         if (!vs)
1974                 return -ENOENT;
1975
1976         vni = vxlan_vni(hdr->vx_vni);
1977         vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1978         if (!vxlan)
1979                 return -ENOENT;
1980
1981         return 0;
1982 }
1983
1984 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1985 {
1986         struct vxlan_dev *vxlan = netdev_priv(dev);
1987         struct arphdr *parp;
1988         u8 *arpptr, *sha;
1989         __be32 sip, tip;
1990         struct neighbour *n;
1991
1992         if (dev->flags & IFF_NOARP)
1993                 goto out;
1994
1995         if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1996                 dev->stats.tx_dropped++;
1997                 goto out;
1998         }
1999         parp = arp_hdr(skb);
2000
2001         if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
2002              parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
2003             parp->ar_pro != htons(ETH_P_IP) ||
2004             parp->ar_op != htons(ARPOP_REQUEST) ||
2005             parp->ar_hln != dev->addr_len ||
2006             parp->ar_pln != 4)
2007                 goto out;
2008         arpptr = (u8 *)parp + sizeof(struct arphdr);
2009         sha = arpptr;
2010         arpptr += dev->addr_len;        /* sha */
2011         memcpy(&sip, arpptr, sizeof(sip));
2012         arpptr += sizeof(sip);
2013         arpptr += dev->addr_len;        /* tha */
2014         memcpy(&tip, arpptr, sizeof(tip));
2015
2016         if (ipv4_is_loopback(tip) ||
2017             ipv4_is_multicast(tip))
2018                 goto out;
2019
2020         n = neigh_lookup(&arp_tbl, &tip, dev);
2021
2022         if (n) {
2023                 struct vxlan_fdb *f;
2024                 struct sk_buff  *reply;
2025
2026                 if (!(n->nud_state & NUD_CONNECTED)) {
2027                         neigh_release(n);
2028                         goto out;
2029                 }
2030
2031                 f = vxlan_find_mac(vxlan, n->ha, vni);
2032                 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2033                         /* bridge-local neighbor */
2034                         neigh_release(n);
2035                         goto out;
2036                 }
2037
2038                 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
2039                                 n->ha, sha);
2040
2041                 neigh_release(n);
2042
2043                 if (reply == NULL)
2044                         goto out;
2045
2046                 skb_reset_mac_header(reply);
2047                 __skb_pull(reply, skb_network_offset(reply));
2048                 reply->ip_summed = CHECKSUM_UNNECESSARY;
2049                 reply->pkt_type = PACKET_HOST;
2050
2051                 if (netif_rx_ni(reply) == NET_RX_DROP)
2052                         dev->stats.rx_dropped++;
2053         } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2054                 union vxlan_addr ipa = {
2055                         .sin.sin_addr.s_addr = tip,
2056                         .sin.sin_family = AF_INET,
2057                 };
2058
2059                 vxlan_ip_miss(dev, &ipa);
2060         }
2061 out:
2062         consume_skb(skb);
2063         return NETDEV_TX_OK;
2064 }
2065
2066 #if IS_ENABLED(CONFIG_IPV6)
2067 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
2068         struct neighbour *n, bool isrouter)
2069 {
2070         struct net_device *dev = request->dev;
2071         struct sk_buff *reply;
2072         struct nd_msg *ns, *na;
2073         struct ipv6hdr *pip6;
2074         u8 *daddr;
2075         int na_olen = 8; /* opt hdr + ETH_ALEN for target */
2076         int ns_olen;
2077         int i, len;
2078
2079         if (dev == NULL || !pskb_may_pull(request, request->len))
2080                 return NULL;
2081
2082         len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
2083                 sizeof(*na) + na_olen + dev->needed_tailroom;
2084         reply = alloc_skb(len, GFP_ATOMIC);
2085         if (reply == NULL)
2086                 return NULL;
2087
2088         reply->protocol = htons(ETH_P_IPV6);
2089         reply->dev = dev;
2090         skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
2091         skb_push(reply, sizeof(struct ethhdr));
2092         skb_reset_mac_header(reply);
2093
2094         ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
2095
2096         daddr = eth_hdr(request)->h_source;
2097         ns_olen = request->len - skb_network_offset(request) -
2098                 sizeof(struct ipv6hdr) - sizeof(*ns);
2099         for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
2100                 if (!ns->opt[i + 1]) {
2101                         kfree_skb(reply);
2102                         return NULL;
2103                 }
2104                 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
2105                         daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
2106                         break;
2107                 }
2108         }
2109
2110         /* Ethernet header */
2111         ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
2112         ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
2113         eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
2114         reply->protocol = htons(ETH_P_IPV6);
2115
2116         skb_pull(reply, sizeof(struct ethhdr));
2117         skb_reset_network_header(reply);
2118         skb_put(reply, sizeof(struct ipv6hdr));
2119
2120         /* IPv6 header */
2121
2122         pip6 = ipv6_hdr(reply);
2123         memset(pip6, 0, sizeof(struct ipv6hdr));
2124         pip6->version = 6;
2125         pip6->priority = ipv6_hdr(request)->priority;
2126         pip6->nexthdr = IPPROTO_ICMPV6;
2127         pip6->hop_limit = 255;
2128         pip6->daddr = ipv6_hdr(request)->saddr;
2129         pip6->saddr = *(struct in6_addr *)n->primary_key;
2130
2131         skb_pull(reply, sizeof(struct ipv6hdr));
2132         skb_reset_transport_header(reply);
2133
2134         /* Neighbor Advertisement */
2135         na = skb_put_zero(reply, sizeof(*na) + na_olen);
2136         na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
2137         na->icmph.icmp6_router = isrouter;
2138         na->icmph.icmp6_override = 1;
2139         na->icmph.icmp6_solicited = 1;
2140         na->target = ns->target;
2141         ether_addr_copy(&na->opt[2], n->ha);
2142         na->opt[0] = ND_OPT_TARGET_LL_ADDR;
2143         na->opt[1] = na_olen >> 3;
2144
2145         na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
2146                 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
2147                 csum_partial(na, sizeof(*na)+na_olen, 0));
2148
2149         pip6->payload_len = htons(sizeof(*na)+na_olen);
2150
2151         skb_push(reply, sizeof(struct ipv6hdr));
2152
2153         reply->ip_summed = CHECKSUM_UNNECESSARY;
2154
2155         return reply;
2156 }
2157
2158 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2159 {
2160         struct vxlan_dev *vxlan = netdev_priv(dev);
2161         const struct in6_addr *daddr;
2162         const struct ipv6hdr *iphdr;
2163         struct inet6_dev *in6_dev;
2164         struct neighbour *n;
2165         struct nd_msg *msg;
2166
2167         in6_dev = __in6_dev_get(dev);
2168         if (!in6_dev)
2169                 goto out;
2170
2171         iphdr = ipv6_hdr(skb);
2172         daddr = &iphdr->daddr;
2173         msg = (struct nd_msg *)(iphdr + 1);
2174
2175         if (ipv6_addr_loopback(daddr) ||
2176             ipv6_addr_is_multicast(&msg->target))
2177                 goto out;
2178
2179         n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2180
2181         if (n) {
2182                 struct vxlan_fdb *f;
2183                 struct sk_buff *reply;
2184
2185                 if (!(n->nud_state & NUD_CONNECTED)) {
2186                         neigh_release(n);
2187                         goto out;
2188                 }
2189
2190                 f = vxlan_find_mac(vxlan, n->ha, vni);
2191                 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2192                         /* bridge-local neighbor */
2193                         neigh_release(n);
2194                         goto out;
2195                 }
2196
2197                 reply = vxlan_na_create(skb, n,
2198                                         !!(f ? f->flags & NTF_ROUTER : 0));
2199
2200                 neigh_release(n);
2201
2202                 if (reply == NULL)
2203                         goto out;
2204
2205                 if (netif_rx_ni(reply) == NET_RX_DROP)
2206                         dev->stats.rx_dropped++;
2207
2208         } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2209                 union vxlan_addr ipa = {
2210                         .sin6.sin6_addr = msg->target,
2211                         .sin6.sin6_family = AF_INET6,
2212                 };
2213
2214                 vxlan_ip_miss(dev, &ipa);
2215         }
2216
2217 out:
2218         consume_skb(skb);
2219         return NETDEV_TX_OK;
2220 }
2221 #endif
2222
2223 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2224 {
2225         struct vxlan_dev *vxlan = netdev_priv(dev);
2226         struct neighbour *n;
2227
2228         if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2229                 return false;
2230
2231         n = NULL;
2232         switch (ntohs(eth_hdr(skb)->h_proto)) {
2233         case ETH_P_IP:
2234         {
2235                 struct iphdr *pip;
2236
2237                 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2238                         return false;
2239                 pip = ip_hdr(skb);
2240                 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2241                 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2242                         union vxlan_addr ipa = {
2243                                 .sin.sin_addr.s_addr = pip->daddr,
2244                                 .sin.sin_family = AF_INET,
2245                         };
2246
2247                         vxlan_ip_miss(dev, &ipa);
2248                         return false;
2249                 }
2250
2251                 break;
2252         }
2253 #if IS_ENABLED(CONFIG_IPV6)
2254         case ETH_P_IPV6:
2255         {
2256                 struct ipv6hdr *pip6;
2257
2258                 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2259                         return false;
2260                 pip6 = ipv6_hdr(skb);
2261                 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2262                 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2263                         union vxlan_addr ipa = {
2264                                 .sin6.sin6_addr = pip6->daddr,
2265                                 .sin6.sin6_family = AF_INET6,
2266                         };
2267
2268                         vxlan_ip_miss(dev, &ipa);
2269                         return false;
2270                 }
2271
2272                 break;
2273         }
2274 #endif
2275         default:
2276                 return false;
2277         }
2278
2279         if (n) {
2280                 bool diff;
2281
2282                 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2283                 if (diff) {
2284                         memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2285                                 dev->addr_len);
2286                         memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2287                 }
2288                 neigh_release(n);
2289                 return diff;
2290         }
2291
2292         return false;
2293 }
2294
2295 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
2296                                 struct vxlan_metadata *md)
2297 {
2298         struct vxlanhdr_gbp *gbp;
2299
2300         if (!md->gbp)
2301                 return;
2302
2303         gbp = (struct vxlanhdr_gbp *)vxh;
2304         vxh->vx_flags |= VXLAN_HF_GBP;
2305
2306         if (md->gbp & VXLAN_GBP_DONT_LEARN)
2307                 gbp->dont_learn = 1;
2308
2309         if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
2310                 gbp->policy_applied = 1;
2311
2312         gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
2313 }
2314
2315 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
2316                                __be16 protocol)
2317 {
2318         struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2319
2320         gpe->np_applied = 1;
2321         gpe->next_protocol = tun_p_from_eth_p(protocol);
2322         if (!gpe->next_protocol)
2323                 return -EPFNOSUPPORT;
2324         return 0;
2325 }
2326
2327 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2328                            int iphdr_len, __be32 vni,
2329                            struct vxlan_metadata *md, u32 vxflags,
2330                            bool udp_sum)
2331 {
2332         struct vxlanhdr *vxh;
2333         int min_headroom;
2334         int err;
2335         int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2336         __be16 inner_protocol = htons(ETH_P_TEB);
2337
2338         if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2339             skb->ip_summed == CHECKSUM_PARTIAL) {
2340                 int csum_start = skb_checksum_start_offset(skb);
2341
2342                 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2343                     !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2344                     (skb->csum_offset == offsetof(struct udphdr, check) ||
2345                      skb->csum_offset == offsetof(struct tcphdr, check)))
2346                         type |= SKB_GSO_TUNNEL_REMCSUM;
2347         }
2348
2349         min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2350                         + VXLAN_HLEN + iphdr_len;
2351
2352         /* Need space for new headers (invalidates iph ptr) */
2353         err = skb_cow_head(skb, min_headroom);
2354         if (unlikely(err))
2355                 return err;
2356
2357         err = iptunnel_handle_offloads(skb, type);
2358         if (err)
2359                 return err;
2360
2361         vxh = __skb_push(skb, sizeof(*vxh));
2362         vxh->vx_flags = VXLAN_HF_VNI;
2363         vxh->vx_vni = vxlan_vni_field(vni);
2364
2365         if (type & SKB_GSO_TUNNEL_REMCSUM) {
2366                 unsigned int start;
2367
2368                 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2369                 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2370                 vxh->vx_flags |= VXLAN_HF_RCO;
2371
2372                 if (!skb_is_gso(skb)) {
2373                         skb->ip_summed = CHECKSUM_NONE;
2374                         skb->encapsulation = 0;
2375                 }
2376         }
2377
2378         if (vxflags & VXLAN_F_GBP)
2379                 vxlan_build_gbp_hdr(vxh, vxflags, md);
2380         if (vxflags & VXLAN_F_GPE) {
2381                 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2382                 if (err < 0)
2383                         return err;
2384                 inner_protocol = skb->protocol;
2385         }
2386
2387         skb_set_inner_protocol(skb, inner_protocol);
2388         return 0;
2389 }
2390
2391 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2392                                       struct vxlan_sock *sock4,
2393                                       struct sk_buff *skb, int oif, u8 tos,
2394                                       __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
2395                                       struct dst_cache *dst_cache,
2396                                       const struct ip_tunnel_info *info)
2397 {
2398         bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2399         struct rtable *rt = NULL;
2400         struct flowi4 fl4;
2401
2402         if (!sock4)
2403                 return ERR_PTR(-EIO);
2404
2405         if (tos && !info)
2406                 use_cache = false;
2407         if (use_cache) {
2408                 rt = dst_cache_get_ip4(dst_cache, saddr);
2409                 if (rt)
2410                         return rt;
2411         }
2412
2413         memset(&fl4, 0, sizeof(fl4));
2414         fl4.flowi4_oif = oif;
2415         fl4.flowi4_tos = RT_TOS(tos);
2416         fl4.flowi4_mark = skb->mark;
2417         fl4.flowi4_proto = IPPROTO_UDP;
2418         fl4.daddr = daddr;
2419         fl4.saddr = *saddr;
2420         fl4.fl4_dport = dport;
2421         fl4.fl4_sport = sport;
2422
2423         rt = ip_route_output_key(vxlan->net, &fl4);
2424         if (!IS_ERR(rt)) {
2425                 if (rt->dst.dev == dev) {
2426                         netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2427                         ip_rt_put(rt);
2428                         return ERR_PTR(-ELOOP);
2429                 }
2430
2431                 *saddr = fl4.saddr;
2432                 if (use_cache)
2433                         dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2434         } else {
2435                 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2436                 return ERR_PTR(-ENETUNREACH);
2437         }
2438         return rt;
2439 }
2440
2441 #if IS_ENABLED(CONFIG_IPV6)
2442 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2443                                           struct net_device *dev,
2444                                           struct vxlan_sock *sock6,
2445                                           struct sk_buff *skb, int oif, u8 tos,
2446                                           __be32 label,
2447                                           const struct in6_addr *daddr,
2448                                           struct in6_addr *saddr,
2449                                           __be16 dport, __be16 sport,
2450                                           struct dst_cache *dst_cache,
2451                                           const struct ip_tunnel_info *info)
2452 {
2453         bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2454         struct dst_entry *ndst;
2455         struct flowi6 fl6;
2456
2457         if (!sock6)
2458                 return ERR_PTR(-EIO);
2459
2460         if (tos && !info)
2461                 use_cache = false;
2462         if (use_cache) {
2463                 ndst = dst_cache_get_ip6(dst_cache, saddr);
2464                 if (ndst)
2465                         return ndst;
2466         }
2467
2468         memset(&fl6, 0, sizeof(fl6));
2469         fl6.flowi6_oif = oif;
2470         fl6.daddr = *daddr;
2471         fl6.saddr = *saddr;
2472         fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
2473         fl6.flowi6_mark = skb->mark;
2474         fl6.flowi6_proto = IPPROTO_UDP;
2475         fl6.fl6_dport = dport;
2476         fl6.fl6_sport = sport;
2477
2478         ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
2479                                                &fl6, NULL);
2480         if (IS_ERR(ndst)) {
2481                 netdev_dbg(dev, "no route to %pI6\n", daddr);
2482                 return ERR_PTR(-ENETUNREACH);
2483         }
2484
2485         if (unlikely(ndst->dev == dev)) {
2486                 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2487                 dst_release(ndst);
2488                 return ERR_PTR(-ELOOP);
2489         }
2490
2491         *saddr = fl6.saddr;
2492         if (use_cache)
2493                 dst_cache_set_ip6(dst_cache, ndst, saddr);
2494         return ndst;
2495 }
2496 #endif
2497
2498 /* Bypass encapsulation if the destination is local */
2499 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2500                                struct vxlan_dev *dst_vxlan, __be32 vni,
2501                                bool snoop)
2502 {
2503         struct pcpu_sw_netstats *tx_stats, *rx_stats;
2504         union vxlan_addr loopback;
2505         union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2506         struct net_device *dev;
2507         int len = skb->len;
2508
2509         tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2510         rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2511         skb->pkt_type = PACKET_HOST;
2512         skb->encapsulation = 0;
2513         skb->dev = dst_vxlan->dev;
2514         __skb_pull(skb, skb_network_offset(skb));
2515
2516         if (remote_ip->sa.sa_family == AF_INET) {
2517                 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2518                 loopback.sa.sa_family =  AF_INET;
2519 #if IS_ENABLED(CONFIG_IPV6)
2520         } else {
2521                 loopback.sin6.sin6_addr = in6addr_loopback;
2522                 loopback.sa.sa_family =  AF_INET6;
2523 #endif
2524         }
2525
2526         rcu_read_lock();
2527         dev = skb->dev;
2528         if (unlikely(!(dev->flags & IFF_UP))) {
2529                 kfree_skb(skb);
2530                 goto drop;
2531         }
2532
2533         if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
2534                 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2535
2536         u64_stats_update_begin(&tx_stats->syncp);
2537         tx_stats->tx_packets++;
2538         tx_stats->tx_bytes += len;
2539         u64_stats_update_end(&tx_stats->syncp);
2540
2541         if (netif_rx(skb) == NET_RX_SUCCESS) {
2542                 u64_stats_update_begin(&rx_stats->syncp);
2543                 rx_stats->rx_packets++;
2544                 rx_stats->rx_bytes += len;
2545                 u64_stats_update_end(&rx_stats->syncp);
2546         } else {
2547 drop:
2548                 dev->stats.rx_dropped++;
2549         }
2550         rcu_read_unlock();
2551 }
2552
2553 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2554                                  struct vxlan_dev *vxlan,
2555                                  union vxlan_addr *daddr,
2556                                  __be16 dst_port, int dst_ifindex, __be32 vni,
2557                                  struct dst_entry *dst,
2558                                  u32 rt_flags)
2559 {
2560 #if IS_ENABLED(CONFIG_IPV6)
2561         /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2562          * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2563          * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2564          */
2565         BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2566 #endif
2567         /* Bypass encapsulation if the destination is local */
2568         if (rt_flags & RTCF_LOCAL &&
2569             !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2570                 struct vxlan_dev *dst_vxlan;
2571
2572                 dst_release(dst);
2573                 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2574                                            daddr->sa.sa_family, dst_port,
2575                                            vxlan->cfg.flags);
2576                 if (!dst_vxlan) {
2577                         dev->stats.tx_errors++;
2578                         kfree_skb(skb);
2579
2580                         return -ENOENT;
2581                 }
2582                 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
2583                 return 1;
2584         }
2585
2586         return 0;
2587 }
2588
2589 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2590                            __be32 default_vni, struct vxlan_rdst *rdst,
2591                            bool did_rsc)
2592 {
2593         struct dst_cache *dst_cache;
2594         struct ip_tunnel_info *info;
2595         struct vxlan_dev *vxlan = netdev_priv(dev);
2596         const struct iphdr *old_iph = ip_hdr(skb);
2597         union vxlan_addr *dst;
2598         union vxlan_addr remote_ip, local_ip;
2599         struct vxlan_metadata _md;
2600         struct vxlan_metadata *md = &_md;
2601         __be16 src_port = 0, dst_port;
2602         struct dst_entry *ndst = NULL;
2603         __be32 vni, label;
2604         __u8 tos, ttl;
2605         int ifindex;
2606         int err;
2607         u32 flags = vxlan->cfg.flags;
2608         bool udp_sum = false;
2609         bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2610
2611         info = skb_tunnel_info(skb);
2612
2613         if (rdst) {
2614                 dst = &rdst->remote_ip;
2615                 if (vxlan_addr_any(dst)) {
2616                         if (did_rsc) {
2617                                 /* short-circuited back to local bridge */
2618                                 vxlan_encap_bypass(skb, vxlan, vxlan,
2619                                                    default_vni, true);
2620                                 return;
2621                         }
2622                         goto drop;
2623                 }
2624
2625                 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2626                 vni = (rdst->remote_vni) ? : default_vni;
2627                 ifindex = rdst->remote_ifindex;
2628                 local_ip = vxlan->cfg.saddr;
2629                 dst_cache = &rdst->dst_cache;
2630                 md->gbp = skb->mark;
2631                 if (flags & VXLAN_F_TTL_INHERIT) {
2632                         ttl = ip_tunnel_get_ttl(old_iph, skb);
2633                 } else {
2634                         ttl = vxlan->cfg.ttl;
2635                         if (!ttl && vxlan_addr_multicast(dst))
2636                                 ttl = 1;
2637                 }
2638
2639                 tos = vxlan->cfg.tos;
2640                 if (tos == 1)
2641                         tos = ip_tunnel_get_dsfield(old_iph, skb);
2642
2643                 if (dst->sa.sa_family == AF_INET)
2644                         udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2645                 else
2646                         udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2647                 label = vxlan->cfg.label;
2648         } else {
2649                 if (!info) {
2650                         WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2651                                   dev->name);
2652                         goto drop;
2653                 }
2654                 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2655                 if (remote_ip.sa.sa_family == AF_INET) {
2656                         remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2657                         local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2658                 } else {
2659                         remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2660                         local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2661                 }
2662                 dst = &remote_ip;
2663                 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2664                 vni = tunnel_id_to_key32(info->key.tun_id);
2665                 ifindex = 0;
2666                 dst_cache = &info->dst_cache;
2667                 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2668                         if (info->options_len < sizeof(*md))
2669                                 goto drop;
2670                         md = ip_tunnel_info_opts(info);
2671                 }
2672                 ttl = info->key.ttl;
2673                 tos = info->key.tos;
2674                 label = info->key.label;
2675                 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2676         }
2677         src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2678                                      vxlan->cfg.port_max, true);
2679
2680         rcu_read_lock();
2681         if (dst->sa.sa_family == AF_INET) {
2682                 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2683                 struct rtable *rt;
2684                 __be16 df = 0;
2685
2686                 if (!ifindex)
2687                         ifindex = sock4->sock->sk->sk_bound_dev_if;
2688
2689                 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2690                                      dst->sin.sin_addr.s_addr,
2691                                      &local_ip.sin.sin_addr.s_addr,
2692                                      dst_port, src_port,
2693                                      dst_cache, info);
2694                 if (IS_ERR(rt)) {
2695                         err = PTR_ERR(rt);
2696                         goto tx_error;
2697                 }
2698
2699                 if (!info) {
2700                         /* Bypass encapsulation if the destination is local */
2701                         err = encap_bypass_if_local(skb, dev, vxlan, dst,
2702                                                     dst_port, ifindex, vni,
2703                                                     &rt->dst, rt->rt_flags);
2704                         if (err)
2705                                 goto out_unlock;
2706
2707                         if (vxlan->cfg.df == VXLAN_DF_SET) {
2708                                 df = htons(IP_DF);
2709                         } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2710                                 struct ethhdr *eth = eth_hdr(skb);
2711
2712                                 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2713                                     (ntohs(eth->h_proto) == ETH_P_IP &&
2714                                      old_iph->frag_off & htons(IP_DF)))
2715                                         df = htons(IP_DF);
2716                         }
2717                 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2718                         df = htons(IP_DF);
2719                 }
2720
2721                 ndst = &rt->dst;
2722                 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM,
2723                                             netif_is_any_bridge_port(dev));
2724                 if (err < 0) {
2725                         goto tx_error;
2726                 } else if (err) {
2727                         if (info) {
2728                                 struct ip_tunnel_info *unclone;
2729                                 struct in_addr src, dst;
2730
2731                                 unclone = skb_tunnel_info_unclone(skb);
2732                                 if (unlikely(!unclone))
2733                                         goto tx_error;
2734
2735                                 src = remote_ip.sin.sin_addr;
2736                                 dst = local_ip.sin.sin_addr;
2737                                 unclone->key.u.ipv4.src = src.s_addr;
2738                                 unclone->key.u.ipv4.dst = dst.s_addr;
2739                         }
2740                         vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2741                         dst_release(ndst);
2742                         goto out_unlock;
2743                 }
2744
2745                 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2746                 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2747                 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2748                                       vni, md, flags, udp_sum);
2749                 if (err < 0)
2750                         goto tx_error;
2751
2752                 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2753                                     dst->sin.sin_addr.s_addr, tos, ttl, df,
2754                                     src_port, dst_port, xnet, !udp_sum);
2755 #if IS_ENABLED(CONFIG_IPV6)
2756         } else {
2757                 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2758
2759                 if (!ifindex)
2760                         ifindex = sock6->sock->sk->sk_bound_dev_if;
2761
2762                 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2763                                         label, &dst->sin6.sin6_addr,
2764                                         &local_ip.sin6.sin6_addr,
2765                                         dst_port, src_port,
2766                                         dst_cache, info);
2767                 if (IS_ERR(ndst)) {
2768                         err = PTR_ERR(ndst);
2769                         ndst = NULL;
2770                         goto tx_error;
2771                 }
2772
2773                 if (!info) {
2774                         u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2775
2776                         err = encap_bypass_if_local(skb, dev, vxlan, dst,
2777                                                     dst_port, ifindex, vni,
2778                                                     ndst, rt6i_flags);
2779                         if (err)
2780                                 goto out_unlock;
2781                 }
2782
2783                 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM,
2784                                             netif_is_any_bridge_port(dev));
2785                 if (err < 0) {
2786                         goto tx_error;
2787                 } else if (err) {
2788                         if (info) {
2789                                 struct ip_tunnel_info *unclone;
2790                                 struct in6_addr src, dst;
2791
2792                                 unclone = skb_tunnel_info_unclone(skb);
2793                                 if (unlikely(!unclone))
2794                                         goto tx_error;
2795
2796                                 src = remote_ip.sin6.sin6_addr;
2797                                 dst = local_ip.sin6.sin6_addr;
2798                                 unclone->key.u.ipv6.src = src;
2799                                 unclone->key.u.ipv6.dst = dst;
2800                         }
2801
2802                         vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2803                         dst_release(ndst);
2804                         goto out_unlock;
2805                 }
2806
2807                 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2808                 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2809                 skb_scrub_packet(skb, xnet);
2810                 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2811                                       vni, md, flags, udp_sum);
2812                 if (err < 0)
2813                         goto tx_error;
2814
2815                 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2816                                      &local_ip.sin6.sin6_addr,
2817                                      &dst->sin6.sin6_addr, tos, ttl,
2818                                      label, src_port, dst_port, !udp_sum);
2819 #endif
2820         }
2821 out_unlock:
2822         rcu_read_unlock();
2823         return;
2824
2825 drop:
2826         dev->stats.tx_dropped++;
2827         dev_kfree_skb(skb);
2828         return;
2829
2830 tx_error:
2831         rcu_read_unlock();
2832         if (err == -ELOOP)
2833                 dev->stats.collisions++;
2834         else if (err == -ENETUNREACH)
2835                 dev->stats.tx_carrier_errors++;
2836         dst_release(ndst);
2837         dev->stats.tx_errors++;
2838         kfree_skb(skb);
2839 }
2840
2841 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2842                           struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2843 {
2844         struct vxlan_rdst nh_rdst;
2845         struct nexthop *nh;
2846         bool do_xmit;
2847         u32 hash;
2848
2849         memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2850         hash = skb_get_hash(skb);
2851
2852         rcu_read_lock();
2853         nh = rcu_dereference(f->nh);
2854         if (!nh) {
2855                 rcu_read_unlock();
2856                 goto drop;
2857         }
2858         do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2859         rcu_read_unlock();
2860
2861         if (likely(do_xmit))
2862                 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2863         else
2864                 goto drop;
2865
2866         return;
2867
2868 drop:
2869         dev->stats.tx_dropped++;
2870         dev_kfree_skb(skb);
2871 }
2872
2873 /* Transmit local packets over Vxlan
2874  *
2875  * Outer IP header inherits ECN and DF from inner header.
2876  * Outer UDP destination is the VXLAN assigned port.
2877  *           source port is based on hash of flow
2878  */
2879 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2880 {
2881         struct vxlan_dev *vxlan = netdev_priv(dev);
2882         struct vxlan_rdst *rdst, *fdst = NULL;
2883         const struct ip_tunnel_info *info;
2884         bool did_rsc = false;
2885         struct vxlan_fdb *f;
2886         struct ethhdr *eth;
2887         __be32 vni = 0;
2888
2889         info = skb_tunnel_info(skb);
2890
2891         skb_reset_mac_header(skb);
2892
2893         if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2894                 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2895                     info->mode & IP_TUNNEL_INFO_TX) {
2896                         vni = tunnel_id_to_key32(info->key.tun_id);
2897                 } else {
2898                         if (info && info->mode & IP_TUNNEL_INFO_TX)
2899                                 vxlan_xmit_one(skb, dev, vni, NULL, false);
2900                         else
2901                                 kfree_skb(skb);
2902                         return NETDEV_TX_OK;
2903                 }
2904         }
2905
2906         if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2907                 eth = eth_hdr(skb);
2908                 if (ntohs(eth->h_proto) == ETH_P_ARP)
2909                         return arp_reduce(dev, skb, vni);
2910 #if IS_ENABLED(CONFIG_IPV6)
2911                 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2912                          pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2913                                             sizeof(struct nd_msg)) &&
2914                          ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2915                         struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2916
2917                         if (m->icmph.icmp6_code == 0 &&
2918                             m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2919                                 return neigh_reduce(dev, skb, vni);
2920                 }
2921 #endif
2922         }
2923
2924         eth = eth_hdr(skb);
2925         f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2926         did_rsc = false;
2927
2928         if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2929             (ntohs(eth->h_proto) == ETH_P_IP ||
2930              ntohs(eth->h_proto) == ETH_P_IPV6)) {
2931                 did_rsc = route_shortcircuit(dev, skb);
2932                 if (did_rsc)
2933                         f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2934         }
2935
2936         if (f == NULL) {
2937                 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2938                 if (f == NULL) {
2939                         if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2940                             !is_multicast_ether_addr(eth->h_dest))
2941                                 vxlan_fdb_miss(vxlan, eth->h_dest);
2942
2943                         dev->stats.tx_dropped++;
2944                         kfree_skb(skb);
2945                         return NETDEV_TX_OK;
2946                 }
2947         }
2948
2949         if (rcu_access_pointer(f->nh)) {
2950                 vxlan_xmit_nh(skb, dev, f,
2951                               (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2952         } else {
2953                 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2954                         struct sk_buff *skb1;
2955
2956                         if (!fdst) {
2957                                 fdst = rdst;
2958                                 continue;
2959                         }
2960                         skb1 = skb_clone(skb, GFP_ATOMIC);
2961                         if (skb1)
2962                                 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2963                 }
2964                 if (fdst)
2965                         vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2966                 else
2967                         kfree_skb(skb);
2968         }
2969
2970         return NETDEV_TX_OK;
2971 }
2972
2973 /* Walk the forwarding table and purge stale entries */
2974 static void vxlan_cleanup(struct timer_list *t)
2975 {
2976         struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2977         unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2978         unsigned int h;
2979
2980         if (!netif_running(vxlan->dev))
2981                 return;
2982
2983         for (h = 0; h < FDB_HASH_SIZE; ++h) {
2984                 struct hlist_node *p, *n;
2985
2986                 spin_lock(&vxlan->hash_lock[h]);
2987                 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2988                         struct vxlan_fdb *f
2989                                 = container_of(p, struct vxlan_fdb, hlist);
2990                         unsigned long timeout;
2991
2992                         if (f->state & (NUD_PERMANENT | NUD_NOARP))
2993                                 continue;
2994
2995                         if (f->flags & NTF_EXT_LEARNED)
2996                                 continue;
2997
2998                         timeout = f->used + vxlan->cfg.age_interval * HZ;
2999                         if (time_before_eq(timeout, jiffies)) {
3000                                 netdev_dbg(vxlan->dev,
3001                                            "garbage collect %pM\n",
3002                                            f->eth_addr);
3003                                 f->state = NUD_STALE;
3004                                 vxlan_fdb_destroy(vxlan, f, true, true);
3005                         } else if (time_before(timeout, next_timer))
3006                                 next_timer = timeout;
3007                 }
3008                 spin_unlock(&vxlan->hash_lock[h]);
3009         }
3010
3011         mod_timer(&vxlan->age_timer, next_timer);
3012 }
3013
3014 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
3015 {
3016         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3017
3018         spin_lock(&vn->sock_lock);
3019         hlist_del_init_rcu(&vxlan->hlist4.hlist);
3020 #if IS_ENABLED(CONFIG_IPV6)
3021         hlist_del_init_rcu(&vxlan->hlist6.hlist);
3022 #endif
3023         spin_unlock(&vn->sock_lock);
3024 }
3025
3026 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
3027                              struct vxlan_dev_node *node)
3028 {
3029         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3030         __be32 vni = vxlan->default_dst.remote_vni;
3031
3032         node->vxlan = vxlan;
3033         spin_lock(&vn->sock_lock);
3034         hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
3035         spin_unlock(&vn->sock_lock);
3036 }
3037
3038 /* Setup stats when device is created */
3039 static int vxlan_init(struct net_device *dev)
3040 {
3041         struct vxlan_dev *vxlan = netdev_priv(dev);
3042         int err;
3043
3044         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
3045         if (!dev->tstats)
3046                 return -ENOMEM;
3047
3048         err = gro_cells_init(&vxlan->gro_cells, dev);
3049         if (err) {
3050                 free_percpu(dev->tstats);
3051                 return err;
3052         }
3053
3054         return 0;
3055 }
3056
3057 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
3058 {
3059         struct vxlan_fdb *f;
3060         u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
3061
3062         spin_lock_bh(&vxlan->hash_lock[hash_index]);
3063         f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
3064         if (f)
3065                 vxlan_fdb_destroy(vxlan, f, true, true);
3066         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
3067 }
3068
3069 static void vxlan_uninit(struct net_device *dev)
3070 {
3071         struct vxlan_dev *vxlan = netdev_priv(dev);
3072
3073         gro_cells_destroy(&vxlan->gro_cells);
3074
3075         vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
3076
3077         free_percpu(dev->tstats);
3078 }
3079
3080 /* Start ageing timer and join group when device is brought up */
3081 static int vxlan_open(struct net_device *dev)
3082 {
3083         struct vxlan_dev *vxlan = netdev_priv(dev);
3084         int ret;
3085
3086         ret = vxlan_sock_add(vxlan);
3087         if (ret < 0)
3088                 return ret;
3089
3090         if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
3091                 ret = vxlan_igmp_join(vxlan);
3092                 if (ret == -EADDRINUSE)
3093                         ret = 0;
3094                 if (ret) {
3095                         vxlan_sock_release(vxlan);
3096                         return ret;
3097                 }
3098         }
3099
3100         if (vxlan->cfg.age_interval)
3101                 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
3102
3103         return ret;
3104 }
3105
3106 /* Purge the forwarding table */
3107 static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
3108 {
3109         unsigned int h;
3110
3111         for (h = 0; h < FDB_HASH_SIZE; ++h) {
3112                 struct hlist_node *p, *n;
3113
3114                 spin_lock_bh(&vxlan->hash_lock[h]);
3115                 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3116                         struct vxlan_fdb *f
3117                                 = container_of(p, struct vxlan_fdb, hlist);
3118                         if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
3119                                 continue;
3120                         /* the all_zeros_mac entry is deleted at vxlan_uninit */
3121                         if (is_zero_ether_addr(f->eth_addr) &&
3122                             f->vni == vxlan->cfg.vni)
3123                                 continue;
3124                         vxlan_fdb_destroy(vxlan, f, true, true);
3125                 }
3126                 spin_unlock_bh(&vxlan->hash_lock[h]);
3127         }
3128 }
3129
3130 /* Cleanup timer and forwarding table on shutdown */
3131 static int vxlan_stop(struct net_device *dev)
3132 {
3133         struct vxlan_dev *vxlan = netdev_priv(dev);
3134         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3135         int ret = 0;
3136
3137         if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
3138             !vxlan_group_used(vn, vxlan))
3139                 ret = vxlan_igmp_leave(vxlan);
3140
3141         del_timer_sync(&vxlan->age_timer);
3142
3143         vxlan_flush(vxlan, false);
3144         vxlan_sock_release(vxlan);
3145
3146         return ret;
3147 }
3148
3149 /* Stub, nothing needs to be done. */
3150 static void vxlan_set_multicast_list(struct net_device *dev)
3151 {
3152 }
3153
3154 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3155 {
3156         struct vxlan_dev *vxlan = netdev_priv(dev);
3157         struct vxlan_rdst *dst = &vxlan->default_dst;
3158         struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3159                                                          dst->remote_ifindex);
3160         bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
3161
3162         /* This check is different than dev->max_mtu, because it looks at
3163          * the lowerdev->mtu, rather than the static dev->max_mtu
3164          */
3165         if (lowerdev) {
3166                 int max_mtu = lowerdev->mtu -
3167                               (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
3168                 if (new_mtu > max_mtu)
3169                         return -EINVAL;
3170         }
3171
3172         dev->mtu = new_mtu;
3173         return 0;
3174 }
3175
3176 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3177 {
3178         struct vxlan_dev *vxlan = netdev_priv(dev);
3179         struct ip_tunnel_info *info = skb_tunnel_info(skb);
3180         __be16 sport, dport;
3181
3182         sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3183                                   vxlan->cfg.port_max, true);
3184         dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3185
3186         if (ip_tunnel_info_af(info) == AF_INET) {
3187                 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3188                 struct rtable *rt;
3189
3190                 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
3191                                      info->key.u.ipv4.dst,
3192                                      &info->key.u.ipv4.src, dport, sport,
3193                                      &info->dst_cache, info);
3194                 if (IS_ERR(rt))
3195                         return PTR_ERR(rt);
3196                 ip_rt_put(rt);
3197         } else {
3198 #if IS_ENABLED(CONFIG_IPV6)
3199                 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3200                 struct dst_entry *ndst;
3201
3202                 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
3203                                         info->key.label, &info->key.u.ipv6.dst,
3204                                         &info->key.u.ipv6.src, dport, sport,
3205                                         &info->dst_cache, info);
3206                 if (IS_ERR(ndst))
3207                         return PTR_ERR(ndst);
3208                 dst_release(ndst);
3209 #else /* !CONFIG_IPV6 */
3210                 return -EPFNOSUPPORT;
3211 #endif
3212         }
3213         info->key.tp_src = sport;
3214         info->key.tp_dst = dport;
3215         return 0;
3216 }
3217
3218 static const struct net_device_ops vxlan_netdev_ether_ops = {
3219         .ndo_init               = vxlan_init,
3220         .ndo_uninit             = vxlan_uninit,
3221         .ndo_open               = vxlan_open,
3222         .ndo_stop               = vxlan_stop,
3223         .ndo_start_xmit         = vxlan_xmit,
3224         .ndo_get_stats64        = dev_get_tstats64,
3225         .ndo_set_rx_mode        = vxlan_set_multicast_list,
3226         .ndo_change_mtu         = vxlan_change_mtu,
3227         .ndo_validate_addr      = eth_validate_addr,
3228         .ndo_set_mac_address    = eth_mac_addr,
3229         .ndo_fdb_add            = vxlan_fdb_add,
3230         .ndo_fdb_del            = vxlan_fdb_delete,
3231         .ndo_fdb_dump           = vxlan_fdb_dump,
3232         .ndo_fdb_get            = vxlan_fdb_get,
3233         .ndo_fill_metadata_dst  = vxlan_fill_metadata_dst,
3234         .ndo_change_proto_down  = dev_change_proto_down_generic,
3235 };
3236
3237 static const struct net_device_ops vxlan_netdev_raw_ops = {
3238         .ndo_init               = vxlan_init,
3239         .ndo_uninit             = vxlan_uninit,
3240         .ndo_open               = vxlan_open,
3241         .ndo_stop               = vxlan_stop,
3242         .ndo_start_xmit         = vxlan_xmit,
3243         .ndo_get_stats64        = dev_get_tstats64,
3244         .ndo_change_mtu         = vxlan_change_mtu,
3245         .ndo_fill_metadata_dst  = vxlan_fill_metadata_dst,
3246 };
3247
3248 /* Info for udev, that this is a virtual tunnel endpoint */
3249 static struct device_type vxlan_type = {
3250         .name = "vxlan",
3251 };
3252
3253 /* Calls the ndo_udp_tunnel_add of the caller in order to
3254  * supply the listening VXLAN udp ports. Callers are expected
3255  * to implement the ndo_udp_tunnel_add.
3256  */
3257 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3258 {
3259         struct vxlan_sock *vs;
3260         struct net *net = dev_net(dev);
3261         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3262         unsigned int i;
3263
3264         spin_lock(&vn->sock_lock);
3265         for (i = 0; i < PORT_HASH_SIZE; ++i) {
3266                 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3267                         unsigned short type;
3268
3269                         if (vs->flags & VXLAN_F_GPE)
3270                                 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3271                         else
3272                                 type = UDP_TUNNEL_TYPE_VXLAN;
3273
3274                         if (push)
3275                                 udp_tunnel_push_rx_port(dev, vs->sock, type);
3276                         else
3277                                 udp_tunnel_drop_rx_port(dev, vs->sock, type);
3278                 }
3279         }
3280         spin_unlock(&vn->sock_lock);
3281 }
3282
3283 /* Initialize the device structure. */
3284 static void vxlan_setup(struct net_device *dev)
3285 {
3286         struct vxlan_dev *vxlan = netdev_priv(dev);
3287         unsigned int h;
3288
3289         eth_hw_addr_random(dev);
3290         ether_setup(dev);
3291
3292         dev->needs_free_netdev = true;
3293         SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3294
3295         dev->features   |= NETIF_F_LLTX;
3296         dev->features   |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3297         dev->features   |= NETIF_F_RXCSUM;
3298         dev->features   |= NETIF_F_GSO_SOFTWARE;
3299
3300         dev->vlan_features = dev->features;
3301         dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3302         dev->hw_features |= NETIF_F_RXCSUM;
3303         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3304         netif_keep_dst(dev);
3305         dev->priv_flags |= IFF_NO_QUEUE;
3306
3307         /* MTU range: 68 - 65535 */
3308         dev->min_mtu = ETH_MIN_MTU;
3309         dev->max_mtu = ETH_MAX_MTU;
3310
3311         INIT_LIST_HEAD(&vxlan->next);
3312
3313         timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3314
3315         vxlan->dev = dev;
3316
3317         for (h = 0; h < FDB_HASH_SIZE; ++h) {
3318                 spin_lock_init(&vxlan->hash_lock[h]);
3319                 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3320         }
3321 }
3322
3323 static void vxlan_ether_setup(struct net_device *dev)
3324 {
3325         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3326         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3327         dev->netdev_ops = &vxlan_netdev_ether_ops;
3328 }
3329
3330 static void vxlan_raw_setup(struct net_device *dev)
3331 {
3332         dev->header_ops = NULL;
3333         dev->type = ARPHRD_NONE;
3334         dev->hard_header_len = 0;
3335         dev->addr_len = 0;
3336         dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3337         dev->netdev_ops = &vxlan_netdev_raw_ops;
3338 }
3339
3340 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3341         [IFLA_VXLAN_ID]         = { .type = NLA_U32 },
3342         [IFLA_VXLAN_GROUP]      = { .len = sizeof_field(struct iphdr, daddr) },
3343         [IFLA_VXLAN_GROUP6]     = { .len = sizeof(struct in6_addr) },
3344         [IFLA_VXLAN_LINK]       = { .type = NLA_U32 },
3345         [IFLA_VXLAN_LOCAL]      = { .len = sizeof_field(struct iphdr, saddr) },
3346         [IFLA_VXLAN_LOCAL6]     = { .len = sizeof(struct in6_addr) },
3347         [IFLA_VXLAN_TOS]        = { .type = NLA_U8 },
3348         [IFLA_VXLAN_TTL]        = { .type = NLA_U8 },
3349         [IFLA_VXLAN_LABEL]      = { .type = NLA_U32 },
3350         [IFLA_VXLAN_LEARNING]   = { .type = NLA_U8 },
3351         [IFLA_VXLAN_AGEING]     = { .type = NLA_U32 },
3352         [IFLA_VXLAN_LIMIT]      = { .type = NLA_U32 },
3353         [IFLA_VXLAN_PORT_RANGE] = { .len  = sizeof(struct ifla_vxlan_port_range) },
3354         [IFLA_VXLAN_PROXY]      = { .type = NLA_U8 },
3355         [IFLA_VXLAN_RSC]        = { .type = NLA_U8 },
3356         [IFLA_VXLAN_L2MISS]     = { .type = NLA_U8 },
3357         [IFLA_VXLAN_L3MISS]     = { .type = NLA_U8 },
3358         [IFLA_VXLAN_COLLECT_METADATA]   = { .type = NLA_U8 },
3359         [IFLA_VXLAN_PORT]       = { .type = NLA_U16 },
3360         [IFLA_VXLAN_UDP_CSUM]   = { .type = NLA_U8 },
3361         [IFLA_VXLAN_UDP_ZERO_CSUM6_TX]  = { .type = NLA_U8 },
3362         [IFLA_VXLAN_UDP_ZERO_CSUM6_RX]  = { .type = NLA_U8 },
3363         [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3364         [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3365         [IFLA_VXLAN_GBP]        = { .type = NLA_FLAG, },
3366         [IFLA_VXLAN_GPE]        = { .type = NLA_FLAG, },
3367         [IFLA_VXLAN_REMCSUM_NOPARTIAL]  = { .type = NLA_FLAG },
3368         [IFLA_VXLAN_TTL_INHERIT]        = { .type = NLA_FLAG },
3369         [IFLA_VXLAN_DF]         = { .type = NLA_U8 },
3370 };
3371
3372 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3373                           struct netlink_ext_ack *extack)
3374 {
3375         if (tb[IFLA_ADDRESS]) {
3376                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3377                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3378                                             "Provided link layer address is not Ethernet");
3379                         return -EINVAL;
3380                 }
3381
3382                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3383                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3384                                             "Provided Ethernet address is not unicast");
3385                         return -EADDRNOTAVAIL;
3386                 }
3387         }
3388
3389         if (tb[IFLA_MTU]) {
3390                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3391
3392                 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3393                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3394                                             "MTU must be between 68 and 65535");
3395                         return -EINVAL;
3396                 }
3397         }
3398
3399         if (!data) {
3400                 NL_SET_ERR_MSG(extack,
3401                                "Required attributes not provided to perform the operation");
3402                 return -EINVAL;
3403         }
3404
3405         if (data[IFLA_VXLAN_ID]) {
3406                 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3407
3408                 if (id >= VXLAN_N_VID) {
3409                         NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3410                                             "VXLAN ID must be lower than 16777216");
3411                         return -ERANGE;
3412                 }
3413         }
3414
3415         if (data[IFLA_VXLAN_PORT_RANGE]) {
3416                 const struct ifla_vxlan_port_range *p
3417                         = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3418
3419                 if (ntohs(p->high) < ntohs(p->low)) {
3420                         NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3421                                             "Invalid source port range");
3422                         return -EINVAL;
3423                 }
3424         }
3425
3426         if (data[IFLA_VXLAN_DF]) {
3427                 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3428
3429                 if (df < 0 || df > VXLAN_DF_MAX) {
3430                         NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3431                                             "Invalid DF attribute");
3432                         return -EINVAL;
3433                 }
3434         }
3435
3436         return 0;
3437 }
3438
3439 static void vxlan_get_drvinfo(struct net_device *netdev,
3440                               struct ethtool_drvinfo *drvinfo)
3441 {
3442         strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3443         strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3444 }
3445
3446 static int vxlan_get_link_ksettings(struct net_device *dev,
3447                                     struct ethtool_link_ksettings *cmd)
3448 {
3449         struct vxlan_dev *vxlan = netdev_priv(dev);
3450         struct vxlan_rdst *dst = &vxlan->default_dst;
3451         struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3452                                                          dst->remote_ifindex);
3453
3454         if (!lowerdev) {
3455                 cmd->base.duplex = DUPLEX_UNKNOWN;
3456                 cmd->base.port = PORT_OTHER;
3457                 cmd->base.speed = SPEED_UNKNOWN;
3458
3459                 return 0;
3460         }
3461
3462         return __ethtool_get_link_ksettings(lowerdev, cmd);
3463 }
3464
3465 static const struct ethtool_ops vxlan_ethtool_ops = {
3466         .get_drvinfo            = vxlan_get_drvinfo,
3467         .get_link               = ethtool_op_get_link,
3468         .get_link_ksettings     = vxlan_get_link_ksettings,
3469 };
3470
3471 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3472                                         __be16 port, u32 flags, int ifindex)
3473 {
3474         struct socket *sock;
3475         struct udp_port_cfg udp_conf;
3476         int err;
3477
3478         memset(&udp_conf, 0, sizeof(udp_conf));
3479
3480         if (ipv6) {
3481                 udp_conf.family = AF_INET6;
3482                 udp_conf.use_udp6_rx_checksums =
3483                     !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3484                 udp_conf.ipv6_v6only = 1;
3485         } else {
3486                 udp_conf.family = AF_INET;
3487         }
3488
3489         udp_conf.local_udp_port = port;
3490         udp_conf.bind_ifindex = ifindex;
3491
3492         /* Open UDP socket */
3493         err = udp_sock_create(net, &udp_conf, &sock);
3494         if (err < 0)
3495                 return ERR_PTR(err);
3496
3497         udp_allow_gso(sock->sk);
3498         return sock;
3499 }
3500
3501 /* Create new listen socket if needed */
3502 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3503                                               __be16 port, u32 flags,
3504                                               int ifindex)
3505 {
3506         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3507         struct vxlan_sock *vs;
3508         struct socket *sock;
3509         unsigned int h;
3510         struct udp_tunnel_sock_cfg tunnel_cfg;
3511
3512         vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3513         if (!vs)
3514                 return ERR_PTR(-ENOMEM);
3515
3516         for (h = 0; h < VNI_HASH_SIZE; ++h)
3517                 INIT_HLIST_HEAD(&vs->vni_list[h]);
3518
3519         sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3520         if (IS_ERR(sock)) {
3521                 kfree(vs);
3522                 return ERR_CAST(sock);
3523         }
3524
3525         vs->sock = sock;
3526         refcount_set(&vs->refcnt, 1);
3527         vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3528
3529         spin_lock(&vn->sock_lock);
3530         hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3531         udp_tunnel_notify_add_rx_port(sock,
3532                                       (vs->flags & VXLAN_F_GPE) ?
3533                                       UDP_TUNNEL_TYPE_VXLAN_GPE :
3534                                       UDP_TUNNEL_TYPE_VXLAN);
3535         spin_unlock(&vn->sock_lock);
3536
3537         /* Mark socket as an encapsulation socket. */
3538         memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3539         tunnel_cfg.sk_user_data = vs;
3540         tunnel_cfg.encap_type = 1;
3541         tunnel_cfg.encap_rcv = vxlan_rcv;
3542         tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3543         tunnel_cfg.encap_destroy = NULL;
3544         tunnel_cfg.gro_receive = vxlan_gro_receive;
3545         tunnel_cfg.gro_complete = vxlan_gro_complete;
3546
3547         setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3548
3549         return vs;
3550 }
3551
3552 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3553 {
3554         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3555         struct vxlan_sock *vs = NULL;
3556         struct vxlan_dev_node *node;
3557         int l3mdev_index = 0;
3558
3559         if (vxlan->cfg.remote_ifindex)
3560                 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3561                         vxlan->net, vxlan->cfg.remote_ifindex);
3562
3563         if (!vxlan->cfg.no_share) {
3564                 spin_lock(&vn->sock_lock);
3565                 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3566                                      vxlan->cfg.dst_port, vxlan->cfg.flags,
3567                                      l3mdev_index);
3568                 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3569                         spin_unlock(&vn->sock_lock);
3570                         return -EBUSY;
3571                 }
3572                 spin_unlock(&vn->sock_lock);
3573         }
3574         if (!vs)
3575                 vs = vxlan_socket_create(vxlan->net, ipv6,
3576                                          vxlan->cfg.dst_port, vxlan->cfg.flags,
3577                                          l3mdev_index);
3578         if (IS_ERR(vs))
3579                 return PTR_ERR(vs);
3580 #if IS_ENABLED(CONFIG_IPV6)
3581         if (ipv6) {
3582                 rcu_assign_pointer(vxlan->vn6_sock, vs);
3583                 node = &vxlan->hlist6;
3584         } else
3585 #endif
3586         {
3587                 rcu_assign_pointer(vxlan->vn4_sock, vs);
3588                 node = &vxlan->hlist4;
3589         }
3590         vxlan_vs_add_dev(vs, vxlan, node);
3591         return 0;
3592 }
3593
3594 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3595 {
3596         bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3597         bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3598         bool ipv4 = !ipv6 || metadata;
3599         int ret = 0;
3600
3601         RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3602 #if IS_ENABLED(CONFIG_IPV6)
3603         RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3604         if (ipv6) {
3605                 ret = __vxlan_sock_add(vxlan, true);
3606                 if (ret < 0 && ret != -EAFNOSUPPORT)
3607                         ipv4 = false;
3608         }
3609 #endif
3610         if (ipv4)
3611                 ret = __vxlan_sock_add(vxlan, false);
3612         if (ret < 0)
3613                 vxlan_sock_release(vxlan);
3614         return ret;
3615 }
3616
3617 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3618                                  struct net_device **lower,
3619                                  struct vxlan_dev *old,
3620                                  struct netlink_ext_ack *extack)
3621 {
3622         struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3623         struct vxlan_dev *tmp;
3624         bool use_ipv6 = false;
3625
3626         if (conf->flags & VXLAN_F_GPE) {
3627                 /* For now, allow GPE only together with
3628                  * COLLECT_METADATA. This can be relaxed later; in such
3629                  * case, the other side of the PtP link will have to be
3630                  * provided.
3631                  */
3632                 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3633                     !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3634                         NL_SET_ERR_MSG(extack,
3635                                        "VXLAN GPE does not support this combination of attributes");
3636                         return -EINVAL;
3637                 }
3638         }
3639
3640         if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3641                 /* Unless IPv6 is explicitly requested, assume IPv4 */
3642                 conf->remote_ip.sa.sa_family = AF_INET;
3643                 conf->saddr.sa.sa_family = AF_INET;
3644         } else if (!conf->remote_ip.sa.sa_family) {
3645                 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3646         } else if (!conf->saddr.sa.sa_family) {
3647                 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3648         }
3649
3650         if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3651                 NL_SET_ERR_MSG(extack,
3652                                "Local and remote address must be from the same family");
3653                 return -EINVAL;
3654         }
3655
3656         if (vxlan_addr_multicast(&conf->saddr)) {
3657                 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3658                 return -EINVAL;
3659         }
3660
3661         if (conf->saddr.sa.sa_family == AF_INET6) {
3662                 if (!IS_ENABLED(CONFIG_IPV6)) {
3663                         NL_SET_ERR_MSG(extack,
3664                                        "IPv6 support not enabled in the kernel");
3665                         return -EPFNOSUPPORT;
3666                 }
3667                 use_ipv6 = true;
3668                 conf->flags |= VXLAN_F_IPV6;
3669
3670                 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3671                         int local_type =
3672                                 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3673                         int remote_type =
3674                                 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3675
3676                         if (local_type & IPV6_ADDR_LINKLOCAL) {
3677                                 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3678                                     (remote_type != IPV6_ADDR_ANY)) {
3679                                         NL_SET_ERR_MSG(extack,
3680                                                        "Invalid combination of local and remote address scopes");
3681                                         return -EINVAL;
3682                                 }
3683
3684                                 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3685                         } else {
3686                                 if (remote_type ==
3687                                     (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3688                                         NL_SET_ERR_MSG(extack,
3689                                                        "Invalid combination of local and remote address scopes");
3690                                         return -EINVAL;
3691                                 }
3692
3693                                 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3694                         }
3695                 }
3696         }
3697
3698         if (conf->label && !use_ipv6) {
3699                 NL_SET_ERR_MSG(extack,
3700                                "Label attribute only applies to IPv6 VXLAN devices");
3701                 return -EINVAL;
3702         }
3703
3704         if (conf->remote_ifindex) {
3705                 struct net_device *lowerdev;
3706
3707                 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3708                 if (!lowerdev) {
3709                         NL_SET_ERR_MSG(extack,
3710                                        "Invalid local interface, device not found");
3711                         return -ENODEV;
3712                 }
3713
3714 #if IS_ENABLED(CONFIG_IPV6)
3715                 if (use_ipv6) {
3716                         struct inet6_dev *idev = __in6_dev_get(lowerdev);
3717
3718                         if (idev && idev->cnf.disable_ipv6) {
3719                                 NL_SET_ERR_MSG(extack,
3720                                                "IPv6 support disabled by administrator");
3721                                 return -EPERM;
3722                         }
3723                 }
3724 #endif
3725
3726                 *lower = lowerdev;
3727         } else {
3728                 if (vxlan_addr_multicast(&conf->remote_ip)) {
3729                         NL_SET_ERR_MSG(extack,
3730                                        "Local interface required for multicast remote destination");
3731
3732                         return -EINVAL;
3733                 }
3734
3735 #if IS_ENABLED(CONFIG_IPV6)
3736                 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3737                         NL_SET_ERR_MSG(extack,
3738                                        "Local interface required for link-local local/remote addresses");
3739                         return -EINVAL;
3740                 }
3741 #endif
3742
3743                 *lower = NULL;
3744         }
3745
3746         if (!conf->dst_port) {
3747                 if (conf->flags & VXLAN_F_GPE)
3748                         conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3749                 else
3750                         conf->dst_port = htons(vxlan_port);
3751         }
3752
3753         if (!conf->age_interval)
3754                 conf->age_interval = FDB_AGE_DEFAULT;
3755
3756         list_for_each_entry(tmp, &vn->vxlan_list, next) {
3757                 if (tmp == old)
3758                         continue;
3759
3760                 if (tmp->cfg.vni != conf->vni)
3761                         continue;
3762                 if (tmp->cfg.dst_port != conf->dst_port)
3763                         continue;
3764                 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3765                     (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3766                         continue;
3767
3768                 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3769                     tmp->cfg.remote_ifindex != conf->remote_ifindex)
3770                         continue;
3771
3772                 NL_SET_ERR_MSG(extack,
3773                                "A VXLAN device with the specified VNI already exists");
3774                 return -EEXIST;
3775         }
3776
3777         return 0;
3778 }
3779
3780 static void vxlan_config_apply(struct net_device *dev,
3781                                struct vxlan_config *conf,
3782                                struct net_device *lowerdev,
3783                                struct net *src_net,
3784                                bool changelink)
3785 {
3786         struct vxlan_dev *vxlan = netdev_priv(dev);
3787         struct vxlan_rdst *dst = &vxlan->default_dst;
3788         unsigned short needed_headroom = ETH_HLEN;
3789         bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3790         int max_mtu = ETH_MAX_MTU;
3791
3792         if (!changelink) {
3793                 if (conf->flags & VXLAN_F_GPE)
3794                         vxlan_raw_setup(dev);
3795                 else
3796                         vxlan_ether_setup(dev);
3797
3798                 if (conf->mtu)
3799                         dev->mtu = conf->mtu;
3800
3801                 vxlan->net = src_net;
3802         }
3803
3804         dst->remote_vni = conf->vni;
3805
3806         memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3807
3808         if (lowerdev) {
3809                 dst->remote_ifindex = conf->remote_ifindex;
3810
3811                 dev->gso_max_size = lowerdev->gso_max_size;
3812                 dev->gso_max_segs = lowerdev->gso_max_segs;
3813
3814                 needed_headroom = lowerdev->hard_header_len;
3815                 needed_headroom += lowerdev->needed_headroom;
3816
3817                 dev->needed_tailroom = lowerdev->needed_tailroom;
3818
3819                 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3820                                            VXLAN_HEADROOM);
3821                 if (max_mtu < ETH_MIN_MTU)
3822                         max_mtu = ETH_MIN_MTU;
3823
3824                 if (!changelink && !conf->mtu)
3825                         dev->mtu = max_mtu;
3826         }
3827
3828         if (dev->mtu > max_mtu)
3829                 dev->mtu = max_mtu;
3830
3831         if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3832                 needed_headroom += VXLAN6_HEADROOM;
3833         else
3834                 needed_headroom += VXLAN_HEADROOM;
3835         dev->needed_headroom = needed_headroom;
3836
3837         memcpy(&vxlan->cfg, conf, sizeof(*conf));
3838 }
3839
3840 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3841                                struct vxlan_config *conf, bool changelink,
3842                                struct netlink_ext_ack *extack)
3843 {
3844         struct vxlan_dev *vxlan = netdev_priv(dev);
3845         struct net_device *lowerdev;
3846         int ret;
3847
3848         ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3849         if (ret)
3850                 return ret;
3851
3852         vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3853
3854         return 0;
3855 }
3856
3857 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3858                               struct vxlan_config *conf,
3859                               struct netlink_ext_ack *extack)
3860 {
3861         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3862         struct vxlan_dev *vxlan = netdev_priv(dev);
3863         struct net_device *remote_dev = NULL;
3864         struct vxlan_fdb *f = NULL;
3865         bool unregister = false;
3866         struct vxlan_rdst *dst;
3867         int err;
3868
3869         dst = &vxlan->default_dst;
3870         err = vxlan_dev_configure(net, dev, conf, false, extack);
3871         if (err)
3872                 return err;
3873
3874         dev->ethtool_ops = &vxlan_ethtool_ops;
3875
3876         /* create an fdb entry for a valid default destination */
3877         if (!vxlan_addr_any(&dst->remote_ip)) {
3878                 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3879                                        &dst->remote_ip,
3880                                        NUD_REACHABLE | NUD_PERMANENT,
3881                                        vxlan->cfg.dst_port,
3882                                        dst->remote_vni,
3883                                        dst->remote_vni,
3884                                        dst->remote_ifindex,
3885                                        NTF_SELF, 0, &f, extack);
3886                 if (err)
3887                         return err;
3888         }
3889
3890         err = register_netdevice(dev);
3891         if (err)
3892                 goto errout;
3893         unregister = true;
3894
3895         if (dst->remote_ifindex) {
3896                 remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3897                 if (!remote_dev) {
3898                         err = -ENODEV;
3899                         goto errout;
3900                 }
3901
3902                 err = netdev_upper_dev_link(remote_dev, dev, extack);
3903                 if (err)
3904                         goto errout;
3905         }
3906
3907         err = rtnl_configure_link(dev, NULL);
3908         if (err < 0)
3909                 goto unlink;
3910
3911         if (f) {
3912                 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
3913
3914                 /* notify default fdb entry */
3915                 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
3916                                        RTM_NEWNEIGH, true, extack);
3917                 if (err) {
3918                         vxlan_fdb_destroy(vxlan, f, false, false);
3919                         if (remote_dev)
3920                                 netdev_upper_dev_unlink(remote_dev, dev);
3921                         goto unregister;
3922                 }
3923         }
3924
3925         list_add(&vxlan->next, &vn->vxlan_list);
3926         if (remote_dev)
3927                 dst->remote_dev = remote_dev;
3928         return 0;
3929 unlink:
3930         if (remote_dev)
3931                 netdev_upper_dev_unlink(remote_dev, dev);
3932 errout:
3933         /* unregister_netdevice() destroys the default FDB entry with deletion
3934          * notification. But the addition notification was not sent yet, so
3935          * destroy the entry by hand here.
3936          */
3937         if (f)
3938                 __vxlan_fdb_free(f);
3939 unregister:
3940         if (unregister)
3941                 unregister_netdevice(dev);
3942         return err;
3943 }
3944
3945 /* Set/clear flags based on attribute */
3946 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3947                           int attrtype, unsigned long mask, bool changelink,
3948                           bool changelink_supported,
3949                           struct netlink_ext_ack *extack)
3950 {
3951         unsigned long flags;
3952
3953         if (!tb[attrtype])
3954                 return 0;
3955
3956         if (changelink && !changelink_supported) {
3957                 vxlan_flag_attr_error(attrtype, extack);
3958                 return -EOPNOTSUPP;
3959         }
3960
3961         if (vxlan_policy[attrtype].type == NLA_FLAG)
3962                 flags = conf->flags | mask;
3963         else if (nla_get_u8(tb[attrtype]))
3964                 flags = conf->flags | mask;
3965         else
3966                 flags = conf->flags & ~mask;
3967
3968         conf->flags = flags;
3969
3970         return 0;
3971 }
3972
3973 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3974                          struct net_device *dev, struct vxlan_config *conf,
3975                          bool changelink, struct netlink_ext_ack *extack)
3976 {
3977         struct vxlan_dev *vxlan = netdev_priv(dev);
3978         int err = 0;
3979
3980         memset(conf, 0, sizeof(*conf));
3981
3982         /* if changelink operation, start with old existing cfg */
3983         if (changelink)
3984                 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3985
3986         if (data[IFLA_VXLAN_ID]) {
3987                 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3988
3989                 if (changelink && (vni != conf->vni)) {
3990                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
3991                         return -EOPNOTSUPP;
3992                 }
3993                 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3994         }
3995
3996         if (data[IFLA_VXLAN_GROUP]) {
3997                 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
3998                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
3999                         return -EOPNOTSUPP;
4000                 }
4001
4002                 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
4003                 conf->remote_ip.sa.sa_family = AF_INET;
4004         } else if (data[IFLA_VXLAN_GROUP6]) {
4005                 if (!IS_ENABLED(CONFIG_IPV6)) {
4006                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
4007                         return -EPFNOSUPPORT;
4008                 }
4009
4010                 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
4011                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
4012                         return -EOPNOTSUPP;
4013                 }
4014
4015                 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
4016                 conf->remote_ip.sa.sa_family = AF_INET6;
4017         }
4018
4019         if (data[IFLA_VXLAN_LOCAL]) {
4020                 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
4021                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
4022                         return -EOPNOTSUPP;
4023                 }
4024
4025                 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
4026                 conf->saddr.sa.sa_family = AF_INET;
4027         } else if (data[IFLA_VXLAN_LOCAL6]) {
4028                 if (!IS_ENABLED(CONFIG_IPV6)) {
4029                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
4030                         return -EPFNOSUPPORT;
4031                 }
4032
4033                 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
4034                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
4035                         return -EOPNOTSUPP;
4036                 }
4037
4038                 /* TODO: respect scope id */
4039                 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
4040                 conf->saddr.sa.sa_family = AF_INET6;
4041         }
4042
4043         if (data[IFLA_VXLAN_LINK])
4044                 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
4045
4046         if (data[IFLA_VXLAN_TOS])
4047                 conf->tos  = nla_get_u8(data[IFLA_VXLAN_TOS]);
4048
4049         if (data[IFLA_VXLAN_TTL])
4050                 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
4051
4052         if (data[IFLA_VXLAN_TTL_INHERIT]) {
4053                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4054                                     VXLAN_F_TTL_INHERIT, changelink, false,
4055                                     extack);
4056                 if (err)
4057                         return err;
4058
4059         }
4060
4061         if (data[IFLA_VXLAN_LABEL])
4062                 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
4063                              IPV6_FLOWLABEL_MASK;
4064
4065         if (data[IFLA_VXLAN_LEARNING]) {
4066                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4067                                     VXLAN_F_LEARN, changelink, true,
4068                                     extack);
4069                 if (err)
4070                         return err;
4071         } else if (!changelink) {
4072                 /* default to learn on a new device */
4073                 conf->flags |= VXLAN_F_LEARN;
4074         }
4075
4076         if (data[IFLA_VXLAN_AGEING])
4077                 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
4078
4079         if (data[IFLA_VXLAN_PROXY]) {
4080                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4081                                     VXLAN_F_PROXY, changelink, false,
4082                                     extack);
4083                 if (err)
4084                         return err;
4085         }
4086
4087         if (data[IFLA_VXLAN_RSC]) {
4088                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4089                                     VXLAN_F_RSC, changelink, false,
4090                                     extack);
4091                 if (err)
4092                         return err;
4093         }
4094
4095         if (data[IFLA_VXLAN_L2MISS]) {
4096                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4097                                     VXLAN_F_L2MISS, changelink, false,
4098                                     extack);
4099                 if (err)
4100                         return err;
4101         }
4102
4103         if (data[IFLA_VXLAN_L3MISS]) {
4104                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4105                                     VXLAN_F_L3MISS, changelink, false,
4106                                     extack);
4107                 if (err)
4108                         return err;
4109         }
4110
4111         if (data[IFLA_VXLAN_LIMIT]) {
4112                 if (changelink) {
4113                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4114                                             "Cannot change limit");
4115                         return -EOPNOTSUPP;
4116                 }
4117                 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4118         }
4119
4120         if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4121                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4122                                     VXLAN_F_COLLECT_METADATA, changelink, false,
4123                                     extack);
4124                 if (err)
4125                         return err;
4126         }
4127
4128         if (data[IFLA_VXLAN_PORT_RANGE]) {
4129                 if (!changelink) {
4130                         const struct ifla_vxlan_port_range *p
4131                                 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4132                         conf->port_min = ntohs(p->low);
4133                         conf->port_max = ntohs(p->high);
4134                 } else {
4135                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4136                                             "Cannot change port range");
4137                         return -EOPNOTSUPP;
4138                 }
4139         }
4140
4141         if (data[IFLA_VXLAN_PORT]) {
4142                 if (changelink) {
4143                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4144                                             "Cannot change port");
4145                         return -EOPNOTSUPP;
4146                 }
4147                 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4148         }
4149
4150         if (data[IFLA_VXLAN_UDP_CSUM]) {
4151                 if (changelink) {
4152                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4153                                             "Cannot change UDP_CSUM flag");
4154                         return -EOPNOTSUPP;
4155                 }
4156                 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4157                         conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4158         }
4159
4160         if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4161                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4162                                     VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4163                                     false, extack);
4164                 if (err)
4165                         return err;
4166         }
4167
4168         if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4169                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4170                                     VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4171                                     false, extack);
4172                 if (err)
4173                         return err;
4174         }
4175
4176         if (data[IFLA_VXLAN_REMCSUM_TX]) {
4177                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4178                                     VXLAN_F_REMCSUM_TX, changelink, false,
4179                                     extack);
4180                 if (err)
4181                         return err;
4182         }
4183
4184         if (data[IFLA_VXLAN_REMCSUM_RX]) {
4185                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4186                                     VXLAN_F_REMCSUM_RX, changelink, false,
4187                                     extack);
4188                 if (err)
4189                         return err;
4190         }
4191
4192         if (data[IFLA_VXLAN_GBP]) {
4193                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4194                                     VXLAN_F_GBP, changelink, false, extack);
4195                 if (err)
4196                         return err;
4197         }
4198
4199         if (data[IFLA_VXLAN_GPE]) {
4200                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4201                                     VXLAN_F_GPE, changelink, false,
4202                                     extack);
4203                 if (err)
4204                         return err;
4205         }
4206
4207         if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4208                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4209                                     VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4210                                     false, extack);
4211                 if (err)
4212                         return err;
4213         }
4214
4215         if (tb[IFLA_MTU]) {
4216                 if (changelink) {
4217                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4218                                             "Cannot change mtu");
4219                         return -EOPNOTSUPP;
4220                 }
4221                 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4222         }
4223
4224         if (data[IFLA_VXLAN_DF])
4225                 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4226
4227         return 0;
4228 }
4229
4230 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
4231                          struct nlattr *tb[], struct nlattr *data[],
4232                          struct netlink_ext_ack *extack)
4233 {
4234         struct vxlan_config conf;
4235         int err;
4236
4237         err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4238         if (err)
4239                 return err;
4240
4241         return __vxlan_dev_create(src_net, dev, &conf, extack);
4242 }
4243
4244 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4245                             struct nlattr *data[],
4246                             struct netlink_ext_ack *extack)
4247 {
4248         struct vxlan_dev *vxlan = netdev_priv(dev);
4249         struct net_device *lowerdev;
4250         struct vxlan_config conf;
4251         struct vxlan_rdst *dst;
4252         int err;
4253
4254         dst = &vxlan->default_dst;
4255         err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4256         if (err)
4257                 return err;
4258
4259         err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4260                                     vxlan, extack);
4261         if (err)
4262                 return err;
4263
4264         if (dst->remote_dev == lowerdev)
4265                 lowerdev = NULL;
4266
4267         err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4268                                              extack);
4269         if (err)
4270                 return err;
4271
4272         /* handle default dst entry */
4273         if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
4274                 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4275
4276                 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4277                 if (!vxlan_addr_any(&conf.remote_ip)) {
4278                         err = vxlan_fdb_update(vxlan, all_zeros_mac,
4279                                                &conf.remote_ip,
4280                                                NUD_REACHABLE | NUD_PERMANENT,
4281                                                NLM_F_APPEND | NLM_F_CREATE,
4282                                                vxlan->cfg.dst_port,
4283                                                conf.vni, conf.vni,
4284                                                conf.remote_ifindex,
4285                                                NTF_SELF, 0, true, extack);
4286                         if (err) {
4287                                 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4288                                 netdev_adjacent_change_abort(dst->remote_dev,
4289                                                              lowerdev, dev);
4290                                 return err;
4291                         }
4292                 }
4293                 if (!vxlan_addr_any(&dst->remote_ip))
4294                         __vxlan_fdb_delete(vxlan, all_zeros_mac,
4295                                            dst->remote_ip,
4296                                            vxlan->cfg.dst_port,
4297                                            dst->remote_vni,
4298                                            dst->remote_vni,
4299                                            dst->remote_ifindex,
4300                                            true);
4301                 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4302         }
4303
4304         if (conf.age_interval != vxlan->cfg.age_interval)
4305                 mod_timer(&vxlan->age_timer, jiffies);
4306
4307         netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4308         if (lowerdev && lowerdev != dst->remote_dev)
4309                 dst->remote_dev = lowerdev;
4310         vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4311         return 0;
4312 }
4313
4314 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4315 {
4316         struct vxlan_dev *vxlan = netdev_priv(dev);
4317
4318         vxlan_flush(vxlan, true);
4319
4320         list_del(&vxlan->next);
4321         unregister_netdevice_queue(dev, head);
4322         if (vxlan->default_dst.remote_dev)
4323                 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4324 }
4325
4326 static size_t vxlan_get_size(const struct net_device *dev)
4327 {
4328
4329         return nla_total_size(sizeof(__u32)) +  /* IFLA_VXLAN_ID */
4330                 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4331                 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
4332                 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4333                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_TTL */
4334                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_TTL_INHERIT */
4335                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_TOS */
4336                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_DF */
4337                 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4338                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_LEARNING */
4339                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_PROXY */
4340                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_RSC */
4341                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_L2MISS */
4342                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_L3MISS */
4343                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_COLLECT_METADATA */
4344                 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
4345                 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
4346                 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4347                 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4348                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4349                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4350                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4351                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4352                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4353                 0;
4354 }
4355
4356 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4357 {
4358         const struct vxlan_dev *vxlan = netdev_priv(dev);
4359         const struct vxlan_rdst *dst = &vxlan->default_dst;
4360         struct ifla_vxlan_port_range ports = {
4361                 .low =  htons(vxlan->cfg.port_min),
4362                 .high = htons(vxlan->cfg.port_max),
4363         };
4364
4365         if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4366                 goto nla_put_failure;
4367
4368         if (!vxlan_addr_any(&dst->remote_ip)) {
4369                 if (dst->remote_ip.sa.sa_family == AF_INET) {
4370                         if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4371                                             dst->remote_ip.sin.sin_addr.s_addr))
4372                                 goto nla_put_failure;
4373 #if IS_ENABLED(CONFIG_IPV6)
4374                 } else {
4375                         if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4376                                              &dst->remote_ip.sin6.sin6_addr))
4377                                 goto nla_put_failure;
4378 #endif
4379                 }
4380         }
4381
4382         if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4383                 goto nla_put_failure;
4384
4385         if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4386                 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4387                         if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4388                                             vxlan->cfg.saddr.sin.sin_addr.s_addr))
4389                                 goto nla_put_failure;
4390 #if IS_ENABLED(CONFIG_IPV6)
4391                 } else {
4392                         if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4393                                              &vxlan->cfg.saddr.sin6.sin6_addr))
4394                                 goto nla_put_failure;
4395 #endif
4396                 }
4397         }
4398
4399         if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4400             nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4401                        !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4402             nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4403             nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4404             nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4405             nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4406                        !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4407             nla_put_u8(skb, IFLA_VXLAN_PROXY,
4408                        !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4409             nla_put_u8(skb, IFLA_VXLAN_RSC,
4410                        !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4411             nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4412                        !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4413             nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4414                        !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4415             nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4416                        !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4417             nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4418             nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4419             nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4420             nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4421                        !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4422             nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4423                        !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4424             nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4425                        !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4426             nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4427                        !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4428             nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4429                        !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
4430                 goto nla_put_failure;
4431
4432         if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4433                 goto nla_put_failure;
4434
4435         if (vxlan->cfg.flags & VXLAN_F_GBP &&
4436             nla_put_flag(skb, IFLA_VXLAN_GBP))
4437                 goto nla_put_failure;
4438
4439         if (vxlan->cfg.flags & VXLAN_F_GPE &&
4440             nla_put_flag(skb, IFLA_VXLAN_GPE))
4441                 goto nla_put_failure;
4442
4443         if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4444             nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4445                 goto nla_put_failure;
4446
4447         return 0;
4448
4449 nla_put_failure:
4450         return -EMSGSIZE;
4451 }
4452
4453 static struct net *vxlan_get_link_net(const struct net_device *dev)
4454 {
4455         struct vxlan_dev *vxlan = netdev_priv(dev);
4456
4457         return vxlan->net;
4458 }
4459
4460 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4461         .kind           = "vxlan",
4462         .maxtype        = IFLA_VXLAN_MAX,
4463         .policy         = vxlan_policy,
4464         .priv_size      = sizeof(struct vxlan_dev),
4465         .setup          = vxlan_setup,
4466         .validate       = vxlan_validate,
4467         .newlink        = vxlan_newlink,
4468         .changelink     = vxlan_changelink,
4469         .dellink        = vxlan_dellink,
4470         .get_size       = vxlan_get_size,
4471         .fill_info      = vxlan_fill_info,
4472         .get_link_net   = vxlan_get_link_net,
4473 };
4474
4475 struct net_device *vxlan_dev_create(struct net *net, const char *name,
4476                                     u8 name_assign_type,
4477                                     struct vxlan_config *conf)
4478 {
4479         struct nlattr *tb[IFLA_MAX + 1];
4480         struct net_device *dev;
4481         int err;
4482
4483         memset(&tb, 0, sizeof(tb));
4484
4485         dev = rtnl_create_link(net, name, name_assign_type,
4486                                &vxlan_link_ops, tb, NULL);
4487         if (IS_ERR(dev))
4488                 return dev;
4489
4490         err = __vxlan_dev_create(net, dev, conf, NULL);
4491         if (err < 0) {
4492                 free_netdev(dev);
4493                 return ERR_PTR(err);
4494         }
4495
4496         err = rtnl_configure_link(dev, NULL);
4497         if (err < 0) {
4498                 LIST_HEAD(list_kill);
4499
4500                 vxlan_dellink(dev, &list_kill);
4501                 unregister_netdevice_many(&list_kill);
4502                 return ERR_PTR(err);
4503         }
4504
4505         return dev;
4506 }
4507 EXPORT_SYMBOL_GPL(vxlan_dev_create);
4508
4509 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4510                                              struct net_device *dev)
4511 {
4512         struct vxlan_dev *vxlan, *next;
4513         LIST_HEAD(list_kill);
4514
4515         list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4516                 struct vxlan_rdst *dst = &vxlan->default_dst;
4517
4518                 /* In case we created vxlan device with carrier
4519                  * and we loose the carrier due to module unload
4520                  * we also need to remove vxlan device. In other
4521                  * cases, it's not necessary and remote_ifindex
4522                  * is 0 here, so no matches.
4523                  */
4524                 if (dst->remote_ifindex == dev->ifindex)
4525                         vxlan_dellink(vxlan->dev, &list_kill);
4526         }
4527
4528         unregister_netdevice_many(&list_kill);
4529 }
4530
4531 static int vxlan_netdevice_event(struct notifier_block *unused,
4532                                  unsigned long event, void *ptr)
4533 {
4534         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4535         struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4536
4537         if (event == NETDEV_UNREGISTER)
4538                 vxlan_handle_lowerdev_unregister(vn, dev);
4539         else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
4540                 vxlan_offload_rx_ports(dev, true);
4541         else if (event == NETDEV_UDP_TUNNEL_DROP_INFO)
4542                 vxlan_offload_rx_ports(dev, false);
4543
4544         return NOTIFY_DONE;
4545 }
4546
4547 static struct notifier_block vxlan_notifier_block __read_mostly = {
4548         .notifier_call = vxlan_netdevice_event,
4549 };
4550
4551 static void
4552 vxlan_fdb_offloaded_set(struct net_device *dev,
4553                         struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4554 {
4555         struct vxlan_dev *vxlan = netdev_priv(dev);
4556         struct vxlan_rdst *rdst;
4557         struct vxlan_fdb *f;
4558         u32 hash_index;
4559
4560         hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4561
4562         spin_lock_bh(&vxlan->hash_lock[hash_index]);
4563
4564         f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4565         if (!f)
4566                 goto out;
4567
4568         rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4569                                    fdb_info->remote_port,
4570                                    fdb_info->remote_vni,
4571                                    fdb_info->remote_ifindex);
4572         if (!rdst)
4573                 goto out;
4574
4575         rdst->offloaded = fdb_info->offloaded;
4576
4577 out:
4578         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4579 }
4580
4581 static int
4582 vxlan_fdb_external_learn_add(struct net_device *dev,
4583                              struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4584 {
4585         struct vxlan_dev *vxlan = netdev_priv(dev);
4586         struct netlink_ext_ack *extack;
4587         u32 hash_index;
4588         int err;
4589
4590         hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4591         extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4592
4593         spin_lock_bh(&vxlan->hash_lock[hash_index]);
4594         err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4595                                NUD_REACHABLE,
4596                                NLM_F_CREATE | NLM_F_REPLACE,
4597                                fdb_info->remote_port,
4598                                fdb_info->vni,
4599                                fdb_info->remote_vni,
4600                                fdb_info->remote_ifindex,
4601                                NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4602                                0, false, extack);
4603         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4604
4605         return err;
4606 }
4607
4608 static int
4609 vxlan_fdb_external_learn_del(struct net_device *dev,
4610                              struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4611 {
4612         struct vxlan_dev *vxlan = netdev_priv(dev);
4613         struct vxlan_fdb *f;
4614         u32 hash_index;
4615         int err = 0;
4616
4617         hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4618         spin_lock_bh(&vxlan->hash_lock[hash_index]);
4619
4620         f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4621         if (!f)
4622                 err = -ENOENT;
4623         else if (f->flags & NTF_EXT_LEARNED)
4624                 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4625                                          fdb_info->remote_ip,
4626                                          fdb_info->remote_port,
4627                                          fdb_info->vni,
4628                                          fdb_info->remote_vni,
4629                                          fdb_info->remote_ifindex,
4630                                          false);
4631
4632         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4633
4634         return err;
4635 }
4636
4637 static int vxlan_switchdev_event(struct notifier_block *unused,
4638                                  unsigned long event, void *ptr)
4639 {
4640         struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4641         struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4642         int err = 0;
4643
4644         switch (event) {
4645         case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4646                 vxlan_fdb_offloaded_set(dev, ptr);
4647                 break;
4648         case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4649                 fdb_info = ptr;
4650                 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4651                 if (err) {
4652                         err = notifier_from_errno(err);
4653                         break;
4654                 }
4655                 fdb_info->offloaded = true;
4656                 vxlan_fdb_offloaded_set(dev, fdb_info);
4657                 break;
4658         case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4659                 fdb_info = ptr;
4660                 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4661                 if (err) {
4662                         err = notifier_from_errno(err);
4663                         break;
4664                 }
4665                 fdb_info->offloaded = false;
4666                 vxlan_fdb_offloaded_set(dev, fdb_info);
4667                 break;
4668         }
4669
4670         return err;
4671 }
4672
4673 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4674         .notifier_call = vxlan_switchdev_event,
4675 };
4676
4677 static void vxlan_fdb_nh_flush(struct nexthop *nh)
4678 {
4679         struct vxlan_fdb *fdb;
4680         struct vxlan_dev *vxlan;
4681         u32 hash_index;
4682
4683         rcu_read_lock();
4684         list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4685                 vxlan = rcu_dereference(fdb->vdev);
4686                 WARN_ON(!vxlan);
4687                 hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4688                                             vxlan->default_dst.remote_vni);
4689                 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4690                 if (!hlist_unhashed(&fdb->hlist))
4691                         vxlan_fdb_destroy(vxlan, fdb, false, false);
4692                 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4693         }
4694         rcu_read_unlock();
4695 }
4696
4697 static int vxlan_nexthop_event(struct notifier_block *nb,
4698                                unsigned long event, void *ptr)
4699 {
4700         struct nh_notifier_info *info = ptr;
4701         struct nexthop *nh;
4702
4703         if (event != NEXTHOP_EVENT_DEL)
4704                 return NOTIFY_DONE;
4705
4706         nh = nexthop_find_by_id(info->net, info->id);
4707         if (!nh)
4708                 return NOTIFY_DONE;
4709
4710         vxlan_fdb_nh_flush(nh);
4711
4712         return NOTIFY_DONE;
4713 }
4714
4715 static __net_init int vxlan_init_net(struct net *net)
4716 {
4717         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4718         unsigned int h;
4719
4720         INIT_LIST_HEAD(&vn->vxlan_list);
4721         spin_lock_init(&vn->sock_lock);
4722         vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event;
4723
4724         for (h = 0; h < PORT_HASH_SIZE; ++h)
4725                 INIT_HLIST_HEAD(&vn->sock_list[h]);
4726
4727         return register_nexthop_notifier(net, &vn->nexthop_notifier_block,
4728                                          NULL);
4729 }
4730
4731 static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
4732 {
4733         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4734         struct vxlan_dev *vxlan, *next;
4735         struct net_device *dev, *aux;
4736
4737         for_each_netdev_safe(net, dev, aux)
4738                 if (dev->rtnl_link_ops == &vxlan_link_ops)
4739                         unregister_netdevice_queue(dev, head);
4740
4741         list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4742                 /* If vxlan->dev is in the same netns, it has already been added
4743                  * to the list by the previous loop.
4744                  */
4745                 if (!net_eq(dev_net(vxlan->dev), net))
4746                         unregister_netdevice_queue(vxlan->dev, head);
4747         }
4748
4749 }
4750
4751 static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4752 {
4753         struct net *net;
4754         LIST_HEAD(list);
4755         unsigned int h;
4756
4757         rtnl_lock();
4758         list_for_each_entry(net, net_list, exit_list) {
4759                 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4760
4761                 unregister_nexthop_notifier(net, &vn->nexthop_notifier_block);
4762         }
4763         list_for_each_entry(net, net_list, exit_list)
4764                 vxlan_destroy_tunnels(net, &list);
4765
4766         unregister_netdevice_many(&list);
4767         rtnl_unlock();
4768
4769         list_for_each_entry(net, net_list, exit_list) {
4770                 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4771
4772                 for (h = 0; h < PORT_HASH_SIZE; ++h)
4773                         WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4774         }
4775 }
4776
4777 static struct pernet_operations vxlan_net_ops = {
4778         .init = vxlan_init_net,
4779         .exit_batch = vxlan_exit_batch_net,
4780         .id   = &vxlan_net_id,
4781         .size = sizeof(struct vxlan_net),
4782 };
4783
4784 static int __init vxlan_init_module(void)
4785 {
4786         int rc;
4787
4788         get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4789
4790         rc = register_pernet_subsys(&vxlan_net_ops);
4791         if (rc)
4792                 goto out1;
4793
4794         rc = register_netdevice_notifier(&vxlan_notifier_block);
4795         if (rc)
4796                 goto out2;
4797
4798         rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4799         if (rc)
4800                 goto out3;
4801
4802         rc = rtnl_link_register(&vxlan_link_ops);
4803         if (rc)
4804                 goto out4;
4805
4806         return 0;
4807 out4:
4808         unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4809 out3:
4810         unregister_netdevice_notifier(&vxlan_notifier_block);
4811 out2:
4812         unregister_pernet_subsys(&vxlan_net_ops);
4813 out1:
4814         return rc;
4815 }
4816 late_initcall(vxlan_init_module);
4817
4818 static void __exit vxlan_cleanup_module(void)
4819 {
4820         rtnl_link_unregister(&vxlan_link_ops);
4821         unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4822         unregister_netdevice_notifier(&vxlan_notifier_block);
4823         unregister_pernet_subsys(&vxlan_net_ops);
4824         /* rcu_barrier() is called by netns */
4825 }
4826 module_exit(vxlan_cleanup_module);
4827
4828 MODULE_LICENSE("GPL");
4829 MODULE_VERSION(VXLAN_VERSION);
4830 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4831 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4832 MODULE_ALIAS_RTNL_LINK("vxlan");