drivers/net/vxlan/vxlan_core.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/gro.h>
  21 #include <net/ipv6_stubs.h>
  22 #include <net/ip.h>
  23 #include <net/icmp.h>
  24 #include <net/rtnetlink.h>
  25 #include <net/inet_ecn.h>
  26 #include <net/net_namespace.h>
  27 #include <net/netns/generic.h>
  28 #include <net/tun_proto.h>
  29 #include <net/vxlan.h>
  30 #include <net/nexthop.h>
  31
  32 #if IS_ENABLED(CONFIG_IPV6)
  33 #include <net/ip6_tunnel.h>
  34 #include <net/ip6_checksum.h>
  35 #endif
  36
  37 #include "vxlan_private.h"
  38
  39 #define VXLAN_VERSION   "0.1"
  40
  41 #define FDB_AGE_DEFAULT 300 /* 5 min */
  42 #define FDB_AGE_INTERVAL (10 * HZ)      /* rescan interval */
  43
  44 /* UDP port for VXLAN traffic.
  45  * The IANA assigned port is 4789, but the Linux default is 8472
  46  * for compatibility with early adopters.
  47  */
  48 static unsigned short vxlan_port __read_mostly = 8472;
  49 module_param_named(udp_port, vxlan_port, ushort, 0444);
  50 MODULE_PARM_DESC(udp_port, "Destination UDP port");
  51
  52 static bool log_ecn_error = true;
  53 module_param(log_ecn_error, bool, 0644);
  54 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  55
  56 unsigned int vxlan_net_id;
  57
  58 const u8 all_zeros_mac[ETH_ALEN + 2];
  59 static struct rtnl_link_ops vxlan_link_ops;
  60
  61 static int vxlan_sock_add(struct vxlan_dev *vxlan);
  62
  63 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
  64
  65 /* salt for hash table */
  66 static u32 vxlan_salt __read_mostly;
  67
  68 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
  69 {
  70         return vs->flags & VXLAN_F_COLLECT_METADATA ||
  71                ip_tunnel_collect_metadata();
  72 }
  73
  74 #if IS_ENABLED(CONFIG_IPV6)
  75 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
  76 {
  77         if (nla_len(nla) >= sizeof(struct in6_addr)) {
  78                 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
  79                 ip->sa.sa_family = AF_INET6;
  80                 return 0;
  81         } else if (nla_len(nla) >= sizeof(__be32)) {
  82                 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
  83                 ip->sa.sa_family = AF_INET;
  84                 return 0;
  85         } else {
  86                 return -EAFNOSUPPORT;
  87         }
  88 }
  89
  90 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
  91                               const union vxlan_addr *ip)
  92 {
  93         if (ip->sa.sa_family == AF_INET6)
  94                 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
  95         else
  96                 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
  97 }
  98
  99 #else /* !CONFIG_IPV6 */
 100
 101 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
 102 {
 103         if (nla_len(nla) >= sizeof(struct in6_addr)) {
 104                 return -EAFNOSUPPORT;
 105         } else if (nla_len(nla) >= sizeof(__be32)) {
 106                 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
 107                 ip->sa.sa_family = AF_INET;
 108                 return 0;
 109         } else {
 110                 return -EAFNOSUPPORT;
 111         }
 112 }
 113
 114 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
 115                               const union vxlan_addr *ip)
 116 {
 117         return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
 118 }
 119 #endif
 120
 121 /* Find VXLAN socket based on network namespace, address family, UDP port,
 122  * enabled unshareable flags and socket device binding (see l3mdev with
 123  * non-default VRF).
 124  */
 125 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
 126                                           __be16 port, u32 flags, int ifindex)
 127 {
 128         struct vxlan_sock *vs;
 129
 130         flags &= VXLAN_F_RCV_FLAGS;
 131
 132         hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
 133                 if (inet_sk(vs->sock->sk)->inet_sport == port &&
 134                     vxlan_get_sk_family(vs) == family &&
 135                     vs->flags == flags &&
 136                     vs->sock->sk->sk_bound_dev_if == ifindex)
 137                         return vs;
 138         }
 139         return NULL;
 140 }
 141
 142 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs,
 143                                            int ifindex, __be32 vni,
 144                                            struct vxlan_vni_node **vninode)
 145 {
 146         struct vxlan_vni_node *vnode;
 147         struct vxlan_dev_node *node;
 148
 149         /* For flow based devices, map all packets to VNI 0 */
 150         if (vs->flags & VXLAN_F_COLLECT_METADATA &&
 151             !(vs->flags & VXLAN_F_VNIFILTER))
 152                 vni = 0;
 153
 154         hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
 155                 if (!node->vxlan)
 156                         continue;
 157                 vnode = NULL;
 158                 if (node->vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
 159                         vnode = vxlan_vnifilter_lookup(node->vxlan, vni);
 160                         if (!vnode)
 161                                 continue;
 162                 } else if (node->vxlan->default_dst.remote_vni != vni) {
 163                         continue;
 164                 }
 165
 166                 if (IS_ENABLED(CONFIG_IPV6)) {
 167                         const struct vxlan_config *cfg = &node->vxlan->cfg;
 168
 169                         if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
 170                             cfg->remote_ifindex != ifindex)
 171                                 continue;
 172                 }
 173
 174                 if (vninode)
 175                         *vninode = vnode;
 176                 return node->vxlan;
 177         }
 178
 179         return NULL;
 180 }
 181
 182 /* Look up VNI in a per net namespace table */
 183 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
 184                                         __be32 vni, sa_family_t family,
 185                                         __be16 port, u32 flags)
 186 {
 187         struct vxlan_sock *vs;
 188
 189         vs = vxlan_find_sock(net, family, port, flags, ifindex);
 190         if (!vs)
 191                 return NULL;
 192
 193         return vxlan_vs_find_vni(vs, ifindex, vni, NULL);
 194 }
 195
 196 /* Fill in neighbour message in skbuff. */
 197 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
 198                           const struct vxlan_fdb *fdb,
 199                           u32 portid, u32 seq, int type, unsigned int flags,
 200                           const struct vxlan_rdst *rdst)
 201 {
 202         unsigned long now = jiffies;
 203         struct nda_cacheinfo ci;
 204         bool send_ip, send_eth;
 205         struct nlmsghdr *nlh;
 206         struct nexthop *nh;
 207         struct ndmsg *ndm;
 208         int nh_family;
 209         u32 nh_id;
 210
 211         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
 212         if (nlh == NULL)
 213                 return -EMSGSIZE;
 214
 215         ndm = nlmsg_data(nlh);
 216         memset(ndm, 0, sizeof(*ndm));
 217
 218         send_eth = send_ip = true;
 219
 220         rcu_read_lock();
 221         nh = rcu_dereference(fdb->nh);
 222         if (nh) {
 223                 nh_family = nexthop_get_family(nh);
 224                 nh_id = nh->id;
 225         }
 226         rcu_read_unlock();
 227
 228         if (type == RTM_GETNEIGH) {
 229                 if (rdst) {
 230                         send_ip = !vxlan_addr_any(&rdst->remote_ip);
 231                         ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
 232                 } else if (nh) {
 233                         ndm->ndm_family = nh_family;
 234                 }
 235                 send_eth = !is_zero_ether_addr(fdb->eth_addr);
 236         } else
 237                 ndm->ndm_family = AF_BRIDGE;
 238         ndm->ndm_state = fdb->state;
 239         ndm->ndm_ifindex = vxlan->dev->ifindex;
 240         ndm->ndm_flags = fdb->flags;
 241         if (rdst && rdst->offloaded)
 242                 ndm->ndm_flags |= NTF_OFFLOADED;
 243         ndm->ndm_type = RTN_UNICAST;
 244
 245         if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
 246             nla_put_s32(skb, NDA_LINK_NETNSID,
 247                         peernet2id(dev_net(vxlan->dev), vxlan->net)))
 248                 goto nla_put_failure;
 249
 250         if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
 251                 goto nla_put_failure;
 252         if (nh) {
 253                 if (nla_put_u32(skb, NDA_NH_ID, nh_id))
 254                         goto nla_put_failure;
 255         } else if (rdst) {
 256                 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST,
 257                                                   &rdst->remote_ip))
 258                         goto nla_put_failure;
 259
 260                 if (rdst->remote_port &&
 261                     rdst->remote_port != vxlan->cfg.dst_port &&
 262                     nla_put_be16(skb, NDA_PORT, rdst->remote_port))
 263                         goto nla_put_failure;
 264                 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
 265                     nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
 266                         goto nla_put_failure;
 267                 if (rdst->remote_ifindex &&
 268                     nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
 269                         goto nla_put_failure;
 270         }
 271
 272         if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
 273             nla_put_u32(skb, NDA_SRC_VNI,
 274                         be32_to_cpu(fdb->vni)))
 275                 goto nla_put_failure;
 276
 277         ci.ndm_used      = jiffies_to_clock_t(now - fdb->used);
 278         ci.ndm_confirmed = 0;
 279         ci.ndm_updated   = jiffies_to_clock_t(now - fdb->updated);
 280         ci.ndm_refcnt    = 0;
 281
 282         if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
 283                 goto nla_put_failure;
 284
 285         nlmsg_end(skb, nlh);
 286         return 0;
 287
 288 nla_put_failure:
 289         nlmsg_cancel(skb, nlh);
 290         return -EMSGSIZE;
 291 }
 292
 293 static inline size_t vxlan_nlmsg_size(void)
 294 {
 295         return NLMSG_ALIGN(sizeof(struct ndmsg))
 296                 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
 297                 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
 298                 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
 299                 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
 300                 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
 301                 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
 302                 + nla_total_size(sizeof(struct nda_cacheinfo));
 303 }
 304
 305 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
 306                                struct vxlan_rdst *rd, int type)
 307 {
 308         struct net *net = dev_net(vxlan->dev);
 309         struct sk_buff *skb;
 310         int err = -ENOBUFS;
 311
 312         skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
 313         if (skb == NULL)
 314                 goto errout;
 315
 316         err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
 317         if (err < 0) {
 318                 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
 319                 WARN_ON(err == -EMSGSIZE);
 320                 kfree_skb(skb);
 321                 goto errout;
 322         }
 323
 324         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
 325         return;
 326 errout:
 327         if (err < 0)
 328                 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
 329 }
 330
 331 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
 332                             const struct vxlan_fdb *fdb,
 333                             const struct vxlan_rdst *rd,
 334                             struct netlink_ext_ack *extack,
 335                             struct switchdev_notifier_vxlan_fdb_info *fdb_info)
 336 {
 337         fdb_info->info.dev = vxlan->dev;
 338         fdb_info->info.extack = extack;
 339         fdb_info->remote_ip = rd->remote_ip;
 340         fdb_info->remote_port = rd->remote_port;
 341         fdb_info->remote_vni = rd->remote_vni;
 342         fdb_info->remote_ifindex = rd->remote_ifindex;
 343         memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
 344         fdb_info->vni = fdb->vni;
 345         fdb_info->offloaded = rd->offloaded;
 346         fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
 347 }
 348
 349 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
 350                                               struct vxlan_fdb *fdb,
 351                                               struct vxlan_rdst *rd,
 352                                               bool adding,
 353                                               struct netlink_ext_ack *extack)
 354 {
 355         struct switchdev_notifier_vxlan_fdb_info info;
 356         enum switchdev_notifier_type notifier_type;
 357         int ret;
 358
 359         if (WARN_ON(!rd))
 360                 return 0;
 361
 362         notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
 363                                : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
 364         vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
 365         ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
 366                                        &info.info, extack);
 367         return notifier_to_errno(ret);
 368 }
 369
 370 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
 371                             struct vxlan_rdst *rd, int type, bool swdev_notify,
 372                             struct netlink_ext_ack *extack)
 373 {
 374         int err;
 375
 376         if (swdev_notify && rd) {
 377                 switch (type) {
 378                 case RTM_NEWNEIGH:
 379                         err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
 380                                                                  true, extack);
 381                         if (err)
 382                                 return err;
 383                         break;
 384                 case RTM_DELNEIGH:
 385                         vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
 386                                                            false, extack);
 387                         break;
 388                 }
 389         }
 390
 391         __vxlan_fdb_notify(vxlan, fdb, rd, type);
 392         return 0;
 393 }
 394
 395 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
 396 {
 397         struct vxlan_dev *vxlan = netdev_priv(dev);
 398         struct vxlan_fdb f = {
 399                 .state = NUD_STALE,
 400         };
 401         struct vxlan_rdst remote = {
 402                 .remote_ip = *ipa, /* goes to NDA_DST */
 403                 .remote_vni = cpu_to_be32(VXLAN_N_VID),
 404         };
 405
 406         vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
 407 }
 408
 409 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
 410 {
 411         struct vxlan_fdb f = {
 412                 .state = NUD_STALE,
 413         };
 414         struct vxlan_rdst remote = { };
 415
 416         memcpy(f.eth_addr, eth_addr, ETH_ALEN);
 417
 418         vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
 419 }
 420
 421 /* Hash Ethernet address */
 422 static u32 eth_hash(const unsigned char *addr)
 423 {
 424         u64 value = get_unaligned((u64 *)addr);
 425
 426         /* only want 6 bytes */
 427 #ifdef __BIG_ENDIAN
 428         value >>= 16;
 429 #else
 430         value <<= 16;
 431 #endif
 432         return hash_64(value, FDB_HASH_BITS);
 433 }
 434
 435 u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
 436 {
 437         /* use 1 byte of OUI and 3 bytes of NIC */
 438         u32 key = get_unaligned((u32 *)(addr + 2));
 439
 440         return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
 441 }
 442
 443 u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
 444 {
 445         if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
 446                 return eth_vni_hash(mac, vni);
 447         else
 448                 return eth_hash(mac);
 449 }
 450
 451 /* Hash chain to use given mac address */
 452 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
 453                                                 const u8 *mac, __be32 vni)
 454 {
 455         return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
 456 }
 457
 458 /* Look up Ethernet address in forwarding table */
 459 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
 460                                           const u8 *mac, __be32 vni)
 461 {
 462         struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
 463         struct vxlan_fdb *f;
 464
 465         hlist_for_each_entry_rcu(f, head, hlist) {
 466                 if (ether_addr_equal(mac, f->eth_addr)) {
 467                         if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
 468                                 if (vni == f->vni)
 469                                         return f;
 470                         } else {
 471                                 return f;
 472                         }
 473                 }
 474         }
 475
 476         return NULL;
 477 }
 478
 479 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
 480                                         const u8 *mac, __be32 vni)
 481 {
 482         struct vxlan_fdb *f;
 483
 484         f = __vxlan_find_mac(vxlan, mac, vni);
 485         if (f && f->used != jiffies)
 486                 f->used = jiffies;
 487
 488         return f;
 489 }
 490
 491 /* caller should hold vxlan->hash_lock */
 492 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
 493                                               union vxlan_addr *ip, __be16 port,
 494                                               __be32 vni, __u32 ifindex)
 495 {
 496         struct vxlan_rdst *rd;
 497
 498         list_for_each_entry(rd, &f->remotes, list) {
 499                 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
 500                     rd->remote_port == port &&
 501                     rd->remote_vni == vni &&
 502                     rd->remote_ifindex == ifindex)
 503                         return rd;
 504         }
 505
 506         return NULL;
 507 }
 508
 509 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
 510                       struct switchdev_notifier_vxlan_fdb_info *fdb_info)
 511 {
 512         struct vxlan_dev *vxlan = netdev_priv(dev);
 513         u8 eth_addr[ETH_ALEN + 2] = { 0 };
 514         struct vxlan_rdst *rdst;
 515         struct vxlan_fdb *f;
 516         int rc = 0;
 517
 518         if (is_multicast_ether_addr(mac) ||
 519             is_zero_ether_addr(mac))
 520                 return -EINVAL;
 521
 522         ether_addr_copy(eth_addr, mac);
 523
 524         rcu_read_lock();
 525
 526         f = __vxlan_find_mac(vxlan, eth_addr, vni);
 527         if (!f) {
 528                 rc = -ENOENT;
 529                 goto out;
 530         }
 531
 532         rdst = first_remote_rcu(f);
 533         vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
 534
 535 out:
 536         rcu_read_unlock();
 537         return rc;
 538 }
 539 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
 540
 541 static int vxlan_fdb_notify_one(struct notifier_block *nb,
 542                                 const struct vxlan_dev *vxlan,
 543                                 const struct vxlan_fdb *f,
 544                                 const struct vxlan_rdst *rdst,
 545                                 struct netlink_ext_ack *extack)
 546 {
 547         struct switchdev_notifier_vxlan_fdb_info fdb_info;
 548         int rc;
 549
 550         vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
 551         rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
 552                                &fdb_info);
 553         return notifier_to_errno(rc);
 554 }
 555
 556 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
 557                      struct notifier_block *nb,
 558                      struct netlink_ext_ack *extack)
 559 {
 560         struct vxlan_dev *vxlan;
 561         struct vxlan_rdst *rdst;
 562         struct vxlan_fdb *f;
 563         unsigned int h;
 564         int rc = 0;
 565
 566         if (!netif_is_vxlan(dev))
 567                 return -EINVAL;
 568         vxlan = netdev_priv(dev);
 569
 570         for (h = 0; h < FDB_HASH_SIZE; ++h) {
 571                 spin_lock_bh(&vxlan->hash_lock[h]);
 572                 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
 573                         if (f->vni == vni) {
 574                                 list_for_each_entry(rdst, &f->remotes, list) {
 575                                         rc = vxlan_fdb_notify_one(nb, vxlan,
 576                                                                   f, rdst,
 577                                                                   extack);
 578                                         if (rc)
 579                                                 goto unlock;
 580                                 }
 581                         }
 582                 }
 583                 spin_unlock_bh(&vxlan->hash_lock[h]);
 584         }
 585         return 0;
 586
 587 unlock:
 588         spin_unlock_bh(&vxlan->hash_lock[h]);
 589         return rc;
 590 }
 591 EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
 592
 593 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
 594 {
 595         struct vxlan_dev *vxlan;
 596         struct vxlan_rdst *rdst;
 597         struct vxlan_fdb *f;
 598         unsigned int h;
 599
 600         if (!netif_is_vxlan(dev))
 601                 return;
 602         vxlan = netdev_priv(dev);
 603
 604         for (h = 0; h < FDB_HASH_SIZE; ++h) {
 605                 spin_lock_bh(&vxlan->hash_lock[h]);
 606                 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
 607                         if (f->vni == vni)
 608                                 list_for_each_entry(rdst, &f->remotes, list)
 609                                         rdst->offloaded = false;
 610                 spin_unlock_bh(&vxlan->hash_lock[h]);
 611         }
 612
 613 }
 614 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
 615
 616 /* Replace destination of unicast mac */
 617 static int vxlan_fdb_replace(struct vxlan_fdb *f,
 618                              union vxlan_addr *ip, __be16 port, __be32 vni,
 619                              __u32 ifindex, struct vxlan_rdst *oldrd)
 620 {
 621         struct vxlan_rdst *rd;
 622
 623         rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
 624         if (rd)
 625                 return 0;
 626
 627         rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
 628         if (!rd)
 629                 return 0;
 630
 631         *oldrd = *rd;
 632         dst_cache_reset(&rd->dst_cache);
 633         rd->remote_ip = *ip;
 634         rd->remote_port = port;
 635         rd->remote_vni = vni;
 636         rd->remote_ifindex = ifindex;
 637         rd->offloaded = false;
 638         return 1;
 639 }
 640
 641 /* Add/update destinations for multicast */
 642 static int vxlan_fdb_append(struct vxlan_fdb *f,
 643                             union vxlan_addr *ip, __be16 port, __be32 vni,
 644                             __u32 ifindex, struct vxlan_rdst **rdp)
 645 {
 646         struct vxlan_rdst *rd;
 647
 648         rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
 649         if (rd)
 650                 return 0;
 651
 652         rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
 653         if (rd == NULL)
 654                 return -ENOMEM;
 655
 656         if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
 657                 kfree(rd);
 658                 return -ENOMEM;
 659         }
 660
 661         rd->remote_ip = *ip;
 662         rd->remote_port = port;
 663         rd->offloaded = false;
 664         rd->remote_vni = vni;
 665         rd->remote_ifindex = ifindex;
 666
 667         list_add_tail_rcu(&rd->list, &f->remotes);
 668
 669         *rdp = rd;
 670         return 1;
 671 }
 672
 673 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
 674                                           unsigned int off,
 675                                           struct vxlanhdr *vh, size_t hdrlen,
 676                                           __be32 vni_field,
 677                                           struct gro_remcsum *grc,
 678                                           bool nopartial)
 679 {
 680         size_t start, offset;
 681
 682         if (skb->remcsum_offload)
 683                 return vh;
 684
 685         if (!NAPI_GRO_CB(skb)->csum_valid)
 686                 return NULL;
 687
 688         start = vxlan_rco_start(vni_field);
 689         offset = start + vxlan_rco_offset(vni_field);
 690
 691         vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
 692                                      start, offset, grc, nopartial);
 693
 694         skb->remcsum_offload = 1;
 695
 696         return vh;
 697 }
 698
 699 static struct sk_buff *vxlan_gro_receive(struct sock *sk,
 700                                          struct list_head *head,
 701                                          struct sk_buff *skb)
 702 {
 703         struct sk_buff *pp = NULL;
 704         struct sk_buff *p;
 705         struct vxlanhdr *vh, *vh2;
 706         unsigned int hlen, off_vx;
 707         int flush = 1;
 708         struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
 709         __be32 flags;
 710         struct gro_remcsum grc;
 711
 712         skb_gro_remcsum_init(&grc);
 713
 714         off_vx = skb_gro_offset(skb);
 715         hlen = off_vx + sizeof(*vh);
 716         vh = skb_gro_header(skb, hlen, off_vx);
 717         if (unlikely(!vh))
 718                 goto out;
 719
 720         skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
 721
 722         flags = vh->vx_flags;
 723
 724         if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
 725                 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
 726                                        vh->vx_vni, &grc,
 727                                        !!(vs->flags &
 728                                           VXLAN_F_REMCSUM_NOPARTIAL));
 729
 730                 if (!vh)
 731                         goto out;
 732         }
 733
 734         skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
 735
 736         list_for_each_entry(p, head, list) {
 737                 if (!NAPI_GRO_CB(p)->same_flow)
 738                         continue;
 739
 740                 vh2 = (struct vxlanhdr *)(p->data + off_vx);
 741                 if (vh->vx_flags != vh2->vx_flags ||
 742                     vh->vx_vni != vh2->vx_vni) {
 743                         NAPI_GRO_CB(p)->same_flow = 0;
 744                         continue;
 745                 }
 746         }
 747
 748         pp = call_gro_receive(eth_gro_receive, head, skb);
 749         flush = 0;
 750
 751 out:
 752         skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
 753
 754         return pp;
 755 }
 756
 757 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
 758 {
 759         /* Sets 'skb->inner_mac_header' since we are always called with
 760          * 'skb->encapsulation' set.
 761          */
 762         return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
 763 }
 764
 765 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
 766                                          __u16 state, __be32 src_vni,
 767                                          __u16 ndm_flags)
 768 {
 769         struct vxlan_fdb *f;
 770
 771         f = kmalloc(sizeof(*f), GFP_ATOMIC);
 772         if (!f)
 773                 return NULL;
 774         f->state = state;
 775         f->flags = ndm_flags;
 776         f->updated = f->used = jiffies;
 777         f->vni = src_vni;
 778         f->nh = NULL;
 779         RCU_INIT_POINTER(f->vdev, vxlan);
 780         INIT_LIST_HEAD(&f->nh_list);
 781         INIT_LIST_HEAD(&f->remotes);
 782         memcpy(f->eth_addr, mac, ETH_ALEN);
 783
 784         return f;
 785 }
 786
 787 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
 788                              __be32 src_vni, struct vxlan_fdb *f)
 789 {
 790         ++vxlan->addrcnt;
 791         hlist_add_head_rcu(&f->hlist,
 792                            vxlan_fdb_head(vxlan, mac, src_vni));
 793 }
 794
 795 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
 796                                u32 nhid, struct netlink_ext_ack *extack)
 797 {
 798         struct nexthop *old_nh = rtnl_dereference(fdb->nh);
 799         struct nexthop *nh;
 800         int err = -EINVAL;
 801
 802         if (old_nh && old_nh->id == nhid)
 803                 return 0;
 804
 805         nh = nexthop_find_by_id(vxlan->net, nhid);
 806         if (!nh) {
 807                 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
 808                 goto err_inval;
 809         }
 810
 811         if (!nexthop_get(nh)) {
 812                 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
 813                 nh = NULL;
 814                 goto err_inval;
 815         }
 816         if (!nexthop_is_fdb(nh)) {
 817                 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
 818                 goto err_inval;
 819         }
 820
 821         if (!nexthop_is_multipath(nh)) {
 822                 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
 823                 goto err_inval;
 824         }
 825
 826         /* check nexthop group family */
 827         switch (vxlan->default_dst.remote_ip.sa.sa_family) {
 828         case AF_INET:
 829                 if (!nexthop_has_v4(nh)) {
 830                         err = -EAFNOSUPPORT;
 831                         NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
 832                         goto err_inval;
 833                 }
 834                 break;
 835         case AF_INET6:
 836                 if (nexthop_has_v4(nh)) {
 837                         err = -EAFNOSUPPORT;
 838                         NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
 839                         goto err_inval;
 840                 }
 841         }
 842
 843         if (old_nh) {
 844                 list_del_rcu(&fdb->nh_list);
 845                 nexthop_put(old_nh);
 846         }
 847         rcu_assign_pointer(fdb->nh, nh);
 848         list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
 849         return 1;
 850
 851 err_inval:
 852         if (nh)
 853                 nexthop_put(nh);
 854         return err;
 855 }
 856
 857 int vxlan_fdb_create(struct vxlan_dev *vxlan,
 858                      const u8 *mac, union vxlan_addr *ip,
 859                      __u16 state, __be16 port, __be32 src_vni,
 860                      __be32 vni, __u32 ifindex, __u16 ndm_flags,
 861                      u32 nhid, struct vxlan_fdb **fdb,
 862                      struct netlink_ext_ack *extack)
 863 {
 864         struct vxlan_rdst *rd = NULL;
 865         struct vxlan_fdb *f;
 866         int rc;
 867
 868         if (vxlan->cfg.addrmax &&
 869             vxlan->addrcnt >= vxlan->cfg.addrmax)
 870                 return -ENOSPC;
 871
 872         netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
 873         f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
 874         if (!f)
 875                 return -ENOMEM;
 876
 877         if (nhid)
 878                 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
 879         else
 880                 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
 881         if (rc < 0)
 882                 goto errout;
 883
 884         *fdb = f;
 885
 886         return 0;
 887
 888 errout:
 889         kfree(f);
 890         return rc;
 891 }
 892
 893 static void __vxlan_fdb_free(struct vxlan_fdb *f)
 894 {
 895         struct vxlan_rdst *rd, *nd;
 896         struct nexthop *nh;
 897
 898         nh = rcu_dereference_raw(f->nh);
 899         if (nh) {
 900                 rcu_assign_pointer(f->nh, NULL);
 901                 rcu_assign_pointer(f->vdev, NULL);
 902                 nexthop_put(nh);
 903         }
 904
 905         list_for_each_entry_safe(rd, nd, &f->remotes, list) {
 906                 dst_cache_destroy(&rd->dst_cache);
 907                 kfree(rd);
 908         }
 909         kfree(f);
 910 }
 911
 912 static void vxlan_fdb_free(struct rcu_head *head)
 913 {
 914         struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
 915
 916         __vxlan_fdb_free(f);
 917 }
 918
 919 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
 920                               bool do_notify, bool swdev_notify)
 921 {
 922         struct vxlan_rdst *rd;
 923
 924         netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
 925
 926         --vxlan->addrcnt;
 927         if (do_notify) {
 928                 if (rcu_access_pointer(f->nh))
 929                         vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
 930                                          swdev_notify, NULL);
 931                 else
 932                         list_for_each_entry(rd, &f->remotes, list)
 933                                 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
 934                                                  swdev_notify, NULL);
 935         }
 936
 937         hlist_del_rcu(&f->hlist);
 938         list_del_rcu(&f->nh_list);
 939         call_rcu(&f->rcu, vxlan_fdb_free);
 940 }
 941
 942 static void vxlan_dst_free(struct rcu_head *head)
 943 {
 944         struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
 945
 946         dst_cache_destroy(&rd->dst_cache);
 947         kfree(rd);
 948 }
 949
 950 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
 951                                      union vxlan_addr *ip,
 952                                      __u16 state, __u16 flags,
 953                                      __be16 port, __be32 vni,
 954                                      __u32 ifindex, __u16 ndm_flags,
 955                                      struct vxlan_fdb *f, u32 nhid,
 956                                      bool swdev_notify,
 957                                      struct netlink_ext_ack *extack)
 958 {
 959         __u16 fdb_flags = (ndm_flags & ~NTF_USE);
 960         struct vxlan_rdst *rd = NULL;
 961         struct vxlan_rdst oldrd;
 962         int notify = 0;
 963         int rc = 0;
 964         int err;
 965
 966         if (nhid && !rcu_access_pointer(f->nh)) {
 967                 NL_SET_ERR_MSG(extack,
 968                                "Cannot replace an existing non nexthop fdb with a nexthop");
 969                 return -EOPNOTSUPP;
 970         }
 971
 972         if (nhid && (flags & NLM_F_APPEND)) {
 973                 NL_SET_ERR_MSG(extack,
 974                                "Cannot append to a nexthop fdb");
 975                 return -EOPNOTSUPP;
 976         }
 977
 978         /* Do not allow an externally learned entry to take over an entry added
 979          * by the user.
 980          */
 981         if (!(fdb_flags & NTF_EXT_LEARNED) ||
 982             !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
 983                 if (f->state != state) {
 984                         f->state = state;
 985                         f->updated = jiffies;
 986                         notify = 1;
 987                 }
 988                 if (f->flags != fdb_flags) {
 989                         f->flags = fdb_flags;
 990                         f->updated = jiffies;
 991                         notify = 1;
 992                 }
 993         }
 994
 995         if ((flags & NLM_F_REPLACE)) {
 996                 /* Only change unicasts */
 997                 if (!(is_multicast_ether_addr(f->eth_addr) ||
 998                       is_zero_ether_addr(f->eth_addr))) {
 999                         if (nhid) {
1000                                 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1001                                 if (rc < 0)
1002                                         return rc;
1003                         } else {
1004                                 rc = vxlan_fdb_replace(f, ip, port, vni,
1005                                                        ifindex, &oldrd);
1006                         }
1007                         notify |= rc;
1008                 } else {
1009                         NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
1010                         return -EOPNOTSUPP;
1011                 }
1012         }
1013         if ((flags & NLM_F_APPEND) &&
1014             (is_multicast_ether_addr(f->eth_addr) ||
1015              is_zero_ether_addr(f->eth_addr))) {
1016                 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1017
1018                 if (rc < 0)
1019                         return rc;
1020                 notify |= rc;
1021         }
1022
1023         if (ndm_flags & NTF_USE)
1024                 f->used = jiffies;
1025
1026         if (notify) {
1027                 if (rd == NULL)
1028                         rd = first_remote_rtnl(f);
1029
1030                 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
1031                                        swdev_notify, extack);
1032                 if (err)
1033                         goto err_notify;
1034         }
1035
1036         return 0;
1037
1038 err_notify:
1039         if (nhid)
1040                 return err;
1041         if ((flags & NLM_F_REPLACE) && rc)
1042                 *rd = oldrd;
1043         else if ((flags & NLM_F_APPEND) && rc) {
1044                 list_del_rcu(&rd->list);
1045                 call_rcu(&rd->rcu, vxlan_dst_free);
1046         }
1047         return err;
1048 }
1049
1050 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
1051                                    const u8 *mac, union vxlan_addr *ip,
1052                                    __u16 state, __u16 flags,
1053                                    __be16 port, __be32 src_vni, __be32 vni,
1054                                    __u32 ifindex, __u16 ndm_flags, u32 nhid,
1055                                    bool swdev_notify,
1056                                    struct netlink_ext_ack *extack)
1057 {
1058         __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1059         struct vxlan_fdb *f;
1060         int rc;
1061
1062         /* Disallow replace to add a multicast entry */
1063         if ((flags & NLM_F_REPLACE) &&
1064             (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1065                 return -EOPNOTSUPP;
1066
1067         netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1068         rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1069                               vni, ifindex, fdb_flags, nhid, &f, extack);
1070         if (rc < 0)
1071                 return rc;
1072
1073         vxlan_fdb_insert(vxlan, mac, src_vni, f);
1074         rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1075                               swdev_notify, extack);
1076         if (rc)
1077                 goto err_notify;
1078
1079         return 0;
1080
1081 err_notify:
1082         vxlan_fdb_destroy(vxlan, f, false, false);
1083         return rc;
1084 }
1085
1086 /* Add new entry to forwarding table -- assumes lock held */
1087 int vxlan_fdb_update(struct vxlan_dev *vxlan,
1088                      const u8 *mac, union vxlan_addr *ip,
1089                      __u16 state, __u16 flags,
1090                      __be16 port, __be32 src_vni, __be32 vni,
1091                      __u32 ifindex, __u16 ndm_flags, u32 nhid,
1092                      bool swdev_notify,
1093                      struct netlink_ext_ack *extack)
1094 {
1095         struct vxlan_fdb *f;
1096
1097         f = __vxlan_find_mac(vxlan, mac, src_vni);
1098         if (f) {
1099                 if (flags & NLM_F_EXCL) {
1100                         netdev_dbg(vxlan->dev,
1101                                    "lost race to create %pM\n", mac);
1102                         return -EEXIST;
1103                 }
1104
1105                 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1106                                                  vni, ifindex, ndm_flags, f,
1107                                                  nhid, swdev_notify, extack);
1108         } else {
1109                 if (!(flags & NLM_F_CREATE))
1110                         return -ENOENT;
1111
1112                 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1113                                                port, src_vni, vni, ifindex,
1114                                                ndm_flags, nhid, swdev_notify,
1115                                                extack);
1116         }
1117 }
1118
1119 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1120                                   struct vxlan_rdst *rd, bool swdev_notify)
1121 {
1122         list_del_rcu(&rd->list);
1123         vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1124         call_rcu(&rd->rcu, vxlan_dst_free);
1125 }
1126
1127 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1128                            union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1129                            __be32 *vni, u32 *ifindex, u32 *nhid,
1130                            struct netlink_ext_ack *extack)
1131 {
1132         struct net *net = dev_net(vxlan->dev);
1133         int err;
1134
1135         if (tb[NDA_NH_ID] &&
1136             (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) {
1137                 NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID");
1138                 return -EINVAL;
1139         }
1140
1141         if (tb[NDA_DST]) {
1142                 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1143                 if (err) {
1144                         NL_SET_ERR_MSG(extack, "Unsupported address family");
1145                         return err;
1146                 }
1147         } else {
1148                 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1149
1150                 if (remote->sa.sa_family == AF_INET) {
1151                         ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1152                         ip->sa.sa_family = AF_INET;
1153 #if IS_ENABLED(CONFIG_IPV6)
1154                 } else {
1155                         ip->sin6.sin6_addr = in6addr_any;
1156                         ip->sa.sa_family = AF_INET6;
1157 #endif
1158                 }
1159         }
1160
1161         if (tb[NDA_PORT]) {
1162                 if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) {
1163                         NL_SET_ERR_MSG(extack, "Invalid vxlan port");
1164                         return -EINVAL;
1165                 }
1166                 *port = nla_get_be16(tb[NDA_PORT]);
1167         } else {
1168                 *port = vxlan->cfg.dst_port;
1169         }
1170
1171         if (tb[NDA_VNI]) {
1172                 if (nla_len(tb[NDA_VNI]) != sizeof(u32)) {
1173                         NL_SET_ERR_MSG(extack, "Invalid vni");
1174                         return -EINVAL;
1175                 }
1176                 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1177         } else {
1178                 *vni = vxlan->default_dst.remote_vni;
1179         }
1180
1181         if (tb[NDA_SRC_VNI]) {
1182                 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) {
1183                         NL_SET_ERR_MSG(extack, "Invalid src vni");
1184                         return -EINVAL;
1185                 }
1186                 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1187         } else {
1188                 *src_vni = vxlan->default_dst.remote_vni;
1189         }
1190
1191         if (tb[NDA_IFINDEX]) {
1192                 struct net_device *tdev;
1193
1194                 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) {
1195                         NL_SET_ERR_MSG(extack, "Invalid ifindex");
1196                         return -EINVAL;
1197                 }
1198                 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1199                 tdev = __dev_get_by_index(net, *ifindex);
1200                 if (!tdev) {
1201                         NL_SET_ERR_MSG(extack, "Device not found");
1202                         return -EADDRNOTAVAIL;
1203                 }
1204         } else {
1205                 *ifindex = 0;
1206         }
1207
1208         if (tb[NDA_NH_ID])
1209                 *nhid = nla_get_u32(tb[NDA_NH_ID]);
1210         else
1211                 *nhid = 0;
1212
1213         return 0;
1214 }
1215
1216 /* Add static entry (via netlink) */
1217 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1218                          struct net_device *dev,
1219                          const unsigned char *addr, u16 vid, u16 flags,
1220                          struct netlink_ext_ack *extack)
1221 {
1222         struct vxlan_dev *vxlan = netdev_priv(dev);
1223         /* struct net *net = dev_net(vxlan->dev); */
1224         union vxlan_addr ip;
1225         __be16 port;
1226         __be32 src_vni, vni;
1227         u32 ifindex, nhid;
1228         u32 hash_index;
1229         int err;
1230
1231         if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1232                 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1233                         ndm->ndm_state);
1234                 return -EINVAL;
1235         }
1236
1237         if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
1238                 return -EINVAL;
1239
1240         err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1241                               &nhid, extack);
1242         if (err)
1243                 return err;
1244
1245         if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1246                 return -EAFNOSUPPORT;
1247
1248         hash_index = fdb_head_index(vxlan, addr, src_vni);
1249         spin_lock_bh(&vxlan->hash_lock[hash_index]);
1250         err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1251                                port, src_vni, vni, ifindex,
1252                                ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1253                                nhid, true, extack);
1254         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1255
1256         return err;
1257 }
1258
1259 int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1260                        const unsigned char *addr, union vxlan_addr ip,
1261                        __be16 port, __be32 src_vni, __be32 vni,
1262                        u32 ifindex, bool swdev_notify)
1263 {
1264         struct vxlan_rdst *rd = NULL;
1265         struct vxlan_fdb *f;
1266         int err = -ENOENT;
1267
1268         f = vxlan_find_mac(vxlan, addr, src_vni);
1269         if (!f)
1270                 return err;
1271
1272         if (!vxlan_addr_any(&ip)) {
1273                 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1274                 if (!rd)
1275                         goto out;
1276         }
1277
1278         /* remove a destination if it's not the only one on the list,
1279          * otherwise destroy the fdb entry
1280          */
1281         if (rd && !list_is_singular(&f->remotes)) {
1282                 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1283                 goto out;
1284         }
1285
1286         vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1287
1288 out:
1289         return 0;
1290 }
1291
1292 /* Delete entry (via netlink) */
1293 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1294                             struct net_device *dev,
1295                             const unsigned char *addr, u16 vid,
1296                             struct netlink_ext_ack *extack)
1297 {
1298         struct vxlan_dev *vxlan = netdev_priv(dev);
1299         union vxlan_addr ip;
1300         __be32 src_vni, vni;
1301         u32 ifindex, nhid;
1302         u32 hash_index;
1303         __be16 port;
1304         int err;
1305
1306         err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1307                               &nhid, extack);
1308         if (err)
1309                 return err;
1310
1311         hash_index = fdb_head_index(vxlan, addr, src_vni);
1312         spin_lock_bh(&vxlan->hash_lock[hash_index]);
1313         err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1314                                  true);
1315         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1316
1317         return err;
1318 }
1319
1320 /* Dump forwarding table */
1321 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1322                           struct net_device *dev,
1323                           struct net_device *filter_dev, int *idx)
1324 {
1325         struct vxlan_dev *vxlan = netdev_priv(dev);
1326         unsigned int h;
1327         int err = 0;
1328
1329         for (h = 0; h < FDB_HASH_SIZE; ++h) {
1330                 struct vxlan_fdb *f;
1331
1332                 rcu_read_lock();
1333                 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1334                         struct vxlan_rdst *rd;
1335
1336                         if (rcu_access_pointer(f->nh)) {
1337                                 if (*idx < cb->args[2])
1338                                         goto skip_nh;
1339                                 err = vxlan_fdb_info(skb, vxlan, f,
1340                                                      NETLINK_CB(cb->skb).portid,
1341                                                      cb->nlh->nlmsg_seq,
1342                                                      RTM_NEWNEIGH,
1343                                                      NLM_F_MULTI, NULL);
1344                                 if (err < 0) {
1345                                         rcu_read_unlock();
1346                                         goto out;
1347                                 }
1348 skip_nh:
1349                                 *idx += 1;
1350                                 continue;
1351                         }
1352
1353                         list_for_each_entry_rcu(rd, &f->remotes, list) {
1354                                 if (*idx < cb->args[2])
1355                                         goto skip;
1356
1357                                 err = vxlan_fdb_info(skb, vxlan, f,
1358                                                      NETLINK_CB(cb->skb).portid,
1359                                                      cb->nlh->nlmsg_seq,
1360                                                      RTM_NEWNEIGH,
1361                                                      NLM_F_MULTI, rd);
1362                                 if (err < 0) {
1363                                         rcu_read_unlock();
1364                                         goto out;
1365                                 }
1366 skip:
1367                                 *idx += 1;
1368                         }
1369                 }
1370                 rcu_read_unlock();
1371         }
1372 out:
1373         return err;
1374 }
1375
1376 static int vxlan_fdb_get(struct sk_buff *skb,
1377                          struct nlattr *tb[],
1378                          struct net_device *dev,
1379                          const unsigned char *addr,
1380                          u16 vid, u32 portid, u32 seq,
1381                          struct netlink_ext_ack *extack)
1382 {
1383         struct vxlan_dev *vxlan = netdev_priv(dev);
1384         struct vxlan_fdb *f;
1385         __be32 vni;
1386         int err;
1387
1388         if (tb[NDA_VNI])
1389                 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1390         else
1391                 vni = vxlan->default_dst.remote_vni;
1392
1393         rcu_read_lock();
1394
1395         f = __vxlan_find_mac(vxlan, addr, vni);
1396         if (!f) {
1397                 NL_SET_ERR_MSG(extack, "Fdb entry not found");
1398                 err = -ENOENT;
1399                 goto errout;
1400         }
1401
1402         err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1403                              RTM_NEWNEIGH, 0, first_remote_rcu(f));
1404 errout:
1405         rcu_read_unlock();
1406         return err;
1407 }
1408
1409 /* Watch incoming packets to learn mapping between Ethernet address
1410  * and Tunnel endpoint.
1411  * Return true if packet is bogus and should be dropped.
1412  */
1413 static bool vxlan_snoop(struct net_device *dev,
1414                         union vxlan_addr *src_ip, const u8 *src_mac,
1415                         u32 src_ifindex, __be32 vni)
1416 {
1417         struct vxlan_dev *vxlan = netdev_priv(dev);
1418         struct vxlan_fdb *f;
1419         u32 ifindex = 0;
1420
1421 #if IS_ENABLED(CONFIG_IPV6)
1422         if (src_ip->sa.sa_family == AF_INET6 &&
1423             (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1424                 ifindex = src_ifindex;
1425 #endif
1426
1427         f = vxlan_find_mac(vxlan, src_mac, vni);
1428         if (likely(f)) {
1429                 struct vxlan_rdst *rdst = first_remote_rcu(f);
1430
1431                 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1432                            rdst->remote_ifindex == ifindex))
1433                         return false;
1434
1435                 /* Don't migrate static entries, drop packets */
1436                 if (f->state & (NUD_PERMANENT | NUD_NOARP))
1437                         return true;
1438
1439                 /* Don't override an fdb with nexthop with a learnt entry */
1440                 if (rcu_access_pointer(f->nh))
1441                         return true;
1442
1443                 if (net_ratelimit())
1444                         netdev_info(dev,
1445                                     "%pM migrated from %pIS to %pIS\n",
1446                                     src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1447
1448                 rdst->remote_ip = *src_ip;
1449                 f->updated = jiffies;
1450                 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1451         } else {
1452                 u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1453
1454                 /* learned new entry */
1455                 spin_lock(&vxlan->hash_lock[hash_index]);
1456
1457                 /* close off race between vxlan_flush and incoming packets */
1458                 if (netif_running(dev))
1459                         vxlan_fdb_update(vxlan, src_mac, src_ip,
1460                                          NUD_REACHABLE,
1461                                          NLM_F_EXCL|NLM_F_CREATE,
1462                                          vxlan->cfg.dst_port,
1463                                          vni,
1464                                          vxlan->default_dst.remote_vni,
1465                                          ifindex, NTF_SELF, 0, true, NULL);
1466                 spin_unlock(&vxlan->hash_lock[hash_index]);
1467         }
1468
1469         return false;
1470 }
1471
1472 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1473 {
1474         struct vxlan_net *vn;
1475
1476         if (!vs)
1477                 return false;
1478         if (!refcount_dec_and_test(&vs->refcnt))
1479                 return false;
1480
1481         vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1482         spin_lock(&vn->sock_lock);
1483         hlist_del_rcu(&vs->hlist);
1484         udp_tunnel_notify_del_rx_port(vs->sock,
1485                                       (vs->flags & VXLAN_F_GPE) ?
1486                                       UDP_TUNNEL_TYPE_VXLAN_GPE :
1487                                       UDP_TUNNEL_TYPE_VXLAN);
1488         spin_unlock(&vn->sock_lock);
1489
1490         return true;
1491 }
1492
1493 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1494 {
1495         struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1496 #if IS_ENABLED(CONFIG_IPV6)
1497         struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1498
1499         RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1500 #endif
1501
1502         RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1503         synchronize_net();
1504
1505         if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
1506                 vxlan_vs_del_vnigrp(vxlan);
1507         else
1508                 vxlan_vs_del_dev(vxlan);
1509
1510         if (__vxlan_sock_release_prep(sock4)) {
1511                 udp_tunnel_sock_release(sock4->sock);
1512                 kfree(sock4);
1513         }
1514
1515 #if IS_ENABLED(CONFIG_IPV6)
1516         if (__vxlan_sock_release_prep(sock6)) {
1517                 udp_tunnel_sock_release(sock6->sock);
1518                 kfree(sock6);
1519         }
1520 #endif
1521 }
1522
1523 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1524                           struct sk_buff *skb, u32 vxflags)
1525 {
1526         size_t start, offset;
1527
1528         if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1529                 goto out;
1530
1531         start = vxlan_rco_start(unparsed->vx_vni);
1532         offset = start + vxlan_rco_offset(unparsed->vx_vni);
1533
1534         if (!pskb_may_pull(skb, offset + sizeof(u16)))
1535                 return false;
1536
1537         skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1538                             !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1539 out:
1540         unparsed->vx_flags &= ~VXLAN_HF_RCO;
1541         unparsed->vx_vni &= VXLAN_VNI_MASK;
1542         return true;
1543 }
1544
1545 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1546                                 struct sk_buff *skb, u32 vxflags,
1547                                 struct vxlan_metadata *md)
1548 {
1549         struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1550         struct metadata_dst *tun_dst;
1551
1552         if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1553                 goto out;
1554
1555         md->gbp = ntohs(gbp->policy_id);
1556
1557         tun_dst = (struct metadata_dst *)skb_dst(skb);
1558         if (tun_dst) {
1559                 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1560                 tun_dst->u.tun_info.options_len = sizeof(*md);
1561         }
1562         if (gbp->dont_learn)
1563                 md->gbp |= VXLAN_GBP_DONT_LEARN;
1564
1565         if (gbp->policy_applied)
1566                 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1567
1568         /* In flow-based mode, GBP is carried in dst_metadata */
1569         if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1570                 skb->mark = md->gbp;
1571 out:
1572         unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1573 }
1574
1575 static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1576                                 __be16 *protocol,
1577                                 struct sk_buff *skb, u32 vxflags)
1578 {
1579         struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1580
1581         /* Need to have Next Protocol set for interfaces in GPE mode. */
1582         if (!gpe->np_applied)
1583                 return false;
1584         /* "The initial version is 0. If a receiver does not support the
1585          * version indicated it MUST drop the packet.
1586          */
1587         if (gpe->version != 0)
1588                 return false;
1589         /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1590          * processing MUST occur." However, we don't implement OAM
1591          * processing, thus drop the packet.
1592          */
1593         if (gpe->oam_flag)
1594                 return false;
1595
1596         *protocol = tun_p_to_eth_p(gpe->next_protocol);
1597         if (!*protocol)
1598                 return false;
1599
1600         unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1601         return true;
1602 }
1603
1604 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1605                           struct vxlan_sock *vs,
1606                           struct sk_buff *skb, __be32 vni)
1607 {
1608         union vxlan_addr saddr;
1609         u32 ifindex = skb->dev->ifindex;
1610
1611         skb_reset_mac_header(skb);
1612         skb->protocol = eth_type_trans(skb, vxlan->dev);
1613         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1614
1615         /* Ignore packet loops (and multicast echo) */
1616         if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1617                 return false;
1618
1619         /* Get address from the outer IP header */
1620         if (vxlan_get_sk_family(vs) == AF_INET) {
1621                 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1622                 saddr.sa.sa_family = AF_INET;
1623 #if IS_ENABLED(CONFIG_IPV6)
1624         } else {
1625                 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1626                 saddr.sa.sa_family = AF_INET6;
1627 #endif
1628         }
1629
1630         if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1631             vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1632                 return false;
1633
1634         return true;
1635 }
1636
1637 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1638                                   struct sk_buff *skb)
1639 {
1640         int err = 0;
1641
1642         if (vxlan_get_sk_family(vs) == AF_INET)
1643                 err = IP_ECN_decapsulate(oiph, skb);
1644 #if IS_ENABLED(CONFIG_IPV6)
1645         else
1646                 err = IP6_ECN_decapsulate(oiph, skb);
1647 #endif
1648
1649         if (unlikely(err) && log_ecn_error) {
1650                 if (vxlan_get_sk_family(vs) == AF_INET)
1651                         net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1652                                              &((struct iphdr *)oiph)->saddr,
1653                                              ((struct iphdr *)oiph)->tos);
1654                 else
1655                         net_info_ratelimited("non-ECT from %pI6\n",
1656                                              &((struct ipv6hdr *)oiph)->saddr);
1657         }
1658         return err <= 1;
1659 }
1660
1661 /* Callback from net/ipv4/udp.c to receive packets */
1662 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1663 {
1664         struct vxlan_vni_node *vninode = NULL;
1665         struct vxlan_dev *vxlan;
1666         struct vxlan_sock *vs;
1667         struct vxlanhdr unparsed;
1668         struct vxlan_metadata _md;
1669         struct vxlan_metadata *md = &_md;
1670         __be16 protocol = htons(ETH_P_TEB);
1671         bool raw_proto = false;
1672         void *oiph;
1673         __be32 vni = 0;
1674
1675         /* Need UDP and VXLAN header to be present */
1676         if (!pskb_may_pull(skb, VXLAN_HLEN))
1677                 goto drop;
1678
1679         unparsed = *vxlan_hdr(skb);
1680         /* VNI flag always required to be set */
1681         if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1682                 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1683                            ntohl(vxlan_hdr(skb)->vx_flags),
1684                            ntohl(vxlan_hdr(skb)->vx_vni));
1685                 /* Return non vxlan pkt */
1686                 goto drop;
1687         }
1688         unparsed.vx_flags &= ~VXLAN_HF_VNI;
1689         unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1690
1691         vs = rcu_dereference_sk_user_data(sk);
1692         if (!vs)
1693                 goto drop;
1694
1695         vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1696
1697         vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode);
1698         if (!vxlan)
1699                 goto drop;
1700
1701         /* For backwards compatibility, only allow reserved fields to be
1702          * used by VXLAN extensions if explicitly requested.
1703          */
1704         if (vs->flags & VXLAN_F_GPE) {
1705                 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1706                         goto drop;
1707                 raw_proto = true;
1708         }
1709
1710         if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1711                                    !net_eq(vxlan->net, dev_net(vxlan->dev))))
1712                 goto drop;
1713
1714         if (vs->flags & VXLAN_F_REMCSUM_RX)
1715                 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags)))
1716                         goto drop;
1717
1718         if (vxlan_collect_metadata(vs)) {
1719                 struct metadata_dst *tun_dst;
1720
1721                 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1722                                          key32_to_tunnel_id(vni), sizeof(*md));
1723
1724                 if (!tun_dst)
1725                         goto drop;
1726
1727                 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1728
1729                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1730         } else {
1731                 memset(md, 0, sizeof(*md));
1732         }
1733
1734         if (vs->flags & VXLAN_F_GBP)
1735                 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1736         /* Note that GBP and GPE can never be active together. This is
1737          * ensured in vxlan_dev_configure.
1738          */
1739
1740         if (unparsed.vx_flags || unparsed.vx_vni) {
1741                 /* If there are any unprocessed flags remaining treat
1742                  * this as a malformed packet. This behavior diverges from
1743                  * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1744                  * in reserved fields are to be ignored. The approach here
1745                  * maintains compatibility with previous stack code, and also
1746                  * is more robust and provides a little more security in
1747                  * adding extensions to VXLAN.
1748                  */
1749                 goto drop;
1750         }
1751
1752         if (!raw_proto) {
1753                 if (!vxlan_set_mac(vxlan, vs, skb, vni))
1754                         goto drop;
1755         } else {
1756                 skb_reset_mac_header(skb);
1757                 skb->dev = vxlan->dev;
1758                 skb->pkt_type = PACKET_HOST;
1759         }
1760
1761         oiph = skb_network_header(skb);
1762         skb_reset_network_header(skb);
1763
1764         if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1765                 ++vxlan->dev->stats.rx_frame_errors;
1766                 ++vxlan->dev->stats.rx_errors;
1767                 vxlan_vnifilter_count(vxlan, vni, vninode,
1768                                       VXLAN_VNI_STATS_RX_ERRORS, 0);
1769                 goto drop;
1770         }
1771
1772         rcu_read_lock();
1773
1774         if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1775                 rcu_read_unlock();
1776                 dev_core_stats_rx_dropped_inc(vxlan->dev);
1777                 vxlan_vnifilter_count(vxlan, vni, vninode,
1778                                       VXLAN_VNI_STATS_RX_DROPS, 0);
1779                 goto drop;
1780         }
1781
1782         dev_sw_netstats_rx_add(vxlan->dev, skb->len);
1783         vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len);
1784         gro_cells_receive(&vxlan->gro_cells, skb);
1785
1786         rcu_read_unlock();
1787
1788         return 0;
1789
1790 drop:
1791         /* Consume bad packet */
1792         kfree_skb(skb);
1793         return 0;
1794 }
1795
1796 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1797 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1798 {
1799         struct vxlan_dev *vxlan;
1800         struct vxlan_sock *vs;
1801         struct vxlanhdr *hdr;
1802         __be32 vni;
1803
1804         if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1805                 return -EINVAL;
1806
1807         hdr = vxlan_hdr(skb);
1808
1809         if (!(hdr->vx_flags & VXLAN_HF_VNI))
1810                 return -EINVAL;
1811
1812         vs = rcu_dereference_sk_user_data(sk);
1813         if (!vs)
1814                 return -ENOENT;
1815
1816         vni = vxlan_vni(hdr->vx_vni);
1817         vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL);
1818         if (!vxlan)
1819                 return -ENOENT;
1820
1821         return 0;
1822 }
1823
1824 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1825 {
1826         struct vxlan_dev *vxlan = netdev_priv(dev);
1827         struct arphdr *parp;
1828         u8 *arpptr, *sha;
1829         __be32 sip, tip;
1830         struct neighbour *n;
1831
1832         if (dev->flags & IFF_NOARP)
1833                 goto out;
1834
1835         if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1836                 dev->stats.tx_dropped++;
1837                 goto out;
1838         }
1839         parp = arp_hdr(skb);
1840
1841         if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1842              parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1843             parp->ar_pro != htons(ETH_P_IP) ||
1844             parp->ar_op != htons(ARPOP_REQUEST) ||
1845             parp->ar_hln != dev->addr_len ||
1846             parp->ar_pln != 4)
1847                 goto out;
1848         arpptr = (u8 *)parp + sizeof(struct arphdr);
1849         sha = arpptr;
1850         arpptr += dev->addr_len;        /* sha */
1851         memcpy(&sip, arpptr, sizeof(sip));
1852         arpptr += sizeof(sip);
1853         arpptr += dev->addr_len;        /* tha */
1854         memcpy(&tip, arpptr, sizeof(tip));
1855
1856         if (ipv4_is_loopback(tip) ||
1857             ipv4_is_multicast(tip))
1858                 goto out;
1859
1860         n = neigh_lookup(&arp_tbl, &tip, dev);
1861
1862         if (n) {
1863                 struct vxlan_fdb *f;
1864                 struct sk_buff  *reply;
1865
1866                 if (!(n->nud_state & NUD_CONNECTED)) {
1867                         neigh_release(n);
1868                         goto out;
1869                 }
1870
1871                 f = vxlan_find_mac(vxlan, n->ha, vni);
1872                 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1873                         /* bridge-local neighbor */
1874                         neigh_release(n);
1875                         goto out;
1876                 }
1877
1878                 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1879                                 n->ha, sha);
1880
1881                 neigh_release(n);
1882
1883                 if (reply == NULL)
1884                         goto out;
1885
1886                 skb_reset_mac_header(reply);
1887                 __skb_pull(reply, skb_network_offset(reply));
1888                 reply->ip_summed = CHECKSUM_UNNECESSARY;
1889                 reply->pkt_type = PACKET_HOST;
1890
1891                 if (netif_rx(reply) == NET_RX_DROP) {
1892                         dev->stats.rx_dropped++;
1893                         vxlan_vnifilter_count(vxlan, vni, NULL,
1894                                               VXLAN_VNI_STATS_RX_DROPS, 0);
1895                 }
1896
1897         } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1898                 union vxlan_addr ipa = {
1899                         .sin.sin_addr.s_addr = tip,
1900                         .sin.sin_family = AF_INET,
1901                 };
1902
1903                 vxlan_ip_miss(dev, &ipa);
1904         }
1905 out:
1906         consume_skb(skb);
1907         return NETDEV_TX_OK;
1908 }
1909
1910 #if IS_ENABLED(CONFIG_IPV6)
1911 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1912         struct neighbour *n, bool isrouter)
1913 {
1914         struct net_device *dev = request->dev;
1915         struct sk_buff *reply;
1916         struct nd_msg *ns, *na;
1917         struct ipv6hdr *pip6;
1918         u8 *daddr;
1919         int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1920         int ns_olen;
1921         int i, len;
1922
1923         if (dev == NULL || !pskb_may_pull(request, request->len))
1924                 return NULL;
1925
1926         len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1927                 sizeof(*na) + na_olen + dev->needed_tailroom;
1928         reply = alloc_skb(len, GFP_ATOMIC);
1929         if (reply == NULL)
1930                 return NULL;
1931
1932         reply->protocol = htons(ETH_P_IPV6);
1933         reply->dev = dev;
1934         skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1935         skb_push(reply, sizeof(struct ethhdr));
1936         skb_reset_mac_header(reply);
1937
1938         ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
1939
1940         daddr = eth_hdr(request)->h_source;
1941         ns_olen = request->len - skb_network_offset(request) -
1942                 sizeof(struct ipv6hdr) - sizeof(*ns);
1943         for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1944                 if (!ns->opt[i + 1]) {
1945                         kfree_skb(reply);
1946                         return NULL;
1947                 }
1948                 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1949                         daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1950                         break;
1951                 }
1952         }
1953
1954         /* Ethernet header */
1955         ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1956         ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1957         eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1958         reply->protocol = htons(ETH_P_IPV6);
1959
1960         skb_pull(reply, sizeof(struct ethhdr));
1961         skb_reset_network_header(reply);
1962         skb_put(reply, sizeof(struct ipv6hdr));
1963
1964         /* IPv6 header */
1965
1966         pip6 = ipv6_hdr(reply);
1967         memset(pip6, 0, sizeof(struct ipv6hdr));
1968         pip6->version = 6;
1969         pip6->priority = ipv6_hdr(request)->priority;
1970         pip6->nexthdr = IPPROTO_ICMPV6;
1971         pip6->hop_limit = 255;
1972         pip6->daddr = ipv6_hdr(request)->saddr;
1973         pip6->saddr = *(struct in6_addr *)n->primary_key;
1974
1975         skb_pull(reply, sizeof(struct ipv6hdr));
1976         skb_reset_transport_header(reply);
1977
1978         /* Neighbor Advertisement */
1979         na = skb_put_zero(reply, sizeof(*na) + na_olen);
1980         na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1981         na->icmph.icmp6_router = isrouter;
1982         na->icmph.icmp6_override = 1;
1983         na->icmph.icmp6_solicited = 1;
1984         na->target = ns->target;
1985         ether_addr_copy(&na->opt[2], n->ha);
1986         na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1987         na->opt[1] = na_olen >> 3;
1988
1989         na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1990                 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1991                 csum_partial(na, sizeof(*na)+na_olen, 0));
1992
1993         pip6->payload_len = htons(sizeof(*na)+na_olen);
1994
1995         skb_push(reply, sizeof(struct ipv6hdr));
1996
1997         reply->ip_summed = CHECKSUM_UNNECESSARY;
1998
1999         return reply;
2000 }
2001
2002 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2003 {
2004         struct vxlan_dev *vxlan = netdev_priv(dev);
2005         const struct in6_addr *daddr;
2006         const struct ipv6hdr *iphdr;
2007         struct inet6_dev *in6_dev;
2008         struct neighbour *n;
2009         struct nd_msg *msg;
2010
2011         rcu_read_lock();
2012         in6_dev = __in6_dev_get(dev);
2013         if (!in6_dev)
2014                 goto out;
2015
2016         iphdr = ipv6_hdr(skb);
2017         daddr = &iphdr->daddr;
2018         msg = (struct nd_msg *)(iphdr + 1);
2019
2020         if (ipv6_addr_loopback(daddr) ||
2021             ipv6_addr_is_multicast(&msg->target))
2022                 goto out;
2023
2024         n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2025
2026         if (n) {
2027                 struct vxlan_fdb *f;
2028                 struct sk_buff *reply;
2029
2030                 if (!(n->nud_state & NUD_CONNECTED)) {
2031                         neigh_release(n);
2032                         goto out;
2033                 }
2034
2035                 f = vxlan_find_mac(vxlan, n->ha, vni);
2036                 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2037                         /* bridge-local neighbor */
2038                         neigh_release(n);
2039                         goto out;
2040                 }
2041
2042                 reply = vxlan_na_create(skb, n,
2043                                         !!(f ? f->flags & NTF_ROUTER : 0));
2044
2045                 neigh_release(n);
2046
2047                 if (reply == NULL)
2048                         goto out;
2049
2050                 if (netif_rx(reply) == NET_RX_DROP) {
2051                         dev->stats.rx_dropped++;
2052                         vxlan_vnifilter_count(vxlan, vni, NULL,
2053                                               VXLAN_VNI_STATS_RX_DROPS, 0);
2054                 }
2055         } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2056                 union vxlan_addr ipa = {
2057                         .sin6.sin6_addr = msg->target,
2058                         .sin6.sin6_family = AF_INET6,
2059                 };
2060
2061                 vxlan_ip_miss(dev, &ipa);
2062         }
2063
2064 out:
2065         rcu_read_unlock();
2066         consume_skb(skb);
2067         return NETDEV_TX_OK;
2068 }
2069 #endif
2070
2071 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2072 {
2073         struct vxlan_dev *vxlan = netdev_priv(dev);
2074         struct neighbour *n;
2075
2076         if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2077                 return false;
2078
2079         n = NULL;
2080         switch (ntohs(eth_hdr(skb)->h_proto)) {
2081         case ETH_P_IP:
2082         {
2083                 struct iphdr *pip;
2084
2085                 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2086                         return false;
2087                 pip = ip_hdr(skb);
2088                 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2089                 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2090                         union vxlan_addr ipa = {
2091                                 .sin.sin_addr.s_addr = pip->daddr,
2092                                 .sin.sin_family = AF_INET,
2093                         };
2094
2095                         vxlan_ip_miss(dev, &ipa);
2096                         return false;
2097                 }
2098
2099                 break;
2100         }
2101 #if IS_ENABLED(CONFIG_IPV6)
2102         case ETH_P_IPV6:
2103         {
2104                 struct ipv6hdr *pip6;
2105
2106                 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2107                         return false;
2108                 pip6 = ipv6_hdr(skb);
2109                 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2110                 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2111                         union vxlan_addr ipa = {
2112                                 .sin6.sin6_addr = pip6->daddr,
2113                                 .sin6.sin6_family = AF_INET6,
2114                         };
2115
2116                         vxlan_ip_miss(dev, &ipa);
2117                         return false;
2118                 }
2119
2120                 break;
2121         }
2122 #endif
2123         default:
2124                 return false;
2125         }
2126
2127         if (n) {
2128                 bool diff;
2129
2130                 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2131                 if (diff) {
2132                         memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2133                                 dev->addr_len);
2134                         memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2135                 }
2136                 neigh_release(n);
2137                 return diff;
2138         }
2139
2140         return false;
2141 }
2142
2143 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
2144                                 struct vxlan_metadata *md)
2145 {
2146         struct vxlanhdr_gbp *gbp;
2147
2148         if (!md->gbp)
2149                 return;
2150
2151         gbp = (struct vxlanhdr_gbp *)vxh;
2152         vxh->vx_flags |= VXLAN_HF_GBP;
2153
2154         if (md->gbp & VXLAN_GBP_DONT_LEARN)
2155                 gbp->dont_learn = 1;
2156
2157         if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
2158                 gbp->policy_applied = 1;
2159
2160         gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
2161 }
2162
2163 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
2164                                __be16 protocol)
2165 {
2166         struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2167
2168         gpe->np_applied = 1;
2169         gpe->next_protocol = tun_p_from_eth_p(protocol);
2170         if (!gpe->next_protocol)
2171                 return -EPFNOSUPPORT;
2172         return 0;
2173 }
2174
2175 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2176                            int iphdr_len, __be32 vni,
2177                            struct vxlan_metadata *md, u32 vxflags,
2178                            bool udp_sum)
2179 {
2180         struct vxlanhdr *vxh;
2181         int min_headroom;
2182         int err;
2183         int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2184         __be16 inner_protocol = htons(ETH_P_TEB);
2185
2186         if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2187             skb->ip_summed == CHECKSUM_PARTIAL) {
2188                 int csum_start = skb_checksum_start_offset(skb);
2189
2190                 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2191                     !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2192                     (skb->csum_offset == offsetof(struct udphdr, check) ||
2193                      skb->csum_offset == offsetof(struct tcphdr, check)))
2194                         type |= SKB_GSO_TUNNEL_REMCSUM;
2195         }
2196
2197         min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2198                         + VXLAN_HLEN + iphdr_len;
2199
2200         /* Need space for new headers (invalidates iph ptr) */
2201         err = skb_cow_head(skb, min_headroom);
2202         if (unlikely(err))
2203                 return err;
2204
2205         err = iptunnel_handle_offloads(skb, type);
2206         if (err)
2207                 return err;
2208
2209         vxh = __skb_push(skb, sizeof(*vxh));
2210         vxh->vx_flags = VXLAN_HF_VNI;
2211         vxh->vx_vni = vxlan_vni_field(vni);
2212
2213         if (type & SKB_GSO_TUNNEL_REMCSUM) {
2214                 unsigned int start;
2215
2216                 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2217                 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2218                 vxh->vx_flags |= VXLAN_HF_RCO;
2219
2220                 if (!skb_is_gso(skb)) {
2221                         skb->ip_summed = CHECKSUM_NONE;
2222                         skb->encapsulation = 0;
2223                 }
2224         }
2225
2226         if (vxflags & VXLAN_F_GBP)
2227                 vxlan_build_gbp_hdr(vxh, vxflags, md);
2228         if (vxflags & VXLAN_F_GPE) {
2229                 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2230                 if (err < 0)
2231                         return err;
2232                 inner_protocol = skb->protocol;
2233         }
2234
2235         skb_set_inner_protocol(skb, inner_protocol);
2236         return 0;
2237 }
2238
2239 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2240                                       struct vxlan_sock *sock4,
2241                                       struct sk_buff *skb, int oif, u8 tos,
2242                                       __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
2243                                       __u8 flow_flags, struct dst_cache *dst_cache,
2244                                       const struct ip_tunnel_info *info)
2245 {
2246         bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2247         struct rtable *rt = NULL;
2248         struct flowi4 fl4;
2249
2250         if (!sock4)
2251                 return ERR_PTR(-EIO);
2252
2253         if (tos && !info)
2254                 use_cache = false;
2255         if (use_cache) {
2256                 rt = dst_cache_get_ip4(dst_cache, saddr);
2257                 if (rt)
2258                         return rt;
2259         }
2260
2261         memset(&fl4, 0, sizeof(fl4));
2262         fl4.flowi4_oif = oif;
2263         fl4.flowi4_tos = RT_TOS(tos);
2264         fl4.flowi4_mark = skb->mark;
2265         fl4.flowi4_proto = IPPROTO_UDP;
2266         fl4.daddr = daddr;
2267         fl4.saddr = *saddr;
2268         fl4.fl4_dport = dport;
2269         fl4.fl4_sport = sport;
2270         fl4.flowi4_flags = flow_flags;
2271
2272         rt = ip_route_output_key(vxlan->net, &fl4);
2273         if (!IS_ERR(rt)) {
2274                 if (rt->dst.dev == dev) {
2275                         netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2276                         ip_rt_put(rt);
2277                         return ERR_PTR(-ELOOP);
2278                 }
2279
2280                 *saddr = fl4.saddr;
2281                 if (use_cache)
2282                         dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2283         } else {
2284                 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2285                 return ERR_PTR(-ENETUNREACH);
2286         }
2287         return rt;
2288 }
2289
2290 #if IS_ENABLED(CONFIG_IPV6)
2291 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2292                                           struct net_device *dev,
2293                                           struct vxlan_sock *sock6,
2294                                           struct sk_buff *skb, int oif, u8 tos,
2295                                           __be32 label,
2296                                           const struct in6_addr *daddr,
2297                                           struct in6_addr *saddr,
2298                                           __be16 dport, __be16 sport,
2299                                           struct dst_cache *dst_cache,
2300                                           const struct ip_tunnel_info *info)
2301 {
2302         bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2303         struct dst_entry *ndst;
2304         struct flowi6 fl6;
2305
2306         if (!sock6)
2307                 return ERR_PTR(-EIO);
2308
2309         if (tos && !info)
2310                 use_cache = false;
2311         if (use_cache) {
2312                 ndst = dst_cache_get_ip6(dst_cache, saddr);
2313                 if (ndst)
2314                         return ndst;
2315         }
2316
2317         memset(&fl6, 0, sizeof(fl6));
2318         fl6.flowi6_oif = oif;
2319         fl6.daddr = *daddr;
2320         fl6.saddr = *saddr;
2321         fl6.flowlabel = ip6_make_flowinfo(tos, label);
2322         fl6.flowi6_mark = skb->mark;
2323         fl6.flowi6_proto = IPPROTO_UDP;
2324         fl6.fl6_dport = dport;
2325         fl6.fl6_sport = sport;
2326
2327         ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
2328                                                &fl6, NULL);
2329         if (IS_ERR(ndst)) {
2330                 netdev_dbg(dev, "no route to %pI6\n", daddr);
2331                 return ERR_PTR(-ENETUNREACH);
2332         }
2333
2334         if (unlikely(ndst->dev == dev)) {
2335                 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2336                 dst_release(ndst);
2337                 return ERR_PTR(-ELOOP);
2338         }
2339
2340         *saddr = fl6.saddr;
2341         if (use_cache)
2342                 dst_cache_set_ip6(dst_cache, ndst, saddr);
2343         return ndst;
2344 }
2345 #endif
2346
2347 /* Bypass encapsulation if the destination is local */
2348 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2349                                struct vxlan_dev *dst_vxlan, __be32 vni,
2350                                bool snoop)
2351 {
2352         struct pcpu_sw_netstats *tx_stats, *rx_stats;
2353         union vxlan_addr loopback;
2354         union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2355         struct net_device *dev;
2356         int len = skb->len;
2357
2358         tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2359         rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2360         skb->pkt_type = PACKET_HOST;
2361         skb->encapsulation = 0;
2362         skb->dev = dst_vxlan->dev;
2363         __skb_pull(skb, skb_network_offset(skb));
2364
2365         if (remote_ip->sa.sa_family == AF_INET) {
2366                 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2367                 loopback.sa.sa_family =  AF_INET;
2368 #if IS_ENABLED(CONFIG_IPV6)
2369         } else {
2370                 loopback.sin6.sin6_addr = in6addr_loopback;
2371                 loopback.sa.sa_family =  AF_INET6;
2372 #endif
2373         }
2374
2375         rcu_read_lock();
2376         dev = skb->dev;
2377         if (unlikely(!(dev->flags & IFF_UP))) {
2378                 kfree_skb(skb);
2379                 goto drop;
2380         }
2381
2382         if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
2383                 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2384
2385         u64_stats_update_begin(&tx_stats->syncp);
2386         u64_stats_inc(&tx_stats->tx_packets);
2387         u64_stats_add(&tx_stats->tx_bytes, len);
2388         u64_stats_update_end(&tx_stats->syncp);
2389         vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len);
2390
2391         if (__netif_rx(skb) == NET_RX_SUCCESS) {
2392                 u64_stats_update_begin(&rx_stats->syncp);
2393                 u64_stats_inc(&rx_stats->rx_packets);
2394                 u64_stats_add(&rx_stats->rx_bytes, len);
2395                 u64_stats_update_end(&rx_stats->syncp);
2396                 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX,
2397                                       len);
2398         } else {
2399 drop:
2400                 dev->stats.rx_dropped++;
2401                 vxlan_vnifilter_count(dst_vxlan, vni, NULL,
2402                                       VXLAN_VNI_STATS_RX_DROPS, 0);
2403         }
2404         rcu_read_unlock();
2405 }
2406
2407 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2408                                  struct vxlan_dev *vxlan,
2409                                  union vxlan_addr *daddr,
2410                                  __be16 dst_port, int dst_ifindex, __be32 vni,
2411                                  struct dst_entry *dst,
2412                                  u32 rt_flags)
2413 {
2414 #if IS_ENABLED(CONFIG_IPV6)
2415         /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2416          * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2417          * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2418          */
2419         BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2420 #endif
2421         /* Bypass encapsulation if the destination is local */
2422         if (rt_flags & RTCF_LOCAL &&
2423             !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2424                 struct vxlan_dev *dst_vxlan;
2425
2426                 dst_release(dst);
2427                 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2428                                            daddr->sa.sa_family, dst_port,
2429                                            vxlan->cfg.flags);
2430                 if (!dst_vxlan) {
2431                         dev->stats.tx_errors++;
2432                         vxlan_vnifilter_count(vxlan, vni, NULL,
2433                                               VXLAN_VNI_STATS_TX_ERRORS, 0);
2434                         kfree_skb(skb);
2435
2436                         return -ENOENT;
2437                 }
2438                 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
2439                 return 1;
2440         }
2441
2442         return 0;
2443 }
2444
2445 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2446                            __be32 default_vni, struct vxlan_rdst *rdst,
2447                            bool did_rsc)
2448 {
2449         struct dst_cache *dst_cache;
2450         struct ip_tunnel_info *info;
2451         struct vxlan_dev *vxlan = netdev_priv(dev);
2452         const struct iphdr *old_iph = ip_hdr(skb);
2453         union vxlan_addr *dst;
2454         union vxlan_addr remote_ip, local_ip;
2455         struct vxlan_metadata _md;
2456         struct vxlan_metadata *md = &_md;
2457         unsigned int pkt_len = skb->len;
2458         __be16 src_port = 0, dst_port;
2459         struct dst_entry *ndst = NULL;
2460         __u8 tos, ttl, flow_flags = 0;
2461         int ifindex;
2462         int err;
2463         u32 flags = vxlan->cfg.flags;
2464         bool udp_sum = false;
2465         bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2466         __be32 vni = 0;
2467 #if IS_ENABLED(CONFIG_IPV6)
2468         __be32 label;
2469 #endif
2470
2471         info = skb_tunnel_info(skb);
2472
2473         if (rdst) {
2474                 dst = &rdst->remote_ip;
2475                 if (vxlan_addr_any(dst)) {
2476                         if (did_rsc) {
2477                                 /* short-circuited back to local bridge */
2478                                 vxlan_encap_bypass(skb, vxlan, vxlan,
2479                                                    default_vni, true);
2480                                 return;
2481                         }
2482                         goto drop;
2483                 }
2484
2485                 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2486                 vni = (rdst->remote_vni) ? : default_vni;
2487                 ifindex = rdst->remote_ifindex;
2488                 local_ip = vxlan->cfg.saddr;
2489                 dst_cache = &rdst->dst_cache;
2490                 md->gbp = skb->mark;
2491                 if (flags & VXLAN_F_TTL_INHERIT) {
2492                         ttl = ip_tunnel_get_ttl(old_iph, skb);
2493                 } else {
2494                         ttl = vxlan->cfg.ttl;
2495                         if (!ttl && vxlan_addr_multicast(dst))
2496                                 ttl = 1;
2497                 }
2498
2499                 tos = vxlan->cfg.tos;
2500                 if (tos == 1)
2501                         tos = ip_tunnel_get_dsfield(old_iph, skb);
2502
2503                 if (dst->sa.sa_family == AF_INET)
2504                         udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2505                 else
2506                         udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2507 #if IS_ENABLED(CONFIG_IPV6)
2508                 label = vxlan->cfg.label;
2509 #endif
2510         } else {
2511                 if (!info) {
2512                         WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2513                                   dev->name);
2514                         goto drop;
2515                 }
2516                 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2517                 if (remote_ip.sa.sa_family == AF_INET) {
2518                         remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2519                         local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2520                 } else {
2521                         remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2522                         local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2523                 }
2524                 dst = &remote_ip;
2525                 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2526                 flow_flags = info->key.flow_flags;
2527                 vni = tunnel_id_to_key32(info->key.tun_id);
2528                 ifindex = 0;
2529                 dst_cache = &info->dst_cache;
2530                 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2531                         if (info->options_len < sizeof(*md))
2532                                 goto drop;
2533                         md = ip_tunnel_info_opts(info);
2534                 }
2535                 ttl = info->key.ttl;
2536                 tos = info->key.tos;
2537 #if IS_ENABLED(CONFIG_IPV6)
2538                 label = info->key.label;
2539 #endif
2540                 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2541         }
2542         src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2543                                      vxlan->cfg.port_max, true);
2544
2545         rcu_read_lock();
2546         if (dst->sa.sa_family == AF_INET) {
2547                 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2548                 struct rtable *rt;
2549                 __be16 df = 0;
2550
2551                 if (!ifindex)
2552                         ifindex = sock4->sock->sk->sk_bound_dev_if;
2553
2554                 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2555                                      dst->sin.sin_addr.s_addr,
2556                                      &local_ip.sin.sin_addr.s_addr,
2557                                      dst_port, src_port, flow_flags,
2558                                      dst_cache, info);
2559                 if (IS_ERR(rt)) {
2560                         err = PTR_ERR(rt);
2561                         goto tx_error;
2562                 }
2563
2564                 if (!info) {
2565                         /* Bypass encapsulation if the destination is local */
2566                         err = encap_bypass_if_local(skb, dev, vxlan, dst,
2567                                                     dst_port, ifindex, vni,
2568                                                     &rt->dst, rt->rt_flags);
2569                         if (err)
2570                                 goto out_unlock;
2571
2572                         if (vxlan->cfg.df == VXLAN_DF_SET) {
2573                                 df = htons(IP_DF);
2574                         } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2575                                 struct ethhdr *eth = eth_hdr(skb);
2576
2577                                 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2578                                     (ntohs(eth->h_proto) == ETH_P_IP &&
2579                                      old_iph->frag_off & htons(IP_DF)))
2580                                         df = htons(IP_DF);
2581                         }
2582                 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2583                         df = htons(IP_DF);
2584                 }
2585
2586                 ndst = &rt->dst;
2587                 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM,
2588                                             netif_is_any_bridge_port(dev));
2589                 if (err < 0) {
2590                         goto tx_error;
2591                 } else if (err) {
2592                         if (info) {
2593                                 struct ip_tunnel_info *unclone;
2594                                 struct in_addr src, dst;
2595
2596                                 unclone = skb_tunnel_info_unclone(skb);
2597                                 if (unlikely(!unclone))
2598                                         goto tx_error;
2599
2600                                 src = remote_ip.sin.sin_addr;
2601                                 dst = local_ip.sin.sin_addr;
2602                                 unclone->key.u.ipv4.src = src.s_addr;
2603                                 unclone->key.u.ipv4.dst = dst.s_addr;
2604                         }
2605                         vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2606                         dst_release(ndst);
2607                         goto out_unlock;
2608                 }
2609
2610                 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2611                 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2612                 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2613                                       vni, md, flags, udp_sum);
2614                 if (err < 0)
2615                         goto tx_error;
2616
2617                 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2618                                     dst->sin.sin_addr.s_addr, tos, ttl, df,
2619                                     src_port, dst_port, xnet, !udp_sum);
2620 #if IS_ENABLED(CONFIG_IPV6)
2621         } else {
2622                 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2623
2624                 if (!ifindex)
2625                         ifindex = sock6->sock->sk->sk_bound_dev_if;
2626
2627                 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2628                                         label, &dst->sin6.sin6_addr,
2629                                         &local_ip.sin6.sin6_addr,
2630                                         dst_port, src_port,
2631                                         dst_cache, info);
2632                 if (IS_ERR(ndst)) {
2633                         err = PTR_ERR(ndst);
2634                         ndst = NULL;
2635                         goto tx_error;
2636                 }
2637
2638                 if (!info) {
2639                         u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2640
2641                         err = encap_bypass_if_local(skb, dev, vxlan, dst,
2642                                                     dst_port, ifindex, vni,
2643                                                     ndst, rt6i_flags);
2644                         if (err)
2645                                 goto out_unlock;
2646                 }
2647
2648                 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM,
2649                                             netif_is_any_bridge_port(dev));
2650                 if (err < 0) {
2651                         goto tx_error;
2652                 } else if (err) {
2653                         if (info) {
2654                                 struct ip_tunnel_info *unclone;
2655                                 struct in6_addr src, dst;
2656
2657                                 unclone = skb_tunnel_info_unclone(skb);
2658                                 if (unlikely(!unclone))
2659                                         goto tx_error;
2660
2661                                 src = remote_ip.sin6.sin6_addr;
2662                                 dst = local_ip.sin6.sin6_addr;
2663                                 unclone->key.u.ipv6.src = src;
2664                                 unclone->key.u.ipv6.dst = dst;
2665                         }
2666
2667                         vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2668                         dst_release(ndst);
2669                         goto out_unlock;
2670                 }
2671
2672                 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2673                 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2674                 skb_scrub_packet(skb, xnet);
2675                 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2676                                       vni, md, flags, udp_sum);
2677                 if (err < 0)
2678                         goto tx_error;
2679
2680                 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2681                                      &local_ip.sin6.sin6_addr,
2682                                      &dst->sin6.sin6_addr, tos, ttl,
2683                                      label, src_port, dst_port, !udp_sum);
2684 #endif
2685         }
2686         vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len);
2687 out_unlock:
2688         rcu_read_unlock();
2689         return;
2690
2691 drop:
2692         dev->stats.tx_dropped++;
2693         vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0);
2694         dev_kfree_skb(skb);
2695         return;
2696
2697 tx_error:
2698         rcu_read_unlock();
2699         if (err == -ELOOP)
2700                 dev->stats.collisions++;
2701         else if (err == -ENETUNREACH)
2702                 dev->stats.tx_carrier_errors++;
2703         dst_release(ndst);
2704         dev->stats.tx_errors++;
2705         vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0);
2706         kfree_skb(skb);
2707 }
2708
2709 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2710                           struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2711 {
2712         struct vxlan_rdst nh_rdst;
2713         struct nexthop *nh;
2714         bool do_xmit;
2715         u32 hash;
2716
2717         memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2718         hash = skb_get_hash(skb);
2719
2720         rcu_read_lock();
2721         nh = rcu_dereference(f->nh);
2722         if (!nh) {
2723                 rcu_read_unlock();
2724                 goto drop;
2725         }
2726         do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2727         rcu_read_unlock();
2728
2729         if (likely(do_xmit))
2730                 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2731         else
2732                 goto drop;
2733
2734         return;
2735
2736 drop:
2737         dev->stats.tx_dropped++;
2738         vxlan_vnifilter_count(netdev_priv(dev), vni, NULL,
2739                               VXLAN_VNI_STATS_TX_DROPS, 0);
2740         dev_kfree_skb(skb);
2741 }
2742
2743 /* Transmit local packets over Vxlan
2744  *
2745  * Outer IP header inherits ECN and DF from inner header.
2746  * Outer UDP destination is the VXLAN assigned port.
2747  *           source port is based on hash of flow
2748  */
2749 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2750 {
2751         struct vxlan_dev *vxlan = netdev_priv(dev);
2752         struct vxlan_rdst *rdst, *fdst = NULL;
2753         const struct ip_tunnel_info *info;
2754         bool did_rsc = false;
2755         struct vxlan_fdb *f;
2756         struct ethhdr *eth;
2757         __be32 vni = 0;
2758
2759         info = skb_tunnel_info(skb);
2760
2761         skb_reset_mac_header(skb);
2762
2763         if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2764                 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2765                     info->mode & IP_TUNNEL_INFO_TX) {
2766                         vni = tunnel_id_to_key32(info->key.tun_id);
2767                 } else {
2768                         if (info && info->mode & IP_TUNNEL_INFO_TX)
2769                                 vxlan_xmit_one(skb, dev, vni, NULL, false);
2770                         else
2771                                 kfree_skb(skb);
2772                         return NETDEV_TX_OK;
2773                 }
2774         }
2775
2776         if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2777                 eth = eth_hdr(skb);
2778                 if (ntohs(eth->h_proto) == ETH_P_ARP)
2779                         return arp_reduce(dev, skb, vni);
2780 #if IS_ENABLED(CONFIG_IPV6)
2781                 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2782                          pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2783                                             sizeof(struct nd_msg)) &&
2784                          ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2785                         struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2786
2787                         if (m->icmph.icmp6_code == 0 &&
2788                             m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2789                                 return neigh_reduce(dev, skb, vni);
2790                 }
2791 #endif
2792         }
2793
2794         eth = eth_hdr(skb);
2795         f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2796         did_rsc = false;
2797
2798         if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2799             (ntohs(eth->h_proto) == ETH_P_IP ||
2800              ntohs(eth->h_proto) == ETH_P_IPV6)) {
2801                 did_rsc = route_shortcircuit(dev, skb);
2802                 if (did_rsc)
2803                         f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2804         }
2805
2806         if (f == NULL) {
2807                 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2808                 if (f == NULL) {
2809                         if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2810                             !is_multicast_ether_addr(eth->h_dest))
2811                                 vxlan_fdb_miss(vxlan, eth->h_dest);
2812
2813                         dev->stats.tx_dropped++;
2814                         vxlan_vnifilter_count(vxlan, vni, NULL,
2815                                               VXLAN_VNI_STATS_TX_DROPS, 0);
2816                         kfree_skb(skb);
2817                         return NETDEV_TX_OK;
2818                 }
2819         }
2820
2821         if (rcu_access_pointer(f->nh)) {
2822                 vxlan_xmit_nh(skb, dev, f,
2823                               (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2824         } else {
2825                 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2826                         struct sk_buff *skb1;
2827
2828                         if (!fdst) {
2829                                 fdst = rdst;
2830                                 continue;
2831                         }
2832                         skb1 = skb_clone(skb, GFP_ATOMIC);
2833                         if (skb1)
2834                                 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2835                 }
2836                 if (fdst)
2837                         vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2838                 else
2839                         kfree_skb(skb);
2840         }
2841
2842         return NETDEV_TX_OK;
2843 }
2844
2845 /* Walk the forwarding table and purge stale entries */
2846 static void vxlan_cleanup(struct timer_list *t)
2847 {
2848         struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2849         unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2850         unsigned int h;
2851
2852         if (!netif_running(vxlan->dev))
2853                 return;
2854
2855         for (h = 0; h < FDB_HASH_SIZE; ++h) {
2856                 struct hlist_node *p, *n;
2857
2858                 spin_lock(&vxlan->hash_lock[h]);
2859                 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2860                         struct vxlan_fdb *f
2861                                 = container_of(p, struct vxlan_fdb, hlist);
2862                         unsigned long timeout;
2863
2864                         if (f->state & (NUD_PERMANENT | NUD_NOARP))
2865                                 continue;
2866
2867                         if (f->flags & NTF_EXT_LEARNED)
2868                                 continue;
2869
2870                         timeout = f->used + vxlan->cfg.age_interval * HZ;
2871                         if (time_before_eq(timeout, jiffies)) {
2872                                 netdev_dbg(vxlan->dev,
2873                                            "garbage collect %pM\n",
2874                                            f->eth_addr);
2875                                 f->state = NUD_STALE;
2876                                 vxlan_fdb_destroy(vxlan, f, true, true);
2877                         } else if (time_before(timeout, next_timer))
2878                                 next_timer = timeout;
2879                 }
2880                 spin_unlock(&vxlan->hash_lock[h]);
2881         }
2882
2883         mod_timer(&vxlan->age_timer, next_timer);
2884 }
2885
2886 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2887 {
2888         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2889
2890         spin_lock(&vn->sock_lock);
2891         hlist_del_init_rcu(&vxlan->hlist4.hlist);
2892 #if IS_ENABLED(CONFIG_IPV6)
2893         hlist_del_init_rcu(&vxlan->hlist6.hlist);
2894 #endif
2895         spin_unlock(&vn->sock_lock);
2896 }
2897
2898 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2899                              struct vxlan_dev_node *node)
2900 {
2901         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2902         __be32 vni = vxlan->default_dst.remote_vni;
2903
2904         node->vxlan = vxlan;
2905         spin_lock(&vn->sock_lock);
2906         hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2907         spin_unlock(&vn->sock_lock);
2908 }
2909
2910 /* Setup stats when device is created */
2911 static int vxlan_init(struct net_device *dev)
2912 {
2913         struct vxlan_dev *vxlan = netdev_priv(dev);
2914         int err;
2915
2916         if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2917                 vxlan_vnigroup_init(vxlan);
2918
2919         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2920         if (!dev->tstats) {
2921                 err = -ENOMEM;
2922                 goto err_vnigroup_uninit;
2923         }
2924
2925         err = gro_cells_init(&vxlan->gro_cells, dev);
2926         if (err)
2927                 goto err_free_percpu;
2928
2929         return 0;
2930
2931 err_free_percpu:
2932         free_percpu(dev->tstats);
2933 err_vnigroup_uninit:
2934         if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2935                 vxlan_vnigroup_uninit(vxlan);
2936         return err;
2937 }
2938
2939 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
2940 {
2941         struct vxlan_fdb *f;
2942         u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
2943
2944         spin_lock_bh(&vxlan->hash_lock[hash_index]);
2945         f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
2946         if (f)
2947                 vxlan_fdb_destroy(vxlan, f, true, true);
2948         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
2949 }
2950
2951 static void vxlan_uninit(struct net_device *dev)
2952 {
2953         struct vxlan_dev *vxlan = netdev_priv(dev);
2954
2955         if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2956                 vxlan_vnigroup_uninit(vxlan);
2957
2958         gro_cells_destroy(&vxlan->gro_cells);
2959
2960         vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
2961
2962         free_percpu(dev->tstats);
2963 }
2964
2965 /* Start ageing timer and join group when device is brought up */
2966 static int vxlan_open(struct net_device *dev)
2967 {
2968         struct vxlan_dev *vxlan = netdev_priv(dev);
2969         int ret;
2970
2971         ret = vxlan_sock_add(vxlan);
2972         if (ret < 0)
2973                 return ret;
2974
2975         ret = vxlan_multicast_join(vxlan);
2976         if (ret) {
2977                 vxlan_sock_release(vxlan);
2978                 return ret;
2979         }
2980
2981         if (vxlan->cfg.age_interval)
2982                 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2983
2984         return ret;
2985 }
2986
2987 /* Purge the forwarding table */
2988 static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
2989 {
2990         unsigned int h;
2991
2992         for (h = 0; h < FDB_HASH_SIZE; ++h) {
2993                 struct hlist_node *p, *n;
2994
2995                 spin_lock_bh(&vxlan->hash_lock[h]);
2996                 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2997                         struct vxlan_fdb *f
2998                                 = container_of(p, struct vxlan_fdb, hlist);
2999                         if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
3000                                 continue;
3001                         /* the all_zeros_mac entry is deleted at vxlan_uninit */
3002                         if (is_zero_ether_addr(f->eth_addr) &&
3003                             f->vni == vxlan->cfg.vni)
3004                                 continue;
3005                         vxlan_fdb_destroy(vxlan, f, true, true);
3006                 }
3007                 spin_unlock_bh(&vxlan->hash_lock[h]);
3008         }
3009 }
3010
3011 /* Cleanup timer and forwarding table on shutdown */
3012 static int vxlan_stop(struct net_device *dev)
3013 {
3014         struct vxlan_dev *vxlan = netdev_priv(dev);
3015
3016         vxlan_multicast_leave(vxlan);
3017
3018         del_timer_sync(&vxlan->age_timer);
3019
3020         vxlan_flush(vxlan, false);
3021         vxlan_sock_release(vxlan);
3022
3023         return 0;
3024 }
3025
3026 /* Stub, nothing needs to be done. */
3027 static void vxlan_set_multicast_list(struct net_device *dev)
3028 {
3029 }
3030
3031 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3032 {
3033         struct vxlan_dev *vxlan = netdev_priv(dev);
3034         struct vxlan_rdst *dst = &vxlan->default_dst;
3035         struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3036                                                          dst->remote_ifindex);
3037         bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
3038
3039         /* This check is different than dev->max_mtu, because it looks at
3040          * the lowerdev->mtu, rather than the static dev->max_mtu
3041          */
3042         if (lowerdev) {
3043                 int max_mtu = lowerdev->mtu -
3044                               (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
3045                 if (new_mtu > max_mtu)
3046                         return -EINVAL;
3047         }
3048
3049         dev->mtu = new_mtu;
3050         return 0;
3051 }
3052
3053 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3054 {
3055         struct vxlan_dev *vxlan = netdev_priv(dev);
3056         struct ip_tunnel_info *info = skb_tunnel_info(skb);
3057         __be16 sport, dport;
3058
3059         sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3060                                   vxlan->cfg.port_max, true);
3061         dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3062
3063         if (ip_tunnel_info_af(info) == AF_INET) {
3064                 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3065                 struct rtable *rt;
3066
3067                 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
3068                                      info->key.u.ipv4.dst,
3069                                      &info->key.u.ipv4.src, dport, sport,
3070                                      info->key.flow_flags, &info->dst_cache,
3071                                      info);
3072                 if (IS_ERR(rt))
3073                         return PTR_ERR(rt);
3074                 ip_rt_put(rt);
3075         } else {
3076 #if IS_ENABLED(CONFIG_IPV6)
3077                 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3078                 struct dst_entry *ndst;
3079
3080                 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
3081                                         info->key.label, &info->key.u.ipv6.dst,
3082                                         &info->key.u.ipv6.src, dport, sport,
3083                                         &info->dst_cache, info);
3084                 if (IS_ERR(ndst))
3085                         return PTR_ERR(ndst);
3086                 dst_release(ndst);
3087 #else /* !CONFIG_IPV6 */
3088                 return -EPFNOSUPPORT;
3089 #endif
3090         }
3091         info->key.tp_src = sport;
3092         info->key.tp_dst = dport;
3093         return 0;
3094 }
3095
3096 static const struct net_device_ops vxlan_netdev_ether_ops = {
3097         .ndo_init               = vxlan_init,
3098         .ndo_uninit             = vxlan_uninit,
3099         .ndo_open               = vxlan_open,
3100         .ndo_stop               = vxlan_stop,
3101         .ndo_start_xmit         = vxlan_xmit,
3102         .ndo_get_stats64        = dev_get_tstats64,
3103         .ndo_set_rx_mode        = vxlan_set_multicast_list,
3104         .ndo_change_mtu         = vxlan_change_mtu,
3105         .ndo_validate_addr      = eth_validate_addr,
3106         .ndo_set_mac_address    = eth_mac_addr,
3107         .ndo_fdb_add            = vxlan_fdb_add,
3108         .ndo_fdb_del            = vxlan_fdb_delete,
3109         .ndo_fdb_dump           = vxlan_fdb_dump,
3110         .ndo_fdb_get            = vxlan_fdb_get,
3111         .ndo_fill_metadata_dst  = vxlan_fill_metadata_dst,
3112 };
3113
3114 static const struct net_device_ops vxlan_netdev_raw_ops = {
3115         .ndo_init               = vxlan_init,
3116         .ndo_uninit             = vxlan_uninit,
3117         .ndo_open               = vxlan_open,
3118         .ndo_stop               = vxlan_stop,
3119         .ndo_start_xmit         = vxlan_xmit,
3120         .ndo_get_stats64        = dev_get_tstats64,
3121         .ndo_change_mtu         = vxlan_change_mtu,
3122         .ndo_fill_metadata_dst  = vxlan_fill_metadata_dst,
3123 };
3124
3125 /* Info for udev, that this is a virtual tunnel endpoint */
3126 static struct device_type vxlan_type = {
3127         .name = "vxlan",
3128 };
3129
3130 /* Calls the ndo_udp_tunnel_add of the caller in order to
3131  * supply the listening VXLAN udp ports. Callers are expected
3132  * to implement the ndo_udp_tunnel_add.
3133  */
3134 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3135 {
3136         struct vxlan_sock *vs;
3137         struct net *net = dev_net(dev);
3138         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3139         unsigned int i;
3140
3141         spin_lock(&vn->sock_lock);
3142         for (i = 0; i < PORT_HASH_SIZE; ++i) {
3143                 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3144                         unsigned short type;
3145
3146                         if (vs->flags & VXLAN_F_GPE)
3147                                 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3148                         else
3149                                 type = UDP_TUNNEL_TYPE_VXLAN;
3150
3151                         if (push)
3152                                 udp_tunnel_push_rx_port(dev, vs->sock, type);
3153                         else
3154                                 udp_tunnel_drop_rx_port(dev, vs->sock, type);
3155                 }
3156         }
3157         spin_unlock(&vn->sock_lock);
3158 }
3159
3160 /* Initialize the device structure. */
3161 static void vxlan_setup(struct net_device *dev)
3162 {
3163         struct vxlan_dev *vxlan = netdev_priv(dev);
3164         unsigned int h;
3165
3166         eth_hw_addr_random(dev);
3167         ether_setup(dev);
3168
3169         dev->needs_free_netdev = true;
3170         SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3171
3172         dev->features   |= NETIF_F_LLTX;
3173         dev->features   |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3174         dev->features   |= NETIF_F_RXCSUM;
3175         dev->features   |= NETIF_F_GSO_SOFTWARE;
3176
3177         dev->vlan_features = dev->features;
3178         dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3179         dev->hw_features |= NETIF_F_RXCSUM;
3180         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3181         netif_keep_dst(dev);
3182         dev->priv_flags |= IFF_NO_QUEUE | IFF_CHANGE_PROTO_DOWN;
3183
3184         /* MTU range: 68 - 65535 */
3185         dev->min_mtu = ETH_MIN_MTU;
3186         dev->max_mtu = ETH_MAX_MTU;
3187
3188         INIT_LIST_HEAD(&vxlan->next);
3189
3190         timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3191
3192         vxlan->dev = dev;
3193
3194         for (h = 0; h < FDB_HASH_SIZE; ++h) {
3195                 spin_lock_init(&vxlan->hash_lock[h]);
3196                 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3197         }
3198 }
3199
3200 static void vxlan_ether_setup(struct net_device *dev)
3201 {
3202         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3203         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3204         dev->netdev_ops = &vxlan_netdev_ether_ops;
3205 }
3206
3207 static void vxlan_raw_setup(struct net_device *dev)
3208 {
3209         dev->header_ops = NULL;
3210         dev->type = ARPHRD_NONE;
3211         dev->hard_header_len = 0;
3212         dev->addr_len = 0;
3213         dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3214         dev->netdev_ops = &vxlan_netdev_raw_ops;
3215 }
3216
3217 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3218         [IFLA_VXLAN_ID]         = { .type = NLA_U32 },
3219         [IFLA_VXLAN_GROUP]      = { .len = sizeof_field(struct iphdr, daddr) },
3220         [IFLA_VXLAN_GROUP6]     = { .len = sizeof(struct in6_addr) },
3221         [IFLA_VXLAN_LINK]       = { .type = NLA_U32 },
3222         [IFLA_VXLAN_LOCAL]      = { .len = sizeof_field(struct iphdr, saddr) },
3223         [IFLA_VXLAN_LOCAL6]     = { .len = sizeof(struct in6_addr) },
3224         [IFLA_VXLAN_TOS]        = { .type = NLA_U8 },
3225         [IFLA_VXLAN_TTL]        = { .type = NLA_U8 },
3226         [IFLA_VXLAN_LABEL]      = { .type = NLA_U32 },
3227         [IFLA_VXLAN_LEARNING]   = { .type = NLA_U8 },
3228         [IFLA_VXLAN_AGEING]     = { .type = NLA_U32 },
3229         [IFLA_VXLAN_LIMIT]      = { .type = NLA_U32 },
3230         [IFLA_VXLAN_PORT_RANGE] = { .len  = sizeof(struct ifla_vxlan_port_range) },
3231         [IFLA_VXLAN_PROXY]      = { .type = NLA_U8 },
3232         [IFLA_VXLAN_RSC]        = { .type = NLA_U8 },
3233         [IFLA_VXLAN_L2MISS]     = { .type = NLA_U8 },
3234         [IFLA_VXLAN_L3MISS]     = { .type = NLA_U8 },
3235         [IFLA_VXLAN_COLLECT_METADATA]   = { .type = NLA_U8 },
3236         [IFLA_VXLAN_PORT]       = { .type = NLA_U16 },
3237         [IFLA_VXLAN_UDP_CSUM]   = { .type = NLA_U8 },
3238         [IFLA_VXLAN_UDP_ZERO_CSUM6_TX]  = { .type = NLA_U8 },
3239         [IFLA_VXLAN_UDP_ZERO_CSUM6_RX]  = { .type = NLA_U8 },
3240         [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3241         [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3242         [IFLA_VXLAN_GBP]        = { .type = NLA_FLAG, },
3243         [IFLA_VXLAN_GPE]        = { .type = NLA_FLAG, },
3244         [IFLA_VXLAN_REMCSUM_NOPARTIAL]  = { .type = NLA_FLAG },
3245         [IFLA_VXLAN_TTL_INHERIT]        = { .type = NLA_FLAG },
3246         [IFLA_VXLAN_DF]         = { .type = NLA_U8 },
3247         [IFLA_VXLAN_VNIFILTER]  = { .type = NLA_U8 },
3248 };
3249
3250 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3251                           struct netlink_ext_ack *extack)
3252 {
3253         if (tb[IFLA_ADDRESS]) {
3254                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3255                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3256                                             "Provided link layer address is not Ethernet");
3257                         return -EINVAL;
3258                 }
3259
3260                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3261                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3262                                             "Provided Ethernet address is not unicast");
3263                         return -EADDRNOTAVAIL;
3264                 }
3265         }
3266
3267         if (tb[IFLA_MTU]) {
3268                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3269
3270                 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3271                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3272                                             "MTU must be between 68 and 65535");
3273                         return -EINVAL;
3274                 }
3275         }
3276
3277         if (!data) {
3278                 NL_SET_ERR_MSG(extack,
3279                                "Required attributes not provided to perform the operation");
3280                 return -EINVAL;
3281         }
3282
3283         if (data[IFLA_VXLAN_ID]) {
3284                 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3285
3286                 if (id >= VXLAN_N_VID) {
3287                         NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3288                                             "VXLAN ID must be lower than 16777216");
3289                         return -ERANGE;
3290                 }
3291         }
3292
3293         if (data[IFLA_VXLAN_PORT_RANGE]) {
3294                 const struct ifla_vxlan_port_range *p
3295                         = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3296
3297                 if (ntohs(p->high) < ntohs(p->low)) {
3298                         NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3299                                             "Invalid source port range");
3300                         return -EINVAL;
3301                 }
3302         }
3303
3304         if (data[IFLA_VXLAN_DF]) {
3305                 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3306
3307                 if (df < 0 || df > VXLAN_DF_MAX) {
3308                         NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3309                                             "Invalid DF attribute");
3310                         return -EINVAL;
3311                 }
3312         }
3313
3314         return 0;
3315 }
3316
3317 static void vxlan_get_drvinfo(struct net_device *netdev,
3318                               struct ethtool_drvinfo *drvinfo)
3319 {
3320         strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3321         strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3322 }
3323
3324 static int vxlan_get_link_ksettings(struct net_device *dev,
3325                                     struct ethtool_link_ksettings *cmd)
3326 {
3327         struct vxlan_dev *vxlan = netdev_priv(dev);
3328         struct vxlan_rdst *dst = &vxlan->default_dst;
3329         struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3330                                                          dst->remote_ifindex);
3331
3332         if (!lowerdev) {
3333                 cmd->base.duplex = DUPLEX_UNKNOWN;
3334                 cmd->base.port = PORT_OTHER;
3335                 cmd->base.speed = SPEED_UNKNOWN;
3336
3337                 return 0;
3338         }
3339
3340         return __ethtool_get_link_ksettings(lowerdev, cmd);
3341 }
3342
3343 static const struct ethtool_ops vxlan_ethtool_ops = {
3344         .get_drvinfo            = vxlan_get_drvinfo,
3345         .get_link               = ethtool_op_get_link,
3346         .get_link_ksettings     = vxlan_get_link_ksettings,
3347 };
3348
3349 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3350                                         __be16 port, u32 flags, int ifindex)
3351 {
3352         struct socket *sock;
3353         struct udp_port_cfg udp_conf;
3354         int err;
3355
3356         memset(&udp_conf, 0, sizeof(udp_conf));
3357
3358         if (ipv6) {
3359                 udp_conf.family = AF_INET6;
3360                 udp_conf.use_udp6_rx_checksums =
3361                     !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3362                 udp_conf.ipv6_v6only = 1;
3363         } else {
3364                 udp_conf.family = AF_INET;
3365         }
3366
3367         udp_conf.local_udp_port = port;
3368         udp_conf.bind_ifindex = ifindex;
3369
3370         /* Open UDP socket */
3371         err = udp_sock_create(net, &udp_conf, &sock);
3372         if (err < 0)
3373                 return ERR_PTR(err);
3374
3375         udp_allow_gso(sock->sk);
3376         return sock;
3377 }
3378
3379 /* Create new listen socket if needed */
3380 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3381                                               __be16 port, u32 flags,
3382                                               int ifindex)
3383 {
3384         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3385         struct vxlan_sock *vs;
3386         struct socket *sock;
3387         unsigned int h;
3388         struct udp_tunnel_sock_cfg tunnel_cfg;
3389
3390         vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3391         if (!vs)
3392                 return ERR_PTR(-ENOMEM);
3393
3394         for (h = 0; h < VNI_HASH_SIZE; ++h)
3395                 INIT_HLIST_HEAD(&vs->vni_list[h]);
3396
3397         sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3398         if (IS_ERR(sock)) {
3399                 kfree(vs);
3400                 return ERR_CAST(sock);
3401         }
3402
3403         vs->sock = sock;
3404         refcount_set(&vs->refcnt, 1);
3405         vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3406
3407         spin_lock(&vn->sock_lock);
3408         hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3409         udp_tunnel_notify_add_rx_port(sock,
3410                                       (vs->flags & VXLAN_F_GPE) ?
3411                                       UDP_TUNNEL_TYPE_VXLAN_GPE :
3412                                       UDP_TUNNEL_TYPE_VXLAN);
3413         spin_unlock(&vn->sock_lock);
3414
3415         /* Mark socket as an encapsulation socket. */
3416         memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3417         tunnel_cfg.sk_user_data = vs;
3418         tunnel_cfg.encap_type = 1;
3419         tunnel_cfg.encap_rcv = vxlan_rcv;
3420         tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3421         tunnel_cfg.encap_destroy = NULL;
3422         tunnel_cfg.gro_receive = vxlan_gro_receive;
3423         tunnel_cfg.gro_complete = vxlan_gro_complete;
3424
3425         setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3426
3427         return vs;
3428 }
3429
3430 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3431 {
3432         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3433         bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3434         struct vxlan_sock *vs = NULL;
3435         struct vxlan_dev_node *node;
3436         int l3mdev_index = 0;
3437
3438         if (vxlan->cfg.remote_ifindex)
3439                 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3440                         vxlan->net, vxlan->cfg.remote_ifindex);
3441
3442         if (!vxlan->cfg.no_share) {
3443                 spin_lock(&vn->sock_lock);
3444                 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3445                                      vxlan->cfg.dst_port, vxlan->cfg.flags,
3446                                      l3mdev_index);
3447                 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3448                         spin_unlock(&vn->sock_lock);
3449                         return -EBUSY;
3450                 }
3451                 spin_unlock(&vn->sock_lock);
3452         }
3453         if (!vs)
3454                 vs = vxlan_socket_create(vxlan->net, ipv6,
3455                                          vxlan->cfg.dst_port, vxlan->cfg.flags,
3456                                          l3mdev_index);
3457         if (IS_ERR(vs))
3458                 return PTR_ERR(vs);
3459 #if IS_ENABLED(CONFIG_IPV6)
3460         if (ipv6) {
3461                 rcu_assign_pointer(vxlan->vn6_sock, vs);
3462                 node = &vxlan->hlist6;
3463         } else
3464 #endif
3465         {
3466                 rcu_assign_pointer(vxlan->vn4_sock, vs);
3467                 node = &vxlan->hlist4;
3468         }
3469
3470         if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER))
3471                 vxlan_vs_add_vnigrp(vxlan, vs, ipv6);
3472         else
3473                 vxlan_vs_add_dev(vs, vxlan, node);
3474
3475         return 0;
3476 }
3477
3478 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3479 {
3480         bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3481         bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3482         bool ipv4 = !ipv6 || metadata;
3483         int ret = 0;
3484
3485         RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3486 #if IS_ENABLED(CONFIG_IPV6)
3487         RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3488         if (ipv6) {
3489                 ret = __vxlan_sock_add(vxlan, true);
3490                 if (ret < 0 && ret != -EAFNOSUPPORT)
3491                         ipv4 = false;
3492         }
3493 #endif
3494         if (ipv4)
3495                 ret = __vxlan_sock_add(vxlan, false);
3496         if (ret < 0)
3497                 vxlan_sock_release(vxlan);
3498         return ret;
3499 }
3500
3501 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan,
3502                      struct vxlan_config *conf, __be32 vni)
3503 {
3504         struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3505         struct vxlan_dev *tmp;
3506
3507         list_for_each_entry(tmp, &vn->vxlan_list, next) {
3508                 if (tmp == vxlan)
3509                         continue;
3510                 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) {
3511                         if (!vxlan_vnifilter_lookup(tmp, vni))
3512                                 continue;
3513                 } else if (tmp->cfg.vni != vni) {
3514                         continue;
3515                 }
3516                 if (tmp->cfg.dst_port != conf->dst_port)
3517                         continue;
3518                 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3519                     (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3520                         continue;
3521
3522                 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3523                     tmp->cfg.remote_ifindex != conf->remote_ifindex)
3524                         continue;
3525
3526                 return -EEXIST;
3527         }
3528
3529         return 0;
3530 }
3531
3532 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3533                                  struct net_device **lower,
3534                                  struct vxlan_dev *old,
3535                                  struct netlink_ext_ack *extack)
3536 {
3537         bool use_ipv6 = false;
3538
3539         if (conf->flags & VXLAN_F_GPE) {
3540                 /* For now, allow GPE only together with
3541                  * COLLECT_METADATA. This can be relaxed later; in such
3542                  * case, the other side of the PtP link will have to be
3543                  * provided.
3544                  */
3545                 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3546                     !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3547                         NL_SET_ERR_MSG(extack,
3548                                        "VXLAN GPE does not support this combination of attributes");
3549                         return -EINVAL;
3550                 }
3551         }
3552
3553         if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3554                 /* Unless IPv6 is explicitly requested, assume IPv4 */
3555                 conf->remote_ip.sa.sa_family = AF_INET;
3556                 conf->saddr.sa.sa_family = AF_INET;
3557         } else if (!conf->remote_ip.sa.sa_family) {
3558                 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3559         } else if (!conf->saddr.sa.sa_family) {
3560                 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3561         }
3562
3563         if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3564                 NL_SET_ERR_MSG(extack,
3565                                "Local and remote address must be from the same family");
3566                 return -EINVAL;
3567         }
3568
3569         if (vxlan_addr_multicast(&conf->saddr)) {
3570                 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3571                 return -EINVAL;
3572         }
3573
3574         if (conf->saddr.sa.sa_family == AF_INET6) {
3575                 if (!IS_ENABLED(CONFIG_IPV6)) {
3576                         NL_SET_ERR_MSG(extack,
3577                                        "IPv6 support not enabled in the kernel");
3578                         return -EPFNOSUPPORT;
3579                 }
3580                 use_ipv6 = true;
3581                 conf->flags |= VXLAN_F_IPV6;
3582
3583                 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3584                         int local_type =
3585                                 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3586                         int remote_type =
3587                                 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3588
3589                         if (local_type & IPV6_ADDR_LINKLOCAL) {
3590                                 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3591                                     (remote_type != IPV6_ADDR_ANY)) {
3592                                         NL_SET_ERR_MSG(extack,
3593                                                        "Invalid combination of local and remote address scopes");
3594                                         return -EINVAL;
3595                                 }
3596
3597                                 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3598                         } else {
3599                                 if (remote_type ==
3600                                     (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3601                                         NL_SET_ERR_MSG(extack,
3602                                                        "Invalid combination of local and remote address scopes");
3603                                         return -EINVAL;
3604                                 }
3605
3606                                 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3607                         }
3608                 }
3609         }
3610
3611         if (conf->label && !use_ipv6) {
3612                 NL_SET_ERR_MSG(extack,
3613                                "Label attribute only applies to IPv6 VXLAN devices");
3614                 return -EINVAL;
3615         }
3616
3617         if (conf->remote_ifindex) {
3618                 struct net_device *lowerdev;
3619
3620                 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3621                 if (!lowerdev) {
3622                         NL_SET_ERR_MSG(extack,
3623                                        "Invalid local interface, device not found");
3624                         return -ENODEV;
3625                 }
3626
3627 #if IS_ENABLED(CONFIG_IPV6)
3628                 if (use_ipv6) {
3629                         struct inet6_dev *idev = __in6_dev_get(lowerdev);
3630
3631                         if (idev && idev->cnf.disable_ipv6) {
3632                                 NL_SET_ERR_MSG(extack,
3633                                                "IPv6 support disabled by administrator");
3634                                 return -EPERM;
3635                         }
3636                 }
3637 #endif
3638
3639                 *lower = lowerdev;
3640         } else {
3641                 if (vxlan_addr_multicast(&conf->remote_ip)) {
3642                         NL_SET_ERR_MSG(extack,
3643                                        "Local interface required for multicast remote destination");
3644
3645                         return -EINVAL;
3646                 }
3647
3648 #if IS_ENABLED(CONFIG_IPV6)
3649                 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3650                         NL_SET_ERR_MSG(extack,
3651                                        "Local interface required for link-local local/remote addresses");
3652                         return -EINVAL;
3653                 }
3654 #endif
3655
3656                 *lower = NULL;
3657         }
3658
3659         if (!conf->dst_port) {
3660                 if (conf->flags & VXLAN_F_GPE)
3661                         conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT);
3662                 else
3663                         conf->dst_port = htons(vxlan_port);
3664         }
3665
3666         if (!conf->age_interval)
3667                 conf->age_interval = FDB_AGE_DEFAULT;
3668
3669         if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) {
3670                 NL_SET_ERR_MSG(extack,
3671                                "A VXLAN device with the specified VNI already exists");
3672                 return -EEXIST;
3673         }
3674
3675         return 0;
3676 }
3677
3678 static void vxlan_config_apply(struct net_device *dev,
3679                                struct vxlan_config *conf,
3680                                struct net_device *lowerdev,
3681                                struct net *src_net,
3682                                bool changelink)
3683 {
3684         struct vxlan_dev *vxlan = netdev_priv(dev);
3685         struct vxlan_rdst *dst = &vxlan->default_dst;
3686         unsigned short needed_headroom = ETH_HLEN;
3687         bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3688         int max_mtu = ETH_MAX_MTU;
3689
3690         if (!changelink) {
3691                 if (conf->flags & VXLAN_F_GPE)
3692                         vxlan_raw_setup(dev);
3693                 else
3694                         vxlan_ether_setup(dev);
3695
3696                 if (conf->mtu)
3697                         dev->mtu = conf->mtu;
3698
3699                 vxlan->net = src_net;
3700         }
3701
3702         dst->remote_vni = conf->vni;
3703
3704         memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3705
3706         if (lowerdev) {
3707                 dst->remote_ifindex = conf->remote_ifindex;
3708
3709                 netif_inherit_tso_max(dev, lowerdev);
3710
3711                 needed_headroom = lowerdev->hard_header_len;
3712                 needed_headroom += lowerdev->needed_headroom;
3713
3714                 dev->needed_tailroom = lowerdev->needed_tailroom;
3715
3716                 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3717                                            VXLAN_HEADROOM);
3718                 if (max_mtu < ETH_MIN_MTU)
3719                         max_mtu = ETH_MIN_MTU;
3720
3721                 if (!changelink && !conf->mtu)
3722                         dev->mtu = max_mtu;
3723         }
3724
3725         if (dev->mtu > max_mtu)
3726                 dev->mtu = max_mtu;
3727
3728         if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3729                 needed_headroom += VXLAN6_HEADROOM;
3730         else
3731                 needed_headroom += VXLAN_HEADROOM;
3732         dev->needed_headroom = needed_headroom;
3733
3734         memcpy(&vxlan->cfg, conf, sizeof(*conf));
3735 }
3736
3737 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3738                                struct vxlan_config *conf, bool changelink,
3739                                struct netlink_ext_ack *extack)
3740 {
3741         struct vxlan_dev *vxlan = netdev_priv(dev);
3742         struct net_device *lowerdev;
3743         int ret;
3744
3745         ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3746         if (ret)
3747                 return ret;
3748
3749         vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3750
3751         return 0;
3752 }
3753
3754 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3755                               struct vxlan_config *conf,
3756                               struct netlink_ext_ack *extack)
3757 {
3758         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3759         struct vxlan_dev *vxlan = netdev_priv(dev);
3760         struct net_device *remote_dev = NULL;
3761         struct vxlan_fdb *f = NULL;
3762         bool unregister = false;
3763         struct vxlan_rdst *dst;
3764         int err;
3765
3766         dst = &vxlan->default_dst;
3767         err = vxlan_dev_configure(net, dev, conf, false, extack);
3768         if (err)
3769                 return err;
3770
3771         dev->ethtool_ops = &vxlan_ethtool_ops;
3772
3773         /* create an fdb entry for a valid default destination */
3774         if (!vxlan_addr_any(&dst->remote_ip)) {
3775                 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3776                                        &dst->remote_ip,
3777                                        NUD_REACHABLE | NUD_PERMANENT,
3778                                        vxlan->cfg.dst_port,
3779                                        dst->remote_vni,
3780                                        dst->remote_vni,
3781                                        dst->remote_ifindex,
3782                                        NTF_SELF, 0, &f, extack);
3783                 if (err)
3784                         return err;
3785         }
3786
3787         err = register_netdevice(dev);
3788         if (err)
3789                 goto errout;
3790         unregister = true;
3791
3792         if (dst->remote_ifindex) {
3793                 remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3794                 if (!remote_dev) {
3795                         err = -ENODEV;
3796                         goto errout;
3797                 }
3798
3799                 err = netdev_upper_dev_link(remote_dev, dev, extack);
3800                 if (err)
3801                         goto errout;
3802         }
3803
3804         err = rtnl_configure_link(dev, NULL, 0, NULL);
3805         if (err < 0)
3806                 goto unlink;
3807
3808         if (f) {
3809                 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
3810
3811                 /* notify default fdb entry */
3812                 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
3813                                        RTM_NEWNEIGH, true, extack);
3814                 if (err) {
3815                         vxlan_fdb_destroy(vxlan, f, false, false);
3816                         if (remote_dev)
3817                                 netdev_upper_dev_unlink(remote_dev, dev);
3818                         goto unregister;
3819                 }
3820         }
3821
3822         list_add(&vxlan->next, &vn->vxlan_list);
3823         if (remote_dev)
3824                 dst->remote_dev = remote_dev;
3825         return 0;
3826 unlink:
3827         if (remote_dev)
3828                 netdev_upper_dev_unlink(remote_dev, dev);
3829 errout:
3830         /* unregister_netdevice() destroys the default FDB entry with deletion
3831          * notification. But the addition notification was not sent yet, so
3832          * destroy the entry by hand here.
3833          */
3834         if (f)
3835                 __vxlan_fdb_free(f);
3836 unregister:
3837         if (unregister)
3838                 unregister_netdevice(dev);
3839         return err;
3840 }
3841
3842 /* Set/clear flags based on attribute */
3843 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3844                           int attrtype, unsigned long mask, bool changelink,
3845                           bool changelink_supported,
3846                           struct netlink_ext_ack *extack)
3847 {
3848         unsigned long flags;
3849
3850         if (!tb[attrtype])
3851                 return 0;
3852
3853         if (changelink && !changelink_supported) {
3854                 vxlan_flag_attr_error(attrtype, extack);
3855                 return -EOPNOTSUPP;
3856         }
3857
3858         if (vxlan_policy[attrtype].type == NLA_FLAG)
3859                 flags = conf->flags | mask;
3860         else if (nla_get_u8(tb[attrtype]))
3861                 flags = conf->flags | mask;
3862         else
3863                 flags = conf->flags & ~mask;
3864
3865         conf->flags = flags;
3866
3867         return 0;
3868 }
3869
3870 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3871                          struct net_device *dev, struct vxlan_config *conf,
3872                          bool changelink, struct netlink_ext_ack *extack)
3873 {
3874         struct vxlan_dev *vxlan = netdev_priv(dev);
3875         int err = 0;
3876
3877         memset(conf, 0, sizeof(*conf));
3878
3879         /* if changelink operation, start with old existing cfg */
3880         if (changelink)
3881                 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3882
3883         if (data[IFLA_VXLAN_ID]) {
3884                 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3885
3886                 if (changelink && (vni != conf->vni)) {
3887                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
3888                         return -EOPNOTSUPP;
3889                 }
3890                 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3891         }
3892
3893         if (data[IFLA_VXLAN_GROUP]) {
3894                 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
3895                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
3896                         return -EOPNOTSUPP;
3897                 }
3898
3899                 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
3900                 conf->remote_ip.sa.sa_family = AF_INET;
3901         } else if (data[IFLA_VXLAN_GROUP6]) {
3902                 if (!IS_ENABLED(CONFIG_IPV6)) {
3903                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
3904                         return -EPFNOSUPPORT;
3905                 }
3906
3907                 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
3908                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
3909                         return -EOPNOTSUPP;
3910                 }
3911
3912                 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3913                 conf->remote_ip.sa.sa_family = AF_INET6;
3914         }
3915
3916         if (data[IFLA_VXLAN_LOCAL]) {
3917                 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
3918                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
3919                         return -EOPNOTSUPP;
3920                 }
3921
3922                 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3923                 conf->saddr.sa.sa_family = AF_INET;
3924         } else if (data[IFLA_VXLAN_LOCAL6]) {
3925                 if (!IS_ENABLED(CONFIG_IPV6)) {
3926                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
3927                         return -EPFNOSUPPORT;
3928                 }
3929
3930                 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
3931                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
3932                         return -EOPNOTSUPP;
3933                 }
3934
3935                 /* TODO: respect scope id */
3936                 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3937                 conf->saddr.sa.sa_family = AF_INET6;
3938         }
3939
3940         if (data[IFLA_VXLAN_LINK])
3941                 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
3942
3943         if (data[IFLA_VXLAN_TOS])
3944                 conf->tos  = nla_get_u8(data[IFLA_VXLAN_TOS]);
3945
3946         if (data[IFLA_VXLAN_TTL])
3947                 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
3948
3949         if (data[IFLA_VXLAN_TTL_INHERIT]) {
3950                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
3951                                     VXLAN_F_TTL_INHERIT, changelink, false,
3952                                     extack);
3953                 if (err)
3954                         return err;
3955
3956         }
3957
3958         if (data[IFLA_VXLAN_LABEL])
3959                 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
3960                              IPV6_FLOWLABEL_MASK;
3961
3962         if (data[IFLA_VXLAN_LEARNING]) {
3963                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
3964                                     VXLAN_F_LEARN, changelink, true,
3965                                     extack);
3966                 if (err)
3967                         return err;
3968         } else if (!changelink) {
3969                 /* default to learn on a new device */
3970                 conf->flags |= VXLAN_F_LEARN;
3971         }
3972
3973         if (data[IFLA_VXLAN_AGEING])
3974                 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3975
3976         if (data[IFLA_VXLAN_PROXY]) {
3977                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
3978                                     VXLAN_F_PROXY, changelink, false,
3979                                     extack);
3980                 if (err)
3981                         return err;
3982         }
3983
3984         if (data[IFLA_VXLAN_RSC]) {
3985                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
3986                                     VXLAN_F_RSC, changelink, false,
3987                                     extack);
3988                 if (err)
3989                         return err;
3990         }
3991
3992         if (data[IFLA_VXLAN_L2MISS]) {
3993                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
3994                                     VXLAN_F_L2MISS, changelink, false,
3995                                     extack);
3996                 if (err)
3997                         return err;
3998         }
3999
4000         if (data[IFLA_VXLAN_L3MISS]) {
4001                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4002                                     VXLAN_F_L3MISS, changelink, false,
4003                                     extack);
4004                 if (err)
4005                         return err;
4006         }
4007
4008         if (data[IFLA_VXLAN_LIMIT]) {
4009                 if (changelink) {
4010                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4011                                             "Cannot change limit");
4012                         return -EOPNOTSUPP;
4013                 }
4014                 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4015         }
4016
4017         if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4018                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4019                                     VXLAN_F_COLLECT_METADATA, changelink, false,
4020                                     extack);
4021                 if (err)
4022                         return err;
4023         }
4024
4025         if (data[IFLA_VXLAN_PORT_RANGE]) {
4026                 if (!changelink) {
4027                         const struct ifla_vxlan_port_range *p
4028                                 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4029                         conf->port_min = ntohs(p->low);
4030                         conf->port_max = ntohs(p->high);
4031                 } else {
4032                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4033                                             "Cannot change port range");
4034                         return -EOPNOTSUPP;
4035                 }
4036         }
4037
4038         if (data[IFLA_VXLAN_PORT]) {
4039                 if (changelink) {
4040                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4041                                             "Cannot change port");
4042                         return -EOPNOTSUPP;
4043                 }
4044                 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4045         }
4046
4047         if (data[IFLA_VXLAN_UDP_CSUM]) {
4048                 if (changelink) {
4049                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4050                                             "Cannot change UDP_CSUM flag");
4051                         return -EOPNOTSUPP;
4052                 }
4053                 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4054                         conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4055         }
4056
4057         if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4058                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4059                                     VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4060                                     false, extack);
4061                 if (err)
4062                         return err;
4063         }
4064
4065         if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4066                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4067                                     VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4068                                     false, extack);
4069                 if (err)
4070                         return err;
4071         }
4072
4073         if (data[IFLA_VXLAN_REMCSUM_TX]) {
4074                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4075                                     VXLAN_F_REMCSUM_TX, changelink, false,
4076                                     extack);
4077                 if (err)
4078                         return err;
4079         }
4080
4081         if (data[IFLA_VXLAN_REMCSUM_RX]) {
4082                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4083                                     VXLAN_F_REMCSUM_RX, changelink, false,
4084                                     extack);
4085                 if (err)
4086                         return err;
4087         }
4088
4089         if (data[IFLA_VXLAN_GBP]) {
4090                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4091                                     VXLAN_F_GBP, changelink, false, extack);
4092                 if (err)
4093                         return err;
4094         }
4095
4096         if (data[IFLA_VXLAN_GPE]) {
4097                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4098                                     VXLAN_F_GPE, changelink, false,
4099                                     extack);
4100                 if (err)
4101                         return err;
4102         }
4103
4104         if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4105                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4106                                     VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4107                                     false, extack);
4108                 if (err)
4109                         return err;
4110         }
4111
4112         if (tb[IFLA_MTU]) {
4113                 if (changelink) {
4114                         NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4115                                             "Cannot change mtu");
4116                         return -EOPNOTSUPP;
4117                 }
4118                 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4119         }
4120
4121         if (data[IFLA_VXLAN_DF])
4122                 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4123
4124         if (data[IFLA_VXLAN_VNIFILTER]) {
4125                 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER,
4126                                     VXLAN_F_VNIFILTER, changelink, false,
4127                                     extack);
4128                 if (err)
4129                         return err;
4130
4131                 if ((conf->flags & VXLAN_F_VNIFILTER) &&
4132                     !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
4133                         NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER],
4134                                             "vxlan vnifilter only valid in collect metadata mode");
4135                         return -EINVAL;
4136                 }
4137         }
4138
4139         return 0;
4140 }
4141
4142 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
4143                          struct nlattr *tb[], struct nlattr *data[],
4144                          struct netlink_ext_ack *extack)
4145 {
4146         struct vxlan_config conf;
4147         int err;
4148
4149         err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4150         if (err)
4151                 return err;
4152
4153         return __vxlan_dev_create(src_net, dev, &conf, extack);
4154 }
4155
4156 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4157                             struct nlattr *data[],
4158                             struct netlink_ext_ack *extack)
4159 {
4160         struct vxlan_dev *vxlan = netdev_priv(dev);
4161         struct net_device *lowerdev;
4162         struct vxlan_config conf;
4163         struct vxlan_rdst *dst;
4164         int err;
4165
4166         dst = &vxlan->default_dst;
4167         err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4168         if (err)
4169                 return err;
4170
4171         err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4172                                     vxlan, extack);
4173         if (err)
4174                 return err;
4175
4176         if (dst->remote_dev == lowerdev)
4177                 lowerdev = NULL;
4178
4179         err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4180                                              extack);
4181         if (err)
4182                 return err;
4183
4184         /* handle default dst entry */
4185         if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
4186                 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4187
4188                 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4189                 if (!vxlan_addr_any(&conf.remote_ip)) {
4190                         err = vxlan_fdb_update(vxlan, all_zeros_mac,
4191                                                &conf.remote_ip,
4192                                                NUD_REACHABLE | NUD_PERMANENT,
4193                                                NLM_F_APPEND | NLM_F_CREATE,
4194                                                vxlan->cfg.dst_port,
4195                                                conf.vni, conf.vni,
4196                                                conf.remote_ifindex,
4197                                                NTF_SELF, 0, true, extack);
4198                         if (err) {
4199                                 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4200                                 netdev_adjacent_change_abort(dst->remote_dev,
4201                                                              lowerdev, dev);
4202                                 return err;
4203                         }
4204                 }
4205                 if (!vxlan_addr_any(&dst->remote_ip))
4206                         __vxlan_fdb_delete(vxlan, all_zeros_mac,
4207                                            dst->remote_ip,
4208                                            vxlan->cfg.dst_port,
4209                                            dst->remote_vni,
4210                                            dst->remote_vni,
4211                                            dst->remote_ifindex,
4212                                            true);
4213                 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4214
4215                 /* If vni filtering device, also update fdb entries of
4216                  * all vnis that were using default remote ip
4217                  */
4218                 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
4219                         err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip,
4220                                                          &conf.remote_ip, extack);
4221                         if (err) {
4222                                 netdev_adjacent_change_abort(dst->remote_dev,
4223                                                              lowerdev, dev);
4224                                 return err;
4225                         }
4226                 }
4227         }
4228
4229         if (conf.age_interval != vxlan->cfg.age_interval)
4230                 mod_timer(&vxlan->age_timer, jiffies);
4231
4232         netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4233         if (lowerdev && lowerdev != dst->remote_dev)
4234                 dst->remote_dev = lowerdev;
4235         vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4236         return 0;
4237 }
4238
4239 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4240 {
4241         struct vxlan_dev *vxlan = netdev_priv(dev);
4242
4243         vxlan_flush(vxlan, true);
4244
4245         list_del(&vxlan->next);
4246         unregister_netdevice_queue(dev, head);
4247         if (vxlan->default_dst.remote_dev)
4248                 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4249 }
4250
4251 static size_t vxlan_get_size(const struct net_device *dev)
4252 {
4253
4254         return nla_total_size(sizeof(__u32)) +  /* IFLA_VXLAN_ID */
4255                 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4256                 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
4257                 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4258                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_TTL */
4259                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_TTL_INHERIT */
4260                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_TOS */
4261                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_DF */
4262                 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4263                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_LEARNING */
4264                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_PROXY */
4265                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_RSC */
4266                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_L2MISS */
4267                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_L3MISS */
4268                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_COLLECT_METADATA */
4269                 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
4270                 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
4271                 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4272                 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4273                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4274                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4275                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4276                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4277                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4278                 0;
4279 }
4280
4281 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4282 {
4283         const struct vxlan_dev *vxlan = netdev_priv(dev);
4284         const struct vxlan_rdst *dst = &vxlan->default_dst;
4285         struct ifla_vxlan_port_range ports = {
4286                 .low =  htons(vxlan->cfg.port_min),
4287                 .high = htons(vxlan->cfg.port_max),
4288         };
4289
4290         if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4291                 goto nla_put_failure;
4292
4293         if (!vxlan_addr_any(&dst->remote_ip)) {
4294                 if (dst->remote_ip.sa.sa_family == AF_INET) {
4295                         if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4296                                             dst->remote_ip.sin.sin_addr.s_addr))
4297                                 goto nla_put_failure;
4298 #if IS_ENABLED(CONFIG_IPV6)
4299                 } else {
4300                         if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4301                                              &dst->remote_ip.sin6.sin6_addr))
4302                                 goto nla_put_failure;
4303 #endif
4304                 }
4305         }
4306
4307         if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4308                 goto nla_put_failure;
4309
4310         if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4311                 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4312                         if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4313                                             vxlan->cfg.saddr.sin.sin_addr.s_addr))
4314                                 goto nla_put_failure;
4315 #if IS_ENABLED(CONFIG_IPV6)
4316                 } else {
4317                         if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4318                                              &vxlan->cfg.saddr.sin6.sin6_addr))
4319                                 goto nla_put_failure;
4320 #endif
4321                 }
4322         }
4323
4324         if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4325             nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4326                        !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4327             nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4328             nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4329             nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4330             nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4331                        !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4332             nla_put_u8(skb, IFLA_VXLAN_PROXY,
4333                        !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4334             nla_put_u8(skb, IFLA_VXLAN_RSC,
4335                        !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4336             nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4337                        !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4338             nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4339                        !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4340             nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4341                        !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4342             nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4343             nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4344             nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4345             nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4346                        !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4347             nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4348                        !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4349             nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4350                        !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4351             nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4352                        !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4353             nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4354                        !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
4355                 goto nla_put_failure;
4356
4357         if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4358                 goto nla_put_failure;
4359
4360         if (vxlan->cfg.flags & VXLAN_F_GBP &&
4361             nla_put_flag(skb, IFLA_VXLAN_GBP))
4362                 goto nla_put_failure;
4363
4364         if (vxlan->cfg.flags & VXLAN_F_GPE &&
4365             nla_put_flag(skb, IFLA_VXLAN_GPE))
4366                 goto nla_put_failure;
4367
4368         if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4369             nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4370                 goto nla_put_failure;
4371
4372         if (vxlan->cfg.flags & VXLAN_F_VNIFILTER &&
4373             nla_put_u8(skb, IFLA_VXLAN_VNIFILTER,
4374                        !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER)))
4375                 goto nla_put_failure;
4376
4377         return 0;
4378
4379 nla_put_failure:
4380         return -EMSGSIZE;
4381 }
4382
4383 static struct net *vxlan_get_link_net(const struct net_device *dev)
4384 {
4385         struct vxlan_dev *vxlan = netdev_priv(dev);
4386
4387         return vxlan->net;
4388 }
4389
4390 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4391         .kind           = "vxlan",
4392         .maxtype        = IFLA_VXLAN_MAX,
4393         .policy         = vxlan_policy,
4394         .priv_size      = sizeof(struct vxlan_dev),
4395         .setup          = vxlan_setup,
4396         .validate       = vxlan_validate,
4397         .newlink        = vxlan_newlink,
4398         .changelink     = vxlan_changelink,
4399         .dellink        = vxlan_dellink,
4400         .get_size       = vxlan_get_size,
4401         .fill_info      = vxlan_fill_info,
4402         .get_link_net   = vxlan_get_link_net,
4403 };
4404
4405 struct net_device *vxlan_dev_create(struct net *net, const char *name,
4406                                     u8 name_assign_type,
4407                                     struct vxlan_config *conf)
4408 {
4409         struct nlattr *tb[IFLA_MAX + 1];
4410         struct net_device *dev;
4411         int err;
4412
4413         memset(&tb, 0, sizeof(tb));
4414
4415         dev = rtnl_create_link(net, name, name_assign_type,
4416                                &vxlan_link_ops, tb, NULL);
4417         if (IS_ERR(dev))
4418                 return dev;
4419
4420         err = __vxlan_dev_create(net, dev, conf, NULL);
4421         if (err < 0) {
4422                 free_netdev(dev);
4423                 return ERR_PTR(err);
4424         }
4425
4426         err = rtnl_configure_link(dev, NULL, 0, NULL);
4427         if (err < 0) {
4428                 LIST_HEAD(list_kill);
4429
4430                 vxlan_dellink(dev, &list_kill);
4431                 unregister_netdevice_many(&list_kill);
4432                 return ERR_PTR(err);
4433         }
4434
4435         return dev;
4436 }
4437 EXPORT_SYMBOL_GPL(vxlan_dev_create);
4438
4439 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4440                                              struct net_device *dev)
4441 {
4442         struct vxlan_dev *vxlan, *next;
4443         LIST_HEAD(list_kill);
4444
4445         list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4446                 struct vxlan_rdst *dst = &vxlan->default_dst;
4447
4448                 /* In case we created vxlan device with carrier
4449                  * and we loose the carrier due to module unload
4450                  * we also need to remove vxlan device. In other
4451                  * cases, it's not necessary and remote_ifindex
4452                  * is 0 here, so no matches.
4453                  */
4454                 if (dst->remote_ifindex == dev->ifindex)
4455                         vxlan_dellink(vxlan->dev, &list_kill);
4456         }
4457
4458         unregister_netdevice_many(&list_kill);
4459 }
4460
4461 static int vxlan_netdevice_event(struct notifier_block *unused,
4462                                  unsigned long event, void *ptr)
4463 {
4464         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4465         struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4466
4467         if (event == NETDEV_UNREGISTER)
4468                 vxlan_handle_lowerdev_unregister(vn, dev);
4469         else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
4470                 vxlan_offload_rx_ports(dev, true);
4471         else if (event == NETDEV_UDP_TUNNEL_DROP_INFO)
4472                 vxlan_offload_rx_ports(dev, false);
4473
4474         return NOTIFY_DONE;
4475 }
4476
4477 static struct notifier_block vxlan_notifier_block __read_mostly = {
4478         .notifier_call = vxlan_netdevice_event,
4479 };
4480
4481 static void
4482 vxlan_fdb_offloaded_set(struct net_device *dev,
4483                         struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4484 {
4485         struct vxlan_dev *vxlan = netdev_priv(dev);
4486         struct vxlan_rdst *rdst;
4487         struct vxlan_fdb *f;
4488         u32 hash_index;
4489
4490         hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4491
4492         spin_lock_bh(&vxlan->hash_lock[hash_index]);
4493
4494         f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4495         if (!f)
4496                 goto out;
4497
4498         rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4499                                    fdb_info->remote_port,
4500                                    fdb_info->remote_vni,
4501                                    fdb_info->remote_ifindex);
4502         if (!rdst)
4503                 goto out;
4504
4505         rdst->offloaded = fdb_info->offloaded;
4506
4507 out:
4508         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4509 }
4510
4511 static int
4512 vxlan_fdb_external_learn_add(struct net_device *dev,
4513                              struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4514 {
4515         struct vxlan_dev *vxlan = netdev_priv(dev);
4516         struct netlink_ext_ack *extack;
4517         u32 hash_index;
4518         int err;
4519
4520         hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4521         extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4522
4523         spin_lock_bh(&vxlan->hash_lock[hash_index]);
4524         err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4525                                NUD_REACHABLE,
4526                                NLM_F_CREATE | NLM_F_REPLACE,
4527                                fdb_info->remote_port,
4528                                fdb_info->vni,
4529                                fdb_info->remote_vni,
4530                                fdb_info->remote_ifindex,
4531                                NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4532                                0, false, extack);
4533         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4534
4535         return err;
4536 }
4537
4538 static int
4539 vxlan_fdb_external_learn_del(struct net_device *dev,
4540                              struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4541 {
4542         struct vxlan_dev *vxlan = netdev_priv(dev);
4543         struct vxlan_fdb *f;
4544         u32 hash_index;
4545         int err = 0;
4546
4547         hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4548         spin_lock_bh(&vxlan->hash_lock[hash_index]);
4549
4550         f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4551         if (!f)
4552                 err = -ENOENT;
4553         else if (f->flags & NTF_EXT_LEARNED)
4554                 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4555                                          fdb_info->remote_ip,
4556                                          fdb_info->remote_port,
4557                                          fdb_info->vni,
4558                                          fdb_info->remote_vni,
4559                                          fdb_info->remote_ifindex,
4560                                          false);
4561
4562         spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4563
4564         return err;
4565 }
4566
4567 static int vxlan_switchdev_event(struct notifier_block *unused,
4568                                  unsigned long event, void *ptr)
4569 {
4570         struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4571         struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4572         int err = 0;
4573
4574         switch (event) {
4575         case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4576                 vxlan_fdb_offloaded_set(dev, ptr);
4577                 break;
4578         case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4579                 fdb_info = ptr;
4580                 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4581                 if (err) {
4582                         err = notifier_from_errno(err);
4583                         break;
4584                 }
4585                 fdb_info->offloaded = true;
4586                 vxlan_fdb_offloaded_set(dev, fdb_info);
4587                 break;
4588         case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4589                 fdb_info = ptr;
4590                 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4591                 if (err) {
4592                         err = notifier_from_errno(err);
4593                         break;
4594                 }
4595                 fdb_info->offloaded = false;
4596                 vxlan_fdb_offloaded_set(dev, fdb_info);
4597                 break;
4598         }
4599
4600         return err;
4601 }
4602
4603 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4604         .notifier_call = vxlan_switchdev_event,
4605 };
4606
4607 static void vxlan_fdb_nh_flush(struct nexthop *nh)
4608 {
4609         struct vxlan_fdb *fdb;
4610         struct vxlan_dev *vxlan;
4611         u32 hash_index;
4612
4613         rcu_read_lock();
4614         list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4615                 vxlan = rcu_dereference(fdb->vdev);
4616                 WARN_ON(!vxlan);
4617                 hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4618                                             vxlan->default_dst.remote_vni);
4619                 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4620                 if (!hlist_unhashed(&fdb->hlist))
4621                         vxlan_fdb_destroy(vxlan, fdb, false, false);
4622                 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4623         }
4624         rcu_read_unlock();
4625 }
4626
4627 static int vxlan_nexthop_event(struct notifier_block *nb,
4628                                unsigned long event, void *ptr)
4629 {
4630         struct nh_notifier_info *info = ptr;
4631         struct nexthop *nh;
4632
4633         if (event != NEXTHOP_EVENT_DEL)
4634                 return NOTIFY_DONE;
4635
4636         nh = nexthop_find_by_id(info->net, info->id);
4637         if (!nh)
4638                 return NOTIFY_DONE;
4639
4640         vxlan_fdb_nh_flush(nh);
4641
4642         return NOTIFY_DONE;
4643 }
4644
4645 static __net_init int vxlan_init_net(struct net *net)
4646 {
4647         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4648         unsigned int h;
4649
4650         INIT_LIST_HEAD(&vn->vxlan_list);
4651         spin_lock_init(&vn->sock_lock);
4652         vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event;
4653
4654         for (h = 0; h < PORT_HASH_SIZE; ++h)
4655                 INIT_HLIST_HEAD(&vn->sock_list[h]);
4656
4657         return register_nexthop_notifier(net, &vn->nexthop_notifier_block,
4658                                          NULL);
4659 }
4660
4661 static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
4662 {
4663         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4664         struct vxlan_dev *vxlan, *next;
4665         struct net_device *dev, *aux;
4666
4667         for_each_netdev_safe(net, dev, aux)
4668                 if (dev->rtnl_link_ops == &vxlan_link_ops)
4669                         unregister_netdevice_queue(dev, head);
4670
4671         list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4672                 /* If vxlan->dev is in the same netns, it has already been added
4673                  * to the list by the previous loop.
4674                  */
4675                 if (!net_eq(dev_net(vxlan->dev), net))
4676                         unregister_netdevice_queue(vxlan->dev, head);
4677         }
4678
4679 }
4680
4681 static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4682 {
4683         struct net *net;
4684         LIST_HEAD(list);
4685         unsigned int h;
4686
4687         list_for_each_entry(net, net_list, exit_list) {
4688                 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4689
4690                 unregister_nexthop_notifier(net, &vn->nexthop_notifier_block);
4691         }
4692         rtnl_lock();
4693         list_for_each_entry(net, net_list, exit_list)
4694                 vxlan_destroy_tunnels(net, &list);
4695
4696         unregister_netdevice_many(&list);
4697         rtnl_unlock();
4698
4699         list_for_each_entry(net, net_list, exit_list) {
4700                 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4701
4702                 for (h = 0; h < PORT_HASH_SIZE; ++h)
4703                         WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4704         }
4705 }
4706
4707 static struct pernet_operations vxlan_net_ops = {
4708         .init = vxlan_init_net,
4709         .exit_batch = vxlan_exit_batch_net,
4710         .id   = &vxlan_net_id,
4711         .size = sizeof(struct vxlan_net),
4712 };
4713
4714 static int __init vxlan_init_module(void)
4715 {
4716         int rc;
4717
4718         get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4719
4720         rc = register_pernet_subsys(&vxlan_net_ops);
4721         if (rc)
4722                 goto out1;
4723
4724         rc = register_netdevice_notifier(&vxlan_notifier_block);
4725         if (rc)
4726                 goto out2;
4727
4728         rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4729         if (rc)
4730                 goto out3;
4731
4732         rc = rtnl_link_register(&vxlan_link_ops);
4733         if (rc)
4734                 goto out4;
4735
4736         vxlan_vnifilter_init();
4737
4738         return 0;
4739 out4:
4740         unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4741 out3:
4742         unregister_netdevice_notifier(&vxlan_notifier_block);
4743 out2:
4744         unregister_pernet_subsys(&vxlan_net_ops);
4745 out1:
4746         return rc;
4747 }
4748 late_initcall(vxlan_init_module);
4749
4750 static void __exit vxlan_cleanup_module(void)
4751 {
4752         vxlan_vnifilter_uninit();
4753         rtnl_link_unregister(&vxlan_link_ops);
4754         unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4755         unregister_netdevice_notifier(&vxlan_notifier_block);
4756         unregister_pernet_subsys(&vxlan_net_ops);
4757         /* rcu_barrier() is called by netns */
4758 }
4759 module_exit(vxlan_cleanup_module);
4760
4761 MODULE_LICENSE("GPL");
4762 MODULE_VERSION(VXLAN_VERSION);
4763 MODULE_AUTHOR("Stephen Hemminger <[email protected]>");
4764 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4765 MODULE_ALIAS_RTNL_LINK("vxlan");