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