linux/drivers/net/vxlan.c
<<
>>
Prefs
   1/*
   2 * VXLAN: Virtual eXtensible Local Area Network
   3 *
   4 * Copyright (c) 2012-2013 Vyatta Inc.
   5 *
   6 * This program is free software; you can redistribute it and/or modify
   7 * it under the terms of the GNU General Public License version 2 as
   8 * published by the Free Software Foundation.
   9 */
  10
  11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  12
  13#include <linux/kernel.h>
  14#include <linux/types.h>
  15#include <linux/module.h>
  16#include <linux/errno.h>
  17#include <linux/slab.h>
  18#include <linux/skbuff.h>
  19#include <linux/rculist.h>
  20#include <linux/netdevice.h>
  21#include <linux/in.h>
  22#include <linux/ip.h>
  23#include <linux/udp.h>
  24#include <linux/igmp.h>
  25#include <linux/etherdevice.h>
  26#include <linux/if_ether.h>
  27#include <linux/if_vlan.h>
  28#include <linux/hash.h>
  29#include <linux/ethtool.h>
  30#include <net/arp.h>
  31#include <net/ndisc.h>
  32#include <net/ip.h>
  33#include <net/ip_tunnels.h>
  34#include <net/icmp.h>
  35#include <net/udp.h>
  36#include <net/rtnetlink.h>
  37#include <net/route.h>
  38#include <net/dsfield.h>
  39#include <net/inet_ecn.h>
  40#include <net/net_namespace.h>
  41#include <net/netns/generic.h>
  42#include <net/vxlan.h>
  43#if IS_ENABLED(CONFIG_IPV6)
  44#include <net/ipv6.h>
  45#include <net/addrconf.h>
  46#include <net/ip6_tunnel.h>
  47#include <net/ip6_checksum.h>
  48#endif
  49
  50#define VXLAN_VERSION   "0.1"
  51
  52#define PORT_HASH_BITS  8
  53#define PORT_HASH_SIZE  (1<<PORT_HASH_BITS)
  54#define VNI_HASH_BITS   10
  55#define VNI_HASH_SIZE   (1<<VNI_HASH_BITS)
  56#define FDB_HASH_BITS   8
  57#define FDB_HASH_SIZE   (1<<FDB_HASH_BITS)
  58#define FDB_AGE_DEFAULT 300 /* 5 min */
  59#define FDB_AGE_INTERVAL (10 * HZ)      /* rescan interval */
  60
  61#define VXLAN_N_VID     (1u << 24)
  62#define VXLAN_VID_MASK  (VXLAN_N_VID - 1)
  63/* IP header + UDP + VXLAN + Ethernet header */
  64#define VXLAN_HEADROOM (20 + 8 + 8 + 14)
  65/* IPv6 header + UDP + VXLAN + Ethernet header */
  66#define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
  67#define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
  68
  69#define VXLAN_FLAGS 0x08000000  /* struct vxlanhdr.vx_flags required value. */
  70
  71/* VXLAN protocol header */
  72struct vxlanhdr {
  73        __be32 vx_flags;
  74        __be32 vx_vni;
  75};
  76
  77/* UDP port for VXLAN traffic.
  78 * The IANA assigned port is 4789, but the Linux default is 8472
  79 * for compatibility with early adopters.
  80 */
  81static unsigned short vxlan_port __read_mostly = 8472;
  82module_param_named(udp_port, vxlan_port, ushort, 0444);
  83MODULE_PARM_DESC(udp_port, "Destination UDP port");
  84
  85static bool log_ecn_error = true;
  86module_param(log_ecn_error, bool, 0644);
  87MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
  88
  89static int vxlan_net_id;
  90
  91static const u8 all_zeros_mac[ETH_ALEN];
  92
  93/* per-network namespace private data for this module */
  94struct vxlan_net {
  95        struct list_head  vxlan_list;
  96        struct hlist_head sock_list[PORT_HASH_SIZE];
  97        spinlock_t        sock_lock;
  98};
  99
 100union vxlan_addr {
 101        struct sockaddr_in sin;
 102        struct sockaddr_in6 sin6;
 103        struct sockaddr sa;
 104};
 105
 106struct vxlan_rdst {
 107        union vxlan_addr         remote_ip;
 108        __be16                   remote_port;
 109        u32                      remote_vni;
 110        u32                      remote_ifindex;
 111        struct list_head         list;
 112        struct rcu_head          rcu;
 113};
 114
 115/* Forwarding table entry */
 116struct vxlan_fdb {
 117        struct hlist_node hlist;        /* linked list of entries */
 118        struct rcu_head   rcu;
 119        unsigned long     updated;      /* jiffies */
 120        unsigned long     used;
 121        struct list_head  remotes;
 122        u16               state;        /* see ndm_state */
 123        u8                flags;        /* see ndm_flags */
 124        u8                eth_addr[ETH_ALEN];
 125};
 126
 127/* Pseudo network device */
 128struct vxlan_dev {
 129        struct hlist_node hlist;        /* vni hash table */
 130        struct list_head  next;         /* vxlan's per namespace list */
 131        struct vxlan_sock *vn_sock;     /* listening socket */
 132        struct net_device *dev;
 133        struct vxlan_rdst default_dst;  /* default destination */
 134        union vxlan_addr  saddr;        /* source address */
 135        __be16            dst_port;
 136        __u16             port_min;     /* source port range */
 137        __u16             port_max;
 138        __u8              tos;          /* TOS override */
 139        __u8              ttl;
 140        u32               flags;        /* VXLAN_F_* below */
 141
 142        struct work_struct sock_work;
 143        struct work_struct igmp_join;
 144        struct work_struct igmp_leave;
 145
 146        unsigned long     age_interval;
 147        struct timer_list age_timer;
 148        spinlock_t        hash_lock;
 149        unsigned int      addrcnt;
 150        unsigned int      addrmax;
 151
 152        struct hlist_head fdb_head[FDB_HASH_SIZE];
 153};
 154
 155#define VXLAN_F_LEARN   0x01
 156#define VXLAN_F_PROXY   0x02
 157#define VXLAN_F_RSC     0x04
 158#define VXLAN_F_L2MISS  0x08
 159#define VXLAN_F_L3MISS  0x10
 160#define VXLAN_F_IPV6    0x20 /* internal flag */
 161
 162/* salt for hash table */
 163static u32 vxlan_salt __read_mostly;
 164static struct workqueue_struct *vxlan_wq;
 165
 166static void vxlan_sock_work(struct work_struct *work);
 167
 168#if IS_ENABLED(CONFIG_IPV6)
 169static inline
 170bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
 171{
 172       if (a->sa.sa_family != b->sa.sa_family)
 173               return false;
 174       if (a->sa.sa_family == AF_INET6)
 175               return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
 176       else
 177               return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
 178}
 179
 180static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
 181{
 182       if (ipa->sa.sa_family == AF_INET6)
 183               return ipv6_addr_any(&ipa->sin6.sin6_addr);
 184       else
 185               return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
 186}
 187
 188static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
 189{
 190       if (ipa->sa.sa_family == AF_INET6)
 191               return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
 192       else
 193               return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
 194}
 195
 196static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
 197{
 198       if (nla_len(nla) >= sizeof(struct in6_addr)) {
 199               nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr));
 200               ip->sa.sa_family = AF_INET6;
 201               return 0;
 202       } else if (nla_len(nla) >= sizeof(__be32)) {
 203               ip->sin.sin_addr.s_addr = nla_get_be32(nla);
 204               ip->sa.sa_family = AF_INET;
 205               return 0;
 206       } else {
 207               return -EAFNOSUPPORT;
 208       }
 209}
 210
 211static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
 212                             const union vxlan_addr *ip)
 213{
 214       if (ip->sa.sa_family == AF_INET6)
 215               return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr);
 216       else
 217               return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
 218}
 219
 220#else /* !CONFIG_IPV6 */
 221
 222static inline
 223bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
 224{
 225       return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
 226}
 227
 228static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
 229{
 230       return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
 231}
 232
 233static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
 234{
 235       return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
 236}
 237
 238static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
 239{
 240       if (nla_len(nla) >= sizeof(struct in6_addr)) {
 241               return -EAFNOSUPPORT;
 242       } else if (nla_len(nla) >= sizeof(__be32)) {
 243               ip->sin.sin_addr.s_addr = nla_get_be32(nla);
 244               ip->sa.sa_family = AF_INET;
 245               return 0;
 246       } else {
 247               return -EAFNOSUPPORT;
 248       }
 249}
 250
 251static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
 252                             const union vxlan_addr *ip)
 253{
 254       return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
 255}
 256#endif
 257
 258/* Virtual Network hash table head */
 259static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
 260{
 261        return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
 262}
 263
 264/* Socket hash table head */
 265static inline struct hlist_head *vs_head(struct net *net, __be16 port)
 266{
 267        struct vxlan_net *vn = net_generic(net, vxlan_net_id);
 268
 269        return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
 270}
 271
 272/* First remote destination for a forwarding entry.
 273 * Guaranteed to be non-NULL because remotes are never deleted.
 274 */
 275static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
 276{
 277        return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
 278}
 279
 280static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
 281{
 282        return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
 283}
 284
 285/* Find VXLAN socket based on network namespace and UDP port */
 286static struct vxlan_sock *vxlan_find_sock(struct net *net, __be16 port)
 287{
 288        struct vxlan_sock *vs;
 289
 290        hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
 291                if (inet_sk(vs->sock->sk)->inet_sport == port)
 292                        return vs;
 293        }
 294        return NULL;
 295}
 296
 297static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
 298{
 299        struct vxlan_dev *vxlan;
 300
 301        hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
 302                if (vxlan->default_dst.remote_vni == id)
 303                        return vxlan;
 304        }
 305
 306        return NULL;
 307}
 308
 309/* Look up VNI in a per net namespace table */
 310static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port)
 311{
 312        struct vxlan_sock *vs;
 313
 314        vs = vxlan_find_sock(net, port);
 315        if (!vs)
 316                return NULL;
 317
 318        return vxlan_vs_find_vni(vs, id);
 319}
 320
 321/* Fill in neighbour message in skbuff. */
 322static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
 323                          const struct vxlan_fdb *fdb,
 324                          u32 portid, u32 seq, int type, unsigned int flags,
 325                          const struct vxlan_rdst *rdst)
 326{
 327        unsigned long now = jiffies;
 328        struct nda_cacheinfo ci;
 329        struct nlmsghdr *nlh;
 330        struct ndmsg *ndm;
 331        bool send_ip, send_eth;
 332
 333        nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
 334        if (nlh == NULL)
 335                return -EMSGSIZE;
 336
 337        ndm = nlmsg_data(nlh);
 338        memset(ndm, 0, sizeof(*ndm));
 339
 340        send_eth = send_ip = true;
 341
 342        if (type == RTM_GETNEIGH) {
 343                ndm->ndm_family = AF_INET;
 344                send_ip = !vxlan_addr_any(&rdst->remote_ip);
 345                send_eth = !is_zero_ether_addr(fdb->eth_addr);
 346        } else
 347                ndm->ndm_family = AF_BRIDGE;
 348        ndm->ndm_state = fdb->state;
 349        ndm->ndm_ifindex = vxlan->dev->ifindex;
 350        ndm->ndm_flags = fdb->flags;
 351        ndm->ndm_type = NDA_DST;
 352
 353        if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
 354                goto nla_put_failure;
 355
 356        if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
 357                goto nla_put_failure;
 358
 359        if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
 360            nla_put_be16(skb, NDA_PORT, rdst->remote_port))
 361                goto nla_put_failure;
 362        if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
 363            nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
 364                goto nla_put_failure;
 365        if (rdst->remote_ifindex &&
 366            nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
 367                goto nla_put_failure;
 368
 369        ci.ndm_used      = jiffies_to_clock_t(now - fdb->used);
 370        ci.ndm_confirmed = 0;
 371        ci.ndm_updated   = jiffies_to_clock_t(now - fdb->updated);
 372        ci.ndm_refcnt    = 0;
 373
 374        if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
 375                goto nla_put_failure;
 376
 377        return nlmsg_end(skb, nlh);
 378
 379nla_put_failure:
 380        nlmsg_cancel(skb, nlh);
 381        return -EMSGSIZE;
 382}
 383
 384static inline size_t vxlan_nlmsg_size(void)
 385{
 386        return NLMSG_ALIGN(sizeof(struct ndmsg))
 387                + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
 388                + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
 389                + nla_total_size(sizeof(__be16)) /* NDA_PORT */
 390                + nla_total_size(sizeof(__be32)) /* NDA_VNI */
 391                + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
 392                + nla_total_size(sizeof(struct nda_cacheinfo));
 393}
 394
 395static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
 396                             struct vxlan_fdb *fdb, int type)
 397{
 398        struct net *net = dev_net(vxlan->dev);
 399        struct sk_buff *skb;
 400        int err = -ENOBUFS;
 401
 402        skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
 403        if (skb == NULL)
 404                goto errout;
 405
 406        err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0,
 407                             first_remote_rtnl(fdb));
 408        if (err < 0) {
 409                /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
 410                WARN_ON(err == -EMSGSIZE);
 411                kfree_skb(skb);
 412                goto errout;
 413        }
 414
 415        rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
 416        return;
 417errout:
 418        if (err < 0)
 419                rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
 420}
 421
 422static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
 423{
 424        struct vxlan_dev *vxlan = netdev_priv(dev);
 425        struct vxlan_fdb f = {
 426                .state = NUD_STALE,
 427        };
 428        struct vxlan_rdst remote = {
 429                .remote_ip = *ipa, /* goes to NDA_DST */
 430                .remote_vni = VXLAN_N_VID,
 431        };
 432
 433        INIT_LIST_HEAD(&f.remotes);
 434        list_add_rcu(&remote.list, &f.remotes);
 435
 436        vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
 437}
 438
 439static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
 440{
 441        struct vxlan_fdb f = {
 442                .state = NUD_STALE,
 443        };
 444
 445        INIT_LIST_HEAD(&f.remotes);
 446        memcpy(f.eth_addr, eth_addr, ETH_ALEN);
 447
 448        vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
 449}
 450
 451/* Hash Ethernet address */
 452static u32 eth_hash(const unsigned char *addr)
 453{
 454        u64 value = get_unaligned((u64 *)addr);
 455
 456        /* only want 6 bytes */
 457#ifdef __BIG_ENDIAN
 458        value >>= 16;
 459#else
 460        value <<= 16;
 461#endif
 462        return hash_64(value, FDB_HASH_BITS);
 463}
 464
 465/* Hash chain to use given mac address */
 466static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
 467                                                const u8 *mac)
 468{
 469        return &vxlan->fdb_head[eth_hash(mac)];
 470}
 471
 472/* Look up Ethernet address in forwarding table */
 473static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
 474                                        const u8 *mac)
 475
 476{
 477        struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
 478        struct vxlan_fdb *f;
 479
 480        hlist_for_each_entry_rcu(f, head, hlist) {
 481                if (ether_addr_equal(mac, f->eth_addr))
 482                        return f;
 483        }
 484
 485        return NULL;
 486}
 487
 488static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
 489                                        const u8 *mac)
 490{
 491        struct vxlan_fdb *f;
 492
 493        f = __vxlan_find_mac(vxlan, mac);
 494        if (f)
 495                f->used = jiffies;
 496
 497        return f;
 498}
 499
 500/* caller should hold vxlan->hash_lock */
 501static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
 502                                              union vxlan_addr *ip, __be16 port,
 503                                              __u32 vni, __u32 ifindex)
 504{
 505        struct vxlan_rdst *rd;
 506
 507        list_for_each_entry(rd, &f->remotes, list) {
 508                if (vxlan_addr_equal(&rd->remote_ip, ip) &&
 509                    rd->remote_port == port &&
 510                    rd->remote_vni == vni &&
 511                    rd->remote_ifindex == ifindex)
 512                        return rd;
 513        }
 514
 515        return NULL;
 516}
 517
 518/* Replace destination of unicast mac */
 519static int vxlan_fdb_replace(struct vxlan_fdb *f,
 520                             union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
 521{
 522        struct vxlan_rdst *rd;
 523
 524        rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
 525        if (rd)
 526                return 0;
 527
 528        rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
 529        if (!rd)
 530                return 0;
 531        rd->remote_ip = *ip;
 532        rd->remote_port = port;
 533        rd->remote_vni = vni;
 534        rd->remote_ifindex = ifindex;
 535        return 1;
 536}
 537
 538/* Add/update destinations for multicast */
 539static int vxlan_fdb_append(struct vxlan_fdb *f,
 540                            union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
 541{
 542        struct vxlan_rdst *rd;
 543
 544        rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
 545        if (rd)
 546                return 0;
 547
 548        rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
 549        if (rd == NULL)
 550                return -ENOBUFS;
 551        rd->remote_ip = *ip;
 552        rd->remote_port = port;
 553        rd->remote_vni = vni;
 554        rd->remote_ifindex = ifindex;
 555
 556        list_add_tail_rcu(&rd->list, &f->remotes);
 557
 558        return 1;
 559}
 560
 561/* Notify netdevs that UDP port started listening */
 562static void vxlan_notify_add_rx_port(struct sock *sk)
 563{
 564        struct net_device *dev;
 565        struct net *net = sock_net(sk);
 566        sa_family_t sa_family = sk->sk_family;
 567        __be16 port = inet_sk(sk)->inet_sport;
 568
 569        rcu_read_lock();
 570        for_each_netdev_rcu(net, dev) {
 571                if (dev->netdev_ops->ndo_add_vxlan_port)
 572                        dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
 573                                                            port);
 574        }
 575        rcu_read_unlock();
 576}
 577
 578/* Notify netdevs that UDP port is no more listening */
 579static void vxlan_notify_del_rx_port(struct sock *sk)
 580{
 581        struct net_device *dev;
 582        struct net *net = sock_net(sk);
 583        sa_family_t sa_family = sk->sk_family;
 584        __be16 port = inet_sk(sk)->inet_sport;
 585
 586        rcu_read_lock();
 587        for_each_netdev_rcu(net, dev) {
 588                if (dev->netdev_ops->ndo_del_vxlan_port)
 589                        dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
 590                                                            port);
 591        }
 592        rcu_read_unlock();
 593}
 594
 595/* Add new entry to forwarding table -- assumes lock held */
 596static int vxlan_fdb_create(struct vxlan_dev *vxlan,
 597                            const u8 *mac, union vxlan_addr *ip,
 598                            __u16 state, __u16 flags,
 599                            __be16 port, __u32 vni, __u32 ifindex,
 600                            __u8 ndm_flags)
 601{
 602        struct vxlan_fdb *f;
 603        int notify = 0;
 604
 605        f = __vxlan_find_mac(vxlan, mac);
 606        if (f) {
 607                if (flags & NLM_F_EXCL) {
 608                        netdev_dbg(vxlan->dev,
 609                                   "lost race to create %pM\n", mac);
 610                        return -EEXIST;
 611                }
 612                if (f->state != state) {
 613                        f->state = state;
 614                        f->updated = jiffies;
 615                        notify = 1;
 616                }
 617                if (f->flags != ndm_flags) {
 618                        f->flags = ndm_flags;
 619                        f->updated = jiffies;
 620                        notify = 1;
 621                }
 622                if ((flags & NLM_F_REPLACE)) {
 623                        /* Only change unicasts */
 624                        if (!(is_multicast_ether_addr(f->eth_addr) ||
 625                             is_zero_ether_addr(f->eth_addr))) {
 626                                int rc = vxlan_fdb_replace(f, ip, port, vni,
 627                                                           ifindex);
 628
 629                                if (rc < 0)
 630                                        return rc;
 631                                notify |= rc;
 632                        } else
 633                                return -EOPNOTSUPP;
 634                }
 635                if ((flags & NLM_F_APPEND) &&
 636                    (is_multicast_ether_addr(f->eth_addr) ||
 637                     is_zero_ether_addr(f->eth_addr))) {
 638                        int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
 639
 640                        if (rc < 0)
 641                                return rc;
 642                        notify |= rc;
 643                }
 644        } else {
 645                if (!(flags & NLM_F_CREATE))
 646                        return -ENOENT;
 647
 648                if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
 649                        return -ENOSPC;
 650
 651                /* Disallow replace to add a multicast entry */
 652                if ((flags & NLM_F_REPLACE) &&
 653                    (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
 654                        return -EOPNOTSUPP;
 655
 656                netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
 657                f = kmalloc(sizeof(*f), GFP_ATOMIC);
 658                if (!f)
 659                        return -ENOMEM;
 660
 661                notify = 1;
 662                f->state = state;
 663                f->flags = ndm_flags;
 664                f->updated = f->used = jiffies;
 665                INIT_LIST_HEAD(&f->remotes);
 666                memcpy(f->eth_addr, mac, ETH_ALEN);
 667
 668                vxlan_fdb_append(f, ip, port, vni, ifindex);
 669
 670                ++vxlan->addrcnt;
 671                hlist_add_head_rcu(&f->hlist,
 672                                   vxlan_fdb_head(vxlan, mac));
 673        }
 674
 675        if (notify)
 676                vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
 677
 678        return 0;
 679}
 680
 681static void vxlan_fdb_free(struct rcu_head *head)
 682{
 683        struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
 684        struct vxlan_rdst *rd, *nd;
 685
 686        list_for_each_entry_safe(rd, nd, &f->remotes, list)
 687                kfree(rd);
 688        kfree(f);
 689}
 690
 691static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
 692{
 693        netdev_dbg(vxlan->dev,
 694                    "delete %pM\n", f->eth_addr);
 695
 696        --vxlan->addrcnt;
 697        vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
 698
 699        hlist_del_rcu(&f->hlist);
 700        call_rcu(&f->rcu, vxlan_fdb_free);
 701}
 702
 703static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
 704                           union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
 705{
 706        struct net *net = dev_net(vxlan->dev);
 707        int err;
 708
 709        if (tb[NDA_DST]) {
 710                err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
 711                if (err)
 712                        return err;
 713        } else {
 714                union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
 715                if (remote->sa.sa_family == AF_INET) {
 716                        ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
 717                        ip->sa.sa_family = AF_INET;
 718#if IS_ENABLED(CONFIG_IPV6)
 719                } else {
 720                        ip->sin6.sin6_addr = in6addr_any;
 721                        ip->sa.sa_family = AF_INET6;
 722#endif
 723                }
 724        }
 725
 726        if (tb[NDA_PORT]) {
 727                if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
 728                        return -EINVAL;
 729                *port = nla_get_be16(tb[NDA_PORT]);
 730        } else {
 731                *port = vxlan->dst_port;
 732        }
 733
 734        if (tb[NDA_VNI]) {
 735                if (nla_len(tb[NDA_VNI]) != sizeof(u32))
 736                        return -EINVAL;
 737                *vni = nla_get_u32(tb[NDA_VNI]);
 738        } else {
 739                *vni = vxlan->default_dst.remote_vni;
 740        }
 741
 742        if (tb[NDA_IFINDEX]) {
 743                struct net_device *tdev;
 744
 745                if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
 746                        return -EINVAL;
 747                *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
 748                tdev = dev_get_by_index(net, *ifindex);
 749                if (!tdev)
 750                        return -EADDRNOTAVAIL;
 751                dev_put(tdev);
 752        } else {
 753                *ifindex = 0;
 754        }
 755
 756        return 0;
 757}
 758
 759/* Add static entry (via netlink) */
 760static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
 761                         struct net_device *dev,
 762                         const unsigned char *addr, u16 flags)
 763{
 764        struct vxlan_dev *vxlan = netdev_priv(dev);
 765        /* struct net *net = dev_net(vxlan->dev); */
 766        union vxlan_addr ip;
 767        __be16 port;
 768        u32 vni, ifindex;
 769        int err;
 770
 771        if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
 772                pr_info("RTM_NEWNEIGH with invalid state %#x\n",
 773                        ndm->ndm_state);
 774                return -EINVAL;
 775        }
 776
 777        if (tb[NDA_DST] == NULL)
 778                return -EINVAL;
 779
 780        err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
 781        if (err)
 782                return err;
 783
 784        spin_lock_bh(&vxlan->hash_lock);
 785        err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
 786                               port, vni, ifindex, ndm->ndm_flags);
 787        spin_unlock_bh(&vxlan->hash_lock);
 788
 789        return err;
 790}
 791
 792/* Delete entry (via netlink) */
 793static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
 794                            struct net_device *dev,
 795                            const unsigned char *addr)
 796{
 797        struct vxlan_dev *vxlan = netdev_priv(dev);
 798        struct vxlan_fdb *f;
 799        struct vxlan_rdst *rd = NULL;
 800        union vxlan_addr ip;
 801        __be16 port;
 802        u32 vni, ifindex;
 803        int err;
 804
 805        err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
 806        if (err)
 807                return err;
 808
 809        err = -ENOENT;
 810
 811        spin_lock_bh(&vxlan->hash_lock);
 812        f = vxlan_find_mac(vxlan, addr);
 813        if (!f)
 814                goto out;
 815
 816        if (!vxlan_addr_any(&ip)) {
 817                rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
 818                if (!rd)
 819                        goto out;
 820        }
 821
 822        err = 0;
 823
 824        /* remove a destination if it's not the only one on the list,
 825         * otherwise destroy the fdb entry
 826         */
 827        if (rd && !list_is_singular(&f->remotes)) {
 828                list_del_rcu(&rd->list);
 829                kfree_rcu(rd, rcu);
 830                goto out;
 831        }
 832
 833        vxlan_fdb_destroy(vxlan, f);
 834
 835out:
 836        spin_unlock_bh(&vxlan->hash_lock);
 837
 838        return err;
 839}
 840
 841/* Dump forwarding table */
 842static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
 843                          struct net_device *dev, int idx)
 844{
 845        struct vxlan_dev *vxlan = netdev_priv(dev);
 846        unsigned int h;
 847
 848        for (h = 0; h < FDB_HASH_SIZE; ++h) {
 849                struct vxlan_fdb *f;
 850                int err;
 851
 852                hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
 853                        struct vxlan_rdst *rd;
 854
 855                        if (idx < cb->args[0])
 856                                goto skip;
 857
 858                        list_for_each_entry_rcu(rd, &f->remotes, list) {
 859                                err = vxlan_fdb_info(skb, vxlan, f,
 860                                                     NETLINK_CB(cb->skb).portid,
 861                                                     cb->nlh->nlmsg_seq,
 862                                                     RTM_NEWNEIGH,
 863                                                     NLM_F_MULTI, rd);
 864                                if (err < 0)
 865                                        goto out;
 866                        }
 867skip:
 868                        ++idx;
 869                }
 870        }
 871out:
 872        return idx;
 873}
 874
 875/* Watch incoming packets to learn mapping between Ethernet address
 876 * and Tunnel endpoint.
 877 * Return true if packet is bogus and should be droppped.
 878 */
 879static bool vxlan_snoop(struct net_device *dev,
 880                        union vxlan_addr *src_ip, const u8 *src_mac)
 881{
 882        struct vxlan_dev *vxlan = netdev_priv(dev);
 883        struct vxlan_fdb *f;
 884
 885        f = vxlan_find_mac(vxlan, src_mac);
 886        if (likely(f)) {
 887                struct vxlan_rdst *rdst = first_remote_rcu(f);
 888
 889                if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
 890                        return false;
 891
 892                /* Don't migrate static entries, drop packets */
 893                if (f->state & NUD_NOARP)
 894                        return true;
 895
 896                if (net_ratelimit())
 897                        netdev_info(dev,
 898                                    "%pM migrated from %pIS to %pIS\n",
 899                                    src_mac, &rdst->remote_ip, &src_ip);
 900
 901                rdst->remote_ip = *src_ip;
 902                f->updated = jiffies;
 903                vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
 904        } else {
 905                /* learned new entry */
 906                spin_lock(&vxlan->hash_lock);
 907
 908                /* close off race between vxlan_flush and incoming packets */
 909                if (netif_running(dev))
 910                        vxlan_fdb_create(vxlan, src_mac, src_ip,
 911                                         NUD_REACHABLE,
 912                                         NLM_F_EXCL|NLM_F_CREATE,
 913                                         vxlan->dst_port,
 914                                         vxlan->default_dst.remote_vni,
 915                                         0, NTF_SELF);
 916                spin_unlock(&vxlan->hash_lock);
 917        }
 918
 919        return false;
 920}
 921
 922/* See if multicast group is already in use by other ID */
 923static bool vxlan_group_used(struct vxlan_net *vn, union vxlan_addr *remote_ip)
 924{
 925        struct vxlan_dev *vxlan;
 926
 927        list_for_each_entry(vxlan, &vn->vxlan_list, next) {
 928                if (!netif_running(vxlan->dev))
 929                        continue;
 930
 931                if (vxlan_addr_equal(&vxlan->default_dst.remote_ip,
 932                                     remote_ip))
 933                        return true;
 934        }
 935
 936        return false;
 937}
 938
 939static void vxlan_sock_hold(struct vxlan_sock *vs)
 940{
 941        atomic_inc(&vs->refcnt);
 942}
 943
 944void vxlan_sock_release(struct vxlan_sock *vs)
 945{
 946        struct sock *sk = vs->sock->sk;
 947        struct net *net = sock_net(sk);
 948        struct vxlan_net *vn = net_generic(net, vxlan_net_id);
 949
 950        if (!atomic_dec_and_test(&vs->refcnt))
 951                return;
 952
 953        spin_lock(&vn->sock_lock);
 954        hlist_del_rcu(&vs->hlist);
 955        rcu_assign_sk_user_data(vs->sock->sk, NULL);
 956        vxlan_notify_del_rx_port(sk);
 957        spin_unlock(&vn->sock_lock);
 958
 959        queue_work(vxlan_wq, &vs->del_work);
 960}
 961EXPORT_SYMBOL_GPL(vxlan_sock_release);
 962
 963/* Callback to update multicast group membership when first VNI on
 964 * multicast asddress is brought up
 965 * Done as workqueue because ip_mc_join_group acquires RTNL.
 966 */
 967static void vxlan_igmp_join(struct work_struct *work)
 968{
 969        struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
 970        struct vxlan_sock *vs = vxlan->vn_sock;
 971        struct sock *sk = vs->sock->sk;
 972        union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
 973        int ifindex = vxlan->default_dst.remote_ifindex;
 974
 975        lock_sock(sk);
 976        if (ip->sa.sa_family == AF_INET) {
 977                struct ip_mreqn mreq = {
 978                        .imr_multiaddr.s_addr   = ip->sin.sin_addr.s_addr,
 979                        .imr_ifindex            = ifindex,
 980                };
 981
 982                ip_mc_join_group(sk, &mreq);
 983#if IS_ENABLED(CONFIG_IPV6)
 984        } else {
 985                ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
 986                                             &ip->sin6.sin6_addr);
 987#endif
 988        }
 989        release_sock(sk);
 990
 991        vxlan_sock_release(vs);
 992        dev_put(vxlan->dev);
 993}
 994
 995/* Inverse of vxlan_igmp_join when last VNI is brought down */
 996static void vxlan_igmp_leave(struct work_struct *work)
 997{
 998        struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
 999        struct vxlan_sock *vs = vxlan->vn_sock;
1000        struct sock *sk = vs->sock->sk;
1001        union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1002        int ifindex = vxlan->default_dst.remote_ifindex;
1003
1004        lock_sock(sk);
1005        if (ip->sa.sa_family == AF_INET) {
1006                struct ip_mreqn mreq = {
1007                        .imr_multiaddr.s_addr   = ip->sin.sin_addr.s_addr,
1008                        .imr_ifindex            = ifindex,
1009                };
1010
1011                ip_mc_leave_group(sk, &mreq);
1012#if IS_ENABLED(CONFIG_IPV6)
1013        } else {
1014                ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1015                                             &ip->sin6.sin6_addr);
1016#endif
1017        }
1018
1019        release_sock(sk);
1020
1021        vxlan_sock_release(vs);
1022        dev_put(vxlan->dev);
1023}
1024
1025/* Callback from net/ipv4/udp.c to receive packets */
1026static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1027{
1028        struct vxlan_sock *vs;
1029        struct vxlanhdr *vxh;
1030        __be16 port;
1031
1032        /* Need Vxlan and inner Ethernet header to be present */
1033        if (!pskb_may_pull(skb, VXLAN_HLEN))
1034                goto error;
1035
1036        /* Return packets with reserved bits set */
1037        vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1038        if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
1039            (vxh->vx_vni & htonl(0xff))) {
1040                netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1041                           ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1042                goto error;
1043        }
1044
1045        if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1046                goto drop;
1047
1048        port = inet_sk(sk)->inet_sport;
1049
1050        vs = rcu_dereference_sk_user_data(sk);
1051        if (!vs)
1052                goto drop;
1053
1054        vs->rcv(vs, skb, vxh->vx_vni);
1055        return 0;
1056
1057drop:
1058        /* Consume bad packet */
1059        kfree_skb(skb);
1060        return 0;
1061
1062error:
1063        /* Return non vxlan pkt */
1064        return 1;
1065}
1066
1067static void vxlan_rcv(struct vxlan_sock *vs,
1068                      struct sk_buff *skb, __be32 vx_vni)
1069{
1070        struct iphdr *oip = NULL;
1071        struct ipv6hdr *oip6 = NULL;
1072        struct vxlan_dev *vxlan;
1073        struct pcpu_tstats *stats;
1074        union vxlan_addr saddr;
1075        __u32 vni;
1076        int err = 0;
1077        union vxlan_addr *remote_ip;
1078
1079        vni = ntohl(vx_vni) >> 8;
1080        /* Is this VNI defined? */
1081        vxlan = vxlan_vs_find_vni(vs, vni);
1082        if (!vxlan)
1083                goto drop;
1084
1085        remote_ip = &vxlan->default_dst.remote_ip;
1086        skb_reset_mac_header(skb);
1087        skb->protocol = eth_type_trans(skb, vxlan->dev);
1088
1089        /* Ignore packet loops (and multicast echo) */
1090        if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1091                goto drop;
1092
1093        /* Re-examine inner Ethernet packet */
1094        if (remote_ip->sa.sa_family == AF_INET) {
1095                oip = ip_hdr(skb);
1096                saddr.sin.sin_addr.s_addr = oip->saddr;
1097                saddr.sa.sa_family = AF_INET;
1098#if IS_ENABLED(CONFIG_IPV6)
1099        } else {
1100                oip6 = ipv6_hdr(skb);
1101                saddr.sin6.sin6_addr = oip6->saddr;
1102                saddr.sa.sa_family = AF_INET6;
1103#endif
1104        }
1105
1106        if ((vxlan->flags & VXLAN_F_LEARN) &&
1107            vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1108                goto drop;
1109
1110        skb_reset_network_header(skb);
1111
1112        /* If the NIC driver gave us an encapsulated packet with
1113         * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
1114         * leave the CHECKSUM_UNNECESSARY, the device checksummed it
1115         * for us. Otherwise force the upper layers to verify it.
1116         */
1117        if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation ||
1118            !(vxlan->dev->features & NETIF_F_RXCSUM))
1119                skb->ip_summed = CHECKSUM_NONE;
1120
1121        skb->encapsulation = 0;
1122
1123        if (oip6)
1124                err = IP6_ECN_decapsulate(oip6, skb);
1125        if (oip)
1126                err = IP_ECN_decapsulate(oip, skb);
1127
1128        if (unlikely(err)) {
1129                if (log_ecn_error) {
1130                        if (oip6)
1131                                net_info_ratelimited("non-ECT from %pI6\n",
1132                                                     &oip6->saddr);
1133                        if (oip)
1134                                net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1135                                                     &oip->saddr, oip->tos);
1136                }
1137                if (err > 1) {
1138                        ++vxlan->dev->stats.rx_frame_errors;
1139                        ++vxlan->dev->stats.rx_errors;
1140                        goto drop;
1141                }
1142        }
1143
1144        stats = this_cpu_ptr(vxlan->dev->tstats);
1145        u64_stats_update_begin(&stats->syncp);
1146        stats->rx_packets++;
1147        stats->rx_bytes += skb->len;
1148        u64_stats_update_end(&stats->syncp);
1149
1150        netif_rx(skb);
1151
1152        return;
1153drop:
1154        /* Consume bad packet */
1155        kfree_skb(skb);
1156}
1157
1158static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1159{
1160        struct vxlan_dev *vxlan = netdev_priv(dev);
1161        struct arphdr *parp;
1162        u8 *arpptr, *sha;
1163        __be32 sip, tip;
1164        struct neighbour *n;
1165
1166        if (dev->flags & IFF_NOARP)
1167                goto out;
1168
1169        if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1170                dev->stats.tx_dropped++;
1171                goto out;
1172        }
1173        parp = arp_hdr(skb);
1174
1175        if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1176             parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1177            parp->ar_pro != htons(ETH_P_IP) ||
1178            parp->ar_op != htons(ARPOP_REQUEST) ||
1179            parp->ar_hln != dev->addr_len ||
1180            parp->ar_pln != 4)
1181                goto out;
1182        arpptr = (u8 *)parp + sizeof(struct arphdr);
1183        sha = arpptr;
1184        arpptr += dev->addr_len;        /* sha */
1185        memcpy(&sip, arpptr, sizeof(sip));
1186        arpptr += sizeof(sip);
1187        arpptr += dev->addr_len;        /* tha */
1188        memcpy(&tip, arpptr, sizeof(tip));
1189
1190        if (ipv4_is_loopback(tip) ||
1191            ipv4_is_multicast(tip))
1192                goto out;
1193
1194        n = neigh_lookup(&arp_tbl, &tip, dev);
1195
1196        if (n) {
1197                struct vxlan_fdb *f;
1198                struct sk_buff  *reply;
1199
1200                if (!(n->nud_state & NUD_CONNECTED)) {
1201                        neigh_release(n);
1202                        goto out;
1203                }
1204
1205                f = vxlan_find_mac(vxlan, n->ha);
1206                if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1207                        /* bridge-local neighbor */
1208                        neigh_release(n);
1209                        goto out;
1210                }
1211
1212                reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1213                                n->ha, sha);
1214
1215                neigh_release(n);
1216
1217                skb_reset_mac_header(reply);
1218                __skb_pull(reply, skb_network_offset(reply));
1219                reply->ip_summed = CHECKSUM_UNNECESSARY;
1220                reply->pkt_type = PACKET_HOST;
1221
1222                if (netif_rx_ni(reply) == NET_RX_DROP)
1223                        dev->stats.rx_dropped++;
1224        } else if (vxlan->flags & VXLAN_F_L3MISS) {
1225                union vxlan_addr ipa = {
1226                        .sin.sin_addr.s_addr = tip,
1227                        .sa.sa_family = AF_INET,
1228                };
1229
1230                vxlan_ip_miss(dev, &ipa);
1231        }
1232out:
1233        consume_skb(skb);
1234        return NETDEV_TX_OK;
1235}
1236
1237#if IS_ENABLED(CONFIG_IPV6)
1238static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1239{
1240        struct vxlan_dev *vxlan = netdev_priv(dev);
1241        struct neighbour *n;
1242        union vxlan_addr ipa;
1243        const struct ipv6hdr *iphdr;
1244        const struct in6_addr *saddr, *daddr;
1245        struct nd_msg *msg;
1246        struct inet6_dev *in6_dev = NULL;
1247
1248        in6_dev = __in6_dev_get(dev);
1249        if (!in6_dev)
1250                goto out;
1251
1252        if (!pskb_may_pull(skb, skb->len))
1253                goto out;
1254
1255        iphdr = ipv6_hdr(skb);
1256        saddr = &iphdr->saddr;
1257        daddr = &iphdr->daddr;
1258
1259        if (ipv6_addr_loopback(daddr) ||
1260            ipv6_addr_is_multicast(daddr))
1261                goto out;
1262
1263        msg = (struct nd_msg *)skb_transport_header(skb);
1264        if (msg->icmph.icmp6_code != 0 ||
1265            msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1266                goto out;
1267
1268        n = neigh_lookup(ipv6_stub->nd_tbl, daddr, dev);
1269
1270        if (n) {
1271                struct vxlan_fdb *f;
1272
1273                if (!(n->nud_state & NUD_CONNECTED)) {
1274                        neigh_release(n);
1275                        goto out;
1276                }
1277
1278                f = vxlan_find_mac(vxlan, n->ha);
1279                if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1280                        /* bridge-local neighbor */
1281                        neigh_release(n);
1282                        goto out;
1283                }
1284
1285                ipv6_stub->ndisc_send_na(dev, n, saddr, &msg->target,
1286                                         !!in6_dev->cnf.forwarding,
1287                                         true, false, false);
1288                neigh_release(n);
1289        } else if (vxlan->flags & VXLAN_F_L3MISS) {
1290                ipa.sin6.sin6_addr = *daddr;
1291                ipa.sa.sa_family = AF_INET6;
1292                vxlan_ip_miss(dev, &ipa);
1293        }
1294
1295out:
1296        consume_skb(skb);
1297        return NETDEV_TX_OK;
1298}
1299#endif
1300
1301static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1302{
1303        struct vxlan_dev *vxlan = netdev_priv(dev);
1304        struct neighbour *n;
1305
1306        if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1307                return false;
1308
1309        n = NULL;
1310        switch (ntohs(eth_hdr(skb)->h_proto)) {
1311        case ETH_P_IP:
1312        {
1313                struct iphdr *pip;
1314
1315                if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1316                        return false;
1317                pip = ip_hdr(skb);
1318                n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1319                if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1320                        union vxlan_addr ipa = {
1321                                .sin.sin_addr.s_addr = pip->daddr,
1322                                .sa.sa_family = AF_INET,
1323                        };
1324
1325                        vxlan_ip_miss(dev, &ipa);
1326                        return false;
1327                }
1328
1329                break;
1330        }
1331#if IS_ENABLED(CONFIG_IPV6)
1332        case ETH_P_IPV6:
1333        {
1334                struct ipv6hdr *pip6;
1335
1336                if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1337                        return false;
1338                pip6 = ipv6_hdr(skb);
1339                n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1340                if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1341                        union vxlan_addr ipa = {
1342                                .sin6.sin6_addr = pip6->daddr,
1343                                .sa.sa_family = AF_INET6,
1344                        };
1345
1346                        vxlan_ip_miss(dev, &ipa);
1347                        return false;
1348                }
1349
1350                break;
1351        }
1352#endif
1353        default:
1354                return false;
1355        }
1356
1357        if (n) {
1358                bool diff;
1359
1360                diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1361                if (diff) {
1362                        memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1363                                dev->addr_len);
1364                        memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1365                }
1366                neigh_release(n);
1367                return diff;
1368        }
1369
1370        return false;
1371}
1372
1373static void vxlan_sock_put(struct sk_buff *skb)
1374{
1375        sock_put(skb->sk);
1376}
1377
1378/* On transmit, associate with the tunnel socket */
1379static void vxlan_set_owner(struct sock *sk, struct sk_buff *skb)
1380{
1381        skb_orphan(skb);
1382        sock_hold(sk);
1383        skb->sk = sk;
1384        skb->destructor = vxlan_sock_put;
1385}
1386
1387/* Compute source port for outgoing packet
1388 *   first choice to use L4 flow hash since it will spread
1389 *     better and maybe available from hardware
1390 *   secondary choice is to use jhash on the Ethernet header
1391 */
1392__be16 vxlan_src_port(__u16 port_min, __u16 port_max, struct sk_buff *skb)
1393{
1394        unsigned int range = (port_max - port_min) + 1;
1395        u32 hash;
1396
1397        hash = skb_get_rxhash(skb);
1398        if (!hash)
1399                hash = jhash(skb->data, 2 * ETH_ALEN,
1400                             (__force u32) skb->protocol);
1401
1402        return htons((((u64) hash * range) >> 32) + port_min);
1403}
1404EXPORT_SYMBOL_GPL(vxlan_src_port);
1405
1406static int handle_offloads(struct sk_buff *skb)
1407{
1408        if (skb_is_gso(skb)) {
1409                int err = skb_unclone(skb, GFP_ATOMIC);
1410                if (unlikely(err))
1411                        return err;
1412
1413                skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
1414        } else if (skb->ip_summed != CHECKSUM_PARTIAL)
1415                skb->ip_summed = CHECKSUM_NONE;
1416
1417        return 0;
1418}
1419
1420#if IS_ENABLED(CONFIG_IPV6)
1421static int vxlan6_xmit_skb(struct vxlan_sock *vs,
1422                           struct dst_entry *dst, struct sk_buff *skb,
1423                           struct net_device *dev, struct in6_addr *saddr,
1424                           struct in6_addr *daddr, __u8 prio, __u8 ttl,
1425                           __be16 src_port, __be16 dst_port, __be32 vni)
1426{
1427        struct ipv6hdr *ip6h;
1428        struct vxlanhdr *vxh;
1429        struct udphdr *uh;
1430        int min_headroom;
1431        int err;
1432
1433        if (!skb->encapsulation) {
1434                skb_reset_inner_headers(skb);
1435                skb->encapsulation = 1;
1436        }
1437
1438        skb_scrub_packet(skb, false);
1439
1440        min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1441                        + VXLAN_HLEN + sizeof(struct ipv6hdr)
1442                        + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1443
1444        /* Need space for new headers (invalidates iph ptr) */
1445        err = skb_cow_head(skb, min_headroom);
1446        if (unlikely(err))
1447                return err;
1448
1449        if (vlan_tx_tag_present(skb)) {
1450                if (WARN_ON(!__vlan_put_tag(skb,
1451                                            skb->vlan_proto,
1452                                            vlan_tx_tag_get(skb))))
1453                        return -ENOMEM;
1454
1455                skb->vlan_tci = 0;
1456        }
1457
1458        vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1459        vxh->vx_flags = htonl(VXLAN_FLAGS);
1460        vxh->vx_vni = vni;
1461
1462        __skb_push(skb, sizeof(*uh));
1463        skb_reset_transport_header(skb);
1464        uh = udp_hdr(skb);
1465
1466        uh->dest = dst_port;
1467        uh->source = src_port;
1468
1469        uh->len = htons(skb->len);
1470        uh->check = 0;
1471
1472        memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1473        IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1474                              IPSKB_REROUTED);
1475        skb_dst_set(skb, dst);
1476
1477        if (!skb_is_gso(skb) && !(dst->dev->features & NETIF_F_IPV6_CSUM)) {
1478                __wsum csum = skb_checksum(skb, 0, skb->len, 0);
1479                skb->ip_summed = CHECKSUM_UNNECESSARY;
1480                uh->check = csum_ipv6_magic(saddr, daddr, skb->len,
1481                                            IPPROTO_UDP, csum);
1482                if (uh->check == 0)
1483                        uh->check = CSUM_MANGLED_0;
1484        } else {
1485                skb->ip_summed = CHECKSUM_PARTIAL;
1486                skb->csum_start = skb_transport_header(skb) - skb->head;
1487                skb->csum_offset = offsetof(struct udphdr, check);
1488                uh->check = ~csum_ipv6_magic(saddr, daddr,
1489                                             skb->len, IPPROTO_UDP, 0);
1490        }
1491
1492        __skb_push(skb, sizeof(*ip6h));
1493        skb_reset_network_header(skb);
1494        ip6h              = ipv6_hdr(skb);
1495        ip6h->version     = 6;
1496        ip6h->priority    = prio;
1497        ip6h->flow_lbl[0] = 0;
1498        ip6h->flow_lbl[1] = 0;
1499        ip6h->flow_lbl[2] = 0;
1500        ip6h->payload_len = htons(skb->len);
1501        ip6h->nexthdr     = IPPROTO_UDP;
1502        ip6h->hop_limit   = ttl;
1503        ip6h->daddr       = *daddr;
1504        ip6h->saddr       = *saddr;
1505
1506        vxlan_set_owner(vs->sock->sk, skb);
1507
1508        err = handle_offloads(skb);
1509        if (err)
1510                return err;
1511
1512        ip6tunnel_xmit(skb, dev);
1513        return 0;
1514}
1515#endif
1516
1517int vxlan_xmit_skb(struct vxlan_sock *vs,
1518                   struct rtable *rt, struct sk_buff *skb,
1519                   __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
1520                   __be16 src_port, __be16 dst_port, __be32 vni)
1521{
1522        struct vxlanhdr *vxh;
1523        struct udphdr *uh;
1524        int min_headroom;
1525        int err;
1526
1527        if (!skb->encapsulation) {
1528                skb_reset_inner_headers(skb);
1529                skb->encapsulation = 1;
1530        }
1531
1532        min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1533                        + VXLAN_HLEN + sizeof(struct iphdr)
1534                        + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1535
1536        /* Need space for new headers (invalidates iph ptr) */
1537        err = skb_cow_head(skb, min_headroom);
1538        if (unlikely(err))
1539                return err;
1540
1541        if (vlan_tx_tag_present(skb)) {
1542                if (WARN_ON(!__vlan_put_tag(skb,
1543                                            skb->vlan_proto,
1544                                            vlan_tx_tag_get(skb))))
1545                        return -ENOMEM;
1546
1547                skb->vlan_tci = 0;
1548        }
1549
1550        vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1551        vxh->vx_flags = htonl(VXLAN_FLAGS);
1552        vxh->vx_vni = vni;
1553
1554        __skb_push(skb, sizeof(*uh));
1555        skb_reset_transport_header(skb);
1556        uh = udp_hdr(skb);
1557
1558        uh->dest = dst_port;
1559        uh->source = src_port;
1560
1561        uh->len = htons(skb->len);
1562        uh->check = 0;
1563
1564        vxlan_set_owner(vs->sock->sk, skb);
1565
1566        err = handle_offloads(skb);
1567        if (err)
1568                return err;
1569
1570        return iptunnel_xmit(rt, skb, src, dst, IPPROTO_UDP, tos, ttl, df,
1571                             false);
1572}
1573EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1574
1575/* Bypass encapsulation if the destination is local */
1576static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1577                               struct vxlan_dev *dst_vxlan)
1578{
1579        struct pcpu_tstats *tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1580        struct pcpu_tstats *rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1581        union vxlan_addr loopback;
1582        union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
1583
1584        skb->pkt_type = PACKET_HOST;
1585        skb->encapsulation = 0;
1586        skb->dev = dst_vxlan->dev;
1587        __skb_pull(skb, skb_network_offset(skb));
1588
1589        if (remote_ip->sa.sa_family == AF_INET) {
1590                loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1591                loopback.sa.sa_family =  AF_INET;
1592#if IS_ENABLED(CONFIG_IPV6)
1593        } else {
1594                loopback.sin6.sin6_addr = in6addr_loopback;
1595                loopback.sa.sa_family =  AF_INET6;
1596#endif
1597        }
1598
1599        if (dst_vxlan->flags & VXLAN_F_LEARN)
1600                vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
1601
1602        u64_stats_update_begin(&tx_stats->syncp);
1603        tx_stats->tx_packets++;
1604        tx_stats->tx_bytes += skb->len;
1605        u64_stats_update_end(&tx_stats->syncp);
1606
1607        if (netif_rx(skb) == NET_RX_SUCCESS) {
1608                u64_stats_update_begin(&rx_stats->syncp);
1609                rx_stats->rx_packets++;
1610                rx_stats->rx_bytes += skb->len;
1611                u64_stats_update_end(&rx_stats->syncp);
1612        } else {
1613                skb->dev->stats.rx_dropped++;
1614        }
1615}
1616
1617static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1618                           struct vxlan_rdst *rdst, bool did_rsc)
1619{
1620        struct vxlan_dev *vxlan = netdev_priv(dev);
1621        struct rtable *rt = NULL;
1622        const struct iphdr *old_iph;
1623        struct flowi4 fl4;
1624        union vxlan_addr *dst;
1625        __be16 src_port = 0, dst_port;
1626        u32 vni;
1627        __be16 df = 0;
1628        __u8 tos, ttl;
1629        int err;
1630
1631        dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
1632        vni = rdst->remote_vni;
1633        dst = &rdst->remote_ip;
1634
1635        if (vxlan_addr_any(dst)) {
1636                if (did_rsc) {
1637                        /* short-circuited back to local bridge */
1638                        vxlan_encap_bypass(skb, vxlan, vxlan);
1639                        return;
1640                }
1641                goto drop;
1642        }
1643
1644        old_iph = ip_hdr(skb);
1645
1646        ttl = vxlan->ttl;
1647        if (!ttl && vxlan_addr_multicast(dst))
1648                ttl = 1;
1649
1650        tos = vxlan->tos;
1651        if (tos == 1)
1652                tos = ip_tunnel_get_dsfield(old_iph, skb);
1653
1654        src_port = vxlan_src_port(vxlan->port_min, vxlan->port_max, skb);
1655
1656        if (dst->sa.sa_family == AF_INET) {
1657                memset(&fl4, 0, sizeof(fl4));
1658                fl4.flowi4_oif = rdst->remote_ifindex;
1659                fl4.flowi4_tos = RT_TOS(tos);
1660                fl4.daddr = dst->sin.sin_addr.s_addr;
1661                fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1662
1663                rt = ip_route_output_key(dev_net(dev), &fl4);
1664                if (IS_ERR(rt)) {
1665                        netdev_dbg(dev, "no route to %pI4\n",
1666                                   &dst->sin.sin_addr.s_addr);
1667                        dev->stats.tx_carrier_errors++;
1668                        goto tx_error;
1669                }
1670
1671                if (rt->dst.dev == dev) {
1672                        netdev_dbg(dev, "circular route to %pI4\n",
1673                                   &dst->sin.sin_addr.s_addr);
1674                        dev->stats.collisions++;
1675                        goto tx_error;
1676                }
1677
1678                /* Bypass encapsulation if the destination is local */
1679                if (rt->rt_flags & RTCF_LOCAL &&
1680                    !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1681                        struct vxlan_dev *dst_vxlan;
1682
1683                        ip_rt_put(rt);
1684                        dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1685                        if (!dst_vxlan)
1686                                goto tx_error;
1687                        vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1688                        return;
1689                }
1690
1691                tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1692                ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
1693
1694                err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
1695                                     fl4.saddr, dst->sin.sin_addr.s_addr,
1696                                     tos, ttl, df, src_port, dst_port,
1697                                     htonl(vni << 8));
1698
1699                if (err < 0)
1700                        goto rt_tx_error;
1701                iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1702#if IS_ENABLED(CONFIG_IPV6)
1703        } else {
1704                struct sock *sk = vxlan->vn_sock->sock->sk;
1705                struct dst_entry *ndst;
1706                struct flowi6 fl6;
1707                u32 flags;
1708
1709                memset(&fl6, 0, sizeof(fl6));
1710                fl6.flowi6_oif = rdst->remote_ifindex;
1711                fl6.daddr = dst->sin6.sin6_addr;
1712                fl6.saddr = vxlan->saddr.sin6.sin6_addr;
1713                fl6.flowi6_proto = IPPROTO_UDP;
1714
1715                if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1716                        netdev_dbg(dev, "no route to %pI6\n",
1717                                   &dst->sin6.sin6_addr);
1718                        dev->stats.tx_carrier_errors++;
1719                        goto tx_error;
1720                }
1721
1722                if (ndst->dev == dev) {
1723                        netdev_dbg(dev, "circular route to %pI6\n",
1724                                   &dst->sin6.sin6_addr);
1725                        dst_release(ndst);
1726                        dev->stats.collisions++;
1727                        goto tx_error;
1728                }
1729
1730                /* Bypass encapsulation if the destination is local */
1731                flags = ((struct rt6_info *)ndst)->rt6i_flags;
1732                if (flags & RTF_LOCAL &&
1733                    !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1734                        struct vxlan_dev *dst_vxlan;
1735
1736                        dst_release(ndst);
1737                        dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1738                        if (!dst_vxlan)
1739                                goto tx_error;
1740                        vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1741                        return;
1742                }
1743
1744                ttl = ttl ? : ip6_dst_hoplimit(ndst);
1745
1746                err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
1747                                      dev, &fl6.saddr, &fl6.daddr, 0, ttl,
1748                                      src_port, dst_port, htonl(vni << 8));
1749#endif
1750        }
1751
1752        return;
1753
1754drop:
1755        dev->stats.tx_dropped++;
1756        goto tx_free;
1757
1758rt_tx_error:
1759        ip_rt_put(rt);
1760tx_error:
1761        dev->stats.tx_errors++;
1762tx_free:
1763        dev_kfree_skb(skb);
1764}
1765
1766/* Transmit local packets over Vxlan
1767 *
1768 * Outer IP header inherits ECN and DF from inner header.
1769 * Outer UDP destination is the VXLAN assigned port.
1770 *           source port is based on hash of flow
1771 */
1772static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1773{
1774        struct vxlan_dev *vxlan = netdev_priv(dev);
1775        struct ethhdr *eth;
1776        bool did_rsc = false;
1777        struct vxlan_rdst *rdst;
1778        struct vxlan_fdb *f;
1779
1780        skb_reset_mac_header(skb);
1781        eth = eth_hdr(skb);
1782
1783        if ((vxlan->flags & VXLAN_F_PROXY)) {
1784                if (ntohs(eth->h_proto) == ETH_P_ARP)
1785                        return arp_reduce(dev, skb);
1786#if IS_ENABLED(CONFIG_IPV6)
1787                else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
1788                         skb->len >= sizeof(struct ipv6hdr) + sizeof(struct nd_msg) &&
1789                         ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
1790                                struct nd_msg *msg;
1791
1792                                msg = (struct nd_msg *)skb_transport_header(skb);
1793                                if (msg->icmph.icmp6_code == 0 &&
1794                                    msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
1795                                        return neigh_reduce(dev, skb);
1796                }
1797#endif
1798        }
1799
1800        f = vxlan_find_mac(vxlan, eth->h_dest);
1801        did_rsc = false;
1802
1803        if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
1804            (ntohs(eth->h_proto) == ETH_P_IP ||
1805             ntohs(eth->h_proto) == ETH_P_IPV6)) {
1806                did_rsc = route_shortcircuit(dev, skb);
1807                if (did_rsc)
1808                        f = vxlan_find_mac(vxlan, eth->h_dest);
1809        }
1810
1811        if (f == NULL) {
1812                f = vxlan_find_mac(vxlan, all_zeros_mac);
1813                if (f == NULL) {
1814                        if ((vxlan->flags & VXLAN_F_L2MISS) &&
1815                            !is_multicast_ether_addr(eth->h_dest))
1816                                vxlan_fdb_miss(vxlan, eth->h_dest);
1817
1818                        dev->stats.tx_dropped++;
1819                        dev_kfree_skb(skb);
1820                        return NETDEV_TX_OK;
1821                }
1822        }
1823
1824        list_for_each_entry_rcu(rdst, &f->remotes, list) {
1825                struct sk_buff *skb1;
1826
1827                skb1 = skb_clone(skb, GFP_ATOMIC);
1828                if (skb1)
1829                        vxlan_xmit_one(skb1, dev, rdst, did_rsc);
1830        }
1831
1832        dev_kfree_skb(skb);
1833        return NETDEV_TX_OK;
1834}
1835
1836/* Walk the forwarding table and purge stale entries */
1837static void vxlan_cleanup(unsigned long arg)
1838{
1839        struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1840        unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1841        unsigned int h;
1842
1843        if (!netif_running(vxlan->dev))
1844                return;
1845
1846        spin_lock_bh(&vxlan->hash_lock);
1847        for (h = 0; h < FDB_HASH_SIZE; ++h) {
1848                struct hlist_node *p, *n;
1849                hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1850                        struct vxlan_fdb *f
1851                                = container_of(p, struct vxlan_fdb, hlist);
1852                        unsigned long timeout;
1853
1854                        if (f->state & NUD_PERMANENT)
1855                                continue;
1856
1857                        timeout = f->used + vxlan->age_interval * HZ;
1858                        if (time_before_eq(timeout, jiffies)) {
1859                                netdev_dbg(vxlan->dev,
1860                                           "garbage collect %pM\n",
1861                                           f->eth_addr);
1862                                f->state = NUD_STALE;
1863                                vxlan_fdb_destroy(vxlan, f);
1864                        } else if (time_before(timeout, next_timer))
1865                                next_timer = timeout;
1866                }
1867        }
1868        spin_unlock_bh(&vxlan->hash_lock);
1869
1870        mod_timer(&vxlan->age_timer, next_timer);
1871}
1872
1873static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
1874{
1875        __u32 vni = vxlan->default_dst.remote_vni;
1876
1877        vxlan->vn_sock = vs;
1878        hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
1879}
1880
1881/* Setup stats when device is created */
1882static int vxlan_init(struct net_device *dev)
1883{
1884        struct vxlan_dev *vxlan = netdev_priv(dev);
1885        struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1886        struct vxlan_sock *vs;
1887
1888        dev->tstats = alloc_percpu(struct pcpu_tstats);
1889        if (!dev->tstats)
1890                return -ENOMEM;
1891
1892        spin_lock(&vn->sock_lock);
1893        vs = vxlan_find_sock(dev_net(dev), vxlan->dst_port);
1894        if (vs) {
1895                /* If we have a socket with same port already, reuse it */
1896                atomic_inc(&vs->refcnt);
1897                vxlan_vs_add_dev(vs, vxlan);
1898        } else {
1899                /* otherwise make new socket outside of RTNL */
1900                dev_hold(dev);
1901                queue_work(vxlan_wq, &vxlan->sock_work);
1902        }
1903        spin_unlock(&vn->sock_lock);
1904
1905        return 0;
1906}
1907
1908static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
1909{
1910        struct vxlan_fdb *f;
1911
1912        spin_lock_bh(&vxlan->hash_lock);
1913        f = __vxlan_find_mac(vxlan, all_zeros_mac);
1914        if (f)
1915                vxlan_fdb_destroy(vxlan, f);
1916        spin_unlock_bh(&vxlan->hash_lock);
1917}
1918
1919static void vxlan_uninit(struct net_device *dev)
1920{
1921        struct vxlan_dev *vxlan = netdev_priv(dev);
1922        struct vxlan_sock *vs = vxlan->vn_sock;
1923
1924        vxlan_fdb_delete_default(vxlan);
1925
1926        if (vs)
1927                vxlan_sock_release(vs);
1928        free_percpu(dev->tstats);
1929}
1930
1931/* Start ageing timer and join group when device is brought up */
1932static int vxlan_open(struct net_device *dev)
1933{
1934        struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1935        struct vxlan_dev *vxlan = netdev_priv(dev);
1936        struct vxlan_sock *vs = vxlan->vn_sock;
1937
1938        /* socket hasn't been created */
1939        if (!vs)
1940                return -ENOTCONN;
1941
1942        if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
1943            vxlan_group_used(vn, &vxlan->default_dst.remote_ip)) {
1944                vxlan_sock_hold(vs);
1945                dev_hold(dev);
1946                queue_work(vxlan_wq, &vxlan->igmp_join);
1947        }
1948
1949        if (vxlan->age_interval)
1950                mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1951
1952        return 0;
1953}
1954
1955/* Purge the forwarding table */
1956static void vxlan_flush(struct vxlan_dev *vxlan)
1957{
1958        unsigned int h;
1959
1960        spin_lock_bh(&vxlan->hash_lock);
1961        for (h = 0; h < FDB_HASH_SIZE; ++h) {
1962                struct hlist_node *p, *n;
1963                hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1964                        struct vxlan_fdb *f
1965                                = container_of(p, struct vxlan_fdb, hlist);
1966                        /* the all_zeros_mac entry is deleted at vxlan_uninit */
1967                        if (!is_zero_ether_addr(f->eth_addr))
1968                                vxlan_fdb_destroy(vxlan, f);
1969                }
1970        }
1971        spin_unlock_bh(&vxlan->hash_lock);
1972}
1973
1974/* Cleanup timer and forwarding table on shutdown */
1975static int vxlan_stop(struct net_device *dev)
1976{
1977        struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1978        struct vxlan_dev *vxlan = netdev_priv(dev);
1979        struct vxlan_sock *vs = vxlan->vn_sock;
1980
1981        if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
1982            ! vxlan_group_used(vn, &vxlan->default_dst.remote_ip)) {
1983                vxlan_sock_hold(vs);
1984                dev_hold(dev);
1985                queue_work(vxlan_wq, &vxlan->igmp_leave);
1986        }
1987
1988        del_timer_sync(&vxlan->age_timer);
1989
1990        vxlan_flush(vxlan);
1991
1992        return 0;
1993}
1994
1995/* Stub, nothing needs to be done. */
1996static void vxlan_set_multicast_list(struct net_device *dev)
1997{
1998}
1999
2000static const struct net_device_ops vxlan_netdev_ops = {
2001        .ndo_init               = vxlan_init,
2002        .ndo_uninit             = vxlan_uninit,
2003        .ndo_open               = vxlan_open,
2004        .ndo_stop               = vxlan_stop,
2005        .ndo_start_xmit         = vxlan_xmit,
2006        .ndo_get_stats64        = ip_tunnel_get_stats64,
2007        .ndo_set_rx_mode        = vxlan_set_multicast_list,
2008        .ndo_change_mtu         = eth_change_mtu,
2009        .ndo_validate_addr      = eth_validate_addr,
2010        .ndo_set_mac_address    = eth_mac_addr,
2011        .ndo_fdb_add            = vxlan_fdb_add,
2012        .ndo_fdb_del            = vxlan_fdb_delete,
2013        .ndo_fdb_dump           = vxlan_fdb_dump,
2014};
2015
2016/* Info for udev, that this is a virtual tunnel endpoint */
2017static struct device_type vxlan_type = {
2018        .name = "vxlan",
2019};
2020
2021/* Calls the ndo_add_vxlan_port of the caller in order to
2022 * supply the listening VXLAN udp ports. Callers are expected
2023 * to implement the ndo_add_vxlan_port.
2024 */
2025void vxlan_get_rx_port(struct net_device *dev)
2026{
2027        struct vxlan_sock *vs;
2028        struct net *net = dev_net(dev);
2029        struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2030        sa_family_t sa_family;
2031        __be16 port;
2032        unsigned int i;
2033
2034        spin_lock(&vn->sock_lock);
2035        for (i = 0; i < PORT_HASH_SIZE; ++i) {
2036                hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2037                        port = inet_sk(vs->sock->sk)->inet_sport;
2038                        sa_family = vs->sock->sk->sk_family;
2039                        dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2040                                                            port);
2041                }
2042        }
2043        spin_unlock(&vn->sock_lock);
2044}
2045EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2046
2047/* Initialize the device structure. */
2048static void vxlan_setup(struct net_device *dev)
2049{
2050        struct vxlan_dev *vxlan = netdev_priv(dev);
2051        unsigned int h;
2052        int low, high;
2053
2054        eth_hw_addr_random(dev);
2055        ether_setup(dev);
2056        if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2057                dev->hard_header_len = ETH_HLEN + VXLAN6_HEADROOM;
2058        else
2059                dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
2060
2061        dev->netdev_ops = &vxlan_netdev_ops;
2062        dev->destructor = free_netdev;
2063        SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2064
2065        dev->tx_queue_len = 0;
2066        dev->features   |= NETIF_F_LLTX;
2067        dev->features   |= NETIF_F_NETNS_LOCAL;
2068        dev->features   |= NETIF_F_SG | NETIF_F_HW_CSUM;
2069        dev->features   |= NETIF_F_RXCSUM;
2070        dev->features   |= NETIF_F_GSO_SOFTWARE;
2071
2072        dev->vlan_features = dev->features;
2073        dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2074        dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2075        dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2076        dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2077        dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
2078        dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2079
2080        INIT_LIST_HEAD(&vxlan->next);
2081        spin_lock_init(&vxlan->hash_lock);
2082        INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2083        INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
2084        INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
2085
2086        init_timer_deferrable(&vxlan->age_timer);
2087        vxlan->age_timer.function = vxlan_cleanup;
2088        vxlan->age_timer.data = (unsigned long) vxlan;
2089
2090        inet_get_local_port_range(&low, &high);
2091        vxlan->port_min = low;
2092        vxlan->port_max = high;
2093        vxlan->dst_port = htons(vxlan_port);
2094
2095        vxlan->dev = dev;
2096
2097        for (h = 0; h < FDB_HASH_SIZE; ++h)
2098                INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2099}
2100
2101static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2102        [IFLA_VXLAN_ID]         = { .type = NLA_U32 },
2103        [IFLA_VXLAN_GROUP]      = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2104        [IFLA_VXLAN_GROUP6]     = { .len = sizeof(struct in6_addr) },
2105        [IFLA_VXLAN_LINK]       = { .type = NLA_U32 },
2106        [IFLA_VXLAN_LOCAL]      = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2107        [IFLA_VXLAN_LOCAL6]     = { .len = sizeof(struct in6_addr) },
2108        [IFLA_VXLAN_TOS]        = { .type = NLA_U8 },
2109        [IFLA_VXLAN_TTL]        = { .type = NLA_U8 },
2110        [IFLA_VXLAN_LEARNING]   = { .type = NLA_U8 },
2111        [IFLA_VXLAN_AGEING]     = { .type = NLA_U32 },
2112        [IFLA_VXLAN_LIMIT]      = { .type = NLA_U32 },
2113        [IFLA_VXLAN_PORT_RANGE] = { .len  = sizeof(struct ifla_vxlan_port_range) },
2114        [IFLA_VXLAN_PROXY]      = { .type = NLA_U8 },
2115        [IFLA_VXLAN_RSC]        = { .type = NLA_U8 },
2116        [IFLA_VXLAN_L2MISS]     = { .type = NLA_U8 },
2117        [IFLA_VXLAN_L3MISS]     = { .type = NLA_U8 },
2118        [IFLA_VXLAN_PORT]       = { .type = NLA_U16 },
2119};
2120
2121static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2122{
2123        if (tb[IFLA_ADDRESS]) {
2124                if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2125                        pr_debug("invalid link address (not ethernet)\n");
2126                        return -EINVAL;
2127                }
2128
2129                if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2130                        pr_debug("invalid all zero ethernet address\n");
2131                        return -EADDRNOTAVAIL;
2132                }
2133        }
2134
2135        if (!data)
2136                return -EINVAL;
2137
2138        if (data[IFLA_VXLAN_ID]) {
2139                __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2140                if (id >= VXLAN_VID_MASK)
2141                        return -ERANGE;
2142        }
2143
2144        if (data[IFLA_VXLAN_PORT_RANGE]) {
2145                const struct ifla_vxlan_port_range *p
2146                        = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2147
2148                if (ntohs(p->high) < ntohs(p->low)) {
2149                        pr_debug("port range %u .. %u not valid\n",
2150                                 ntohs(p->low), ntohs(p->high));
2151                        return -EINVAL;
2152                }
2153        }
2154
2155        return 0;
2156}
2157
2158static void vxlan_get_drvinfo(struct net_device *netdev,
2159                              struct ethtool_drvinfo *drvinfo)
2160{
2161        strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2162        strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2163}
2164
2165static const struct ethtool_ops vxlan_ethtool_ops = {
2166        .get_drvinfo    = vxlan_get_drvinfo,
2167        .get_link       = ethtool_op_get_link,
2168};
2169
2170static void vxlan_del_work(struct work_struct *work)
2171{
2172        struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2173
2174        sk_release_kernel(vs->sock->sk);
2175        kfree_rcu(vs, rcu);
2176}
2177
2178#if IS_ENABLED(CONFIG_IPV6)
2179/* Create UDP socket for encapsulation receive. AF_INET6 socket
2180 * could be used for both IPv4 and IPv6 communications, but
2181 * users may set bindv6only=1.
2182 */
2183static int create_v6_sock(struct net *net, __be16 port, struct socket **psock)
2184{
2185        struct sock *sk;
2186        struct socket *sock;
2187        struct sockaddr_in6 vxlan_addr = {
2188                .sin6_family = AF_INET6,
2189                .sin6_port = port,
2190        };
2191        int rc, val = 1;
2192
2193        rc = sock_create_kern(AF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
2194        if (rc < 0) {
2195                pr_debug("UDPv6 socket create failed\n");
2196                return rc;
2197        }
2198
2199        /* Put in proper namespace */
2200        sk = sock->sk;
2201        sk_change_net(sk, net);
2202
2203        kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
2204                          (char *)&val, sizeof(val));
2205        rc = kernel_bind(sock, (struct sockaddr *)&vxlan_addr,
2206                         sizeof(struct sockaddr_in6));
2207        if (rc < 0) {
2208                pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2209                         &vxlan_addr.sin6_addr, ntohs(vxlan_addr.sin6_port), rc);
2210                sk_release_kernel(sk);
2211                return rc;
2212        }
2213        /* At this point, IPv6 module should have been loaded in
2214         * sock_create_kern().
2215         */
2216        BUG_ON(!ipv6_stub);
2217
2218        *psock = sock;
2219        /* Disable multicast loopback */
2220        inet_sk(sk)->mc_loop = 0;
2221        return 0;
2222}
2223
2224#else
2225
2226static int create_v6_sock(struct net *net, __be16 port, struct socket **psock)
2227{
2228                return -EPFNOSUPPORT;
2229}
2230#endif
2231
2232static int create_v4_sock(struct net *net, __be16 port, struct socket **psock)
2233{
2234        struct sock *sk;
2235        struct socket *sock;
2236        struct sockaddr_in vxlan_addr = {
2237                .sin_family = AF_INET,
2238                .sin_addr.s_addr = htonl(INADDR_ANY),
2239                .sin_port = port,
2240        };
2241        int rc;
2242
2243        /* Create UDP socket for encapsulation receive. */
2244        rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
2245        if (rc < 0) {
2246                pr_debug("UDP socket create failed\n");
2247                return rc;
2248        }
2249
2250        /* Put in proper namespace */
2251        sk = sock->sk;
2252        sk_change_net(sk, net);
2253
2254        rc = kernel_bind(sock, (struct sockaddr *) &vxlan_addr,
2255                         sizeof(vxlan_addr));
2256        if (rc < 0) {
2257                pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2258                         &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
2259                sk_release_kernel(sk);
2260                return rc;
2261        }
2262
2263        *psock = sock;
2264        /* Disable multicast loopback */
2265        inet_sk(sk)->mc_loop = 0;
2266        return 0;
2267}
2268
2269/* Create new listen socket if needed */
2270static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
2271                                              vxlan_rcv_t *rcv, void *data, bool ipv6)
2272{
2273        struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2274        struct vxlan_sock *vs;
2275        struct socket *sock;
2276        struct sock *sk;
2277        int rc = 0;
2278        unsigned int h;
2279
2280        vs = kmalloc(sizeof(*vs), GFP_KERNEL);
2281        if (!vs)
2282                return ERR_PTR(-ENOMEM);
2283
2284        for (h = 0; h < VNI_HASH_SIZE; ++h)
2285                INIT_HLIST_HEAD(&vs->vni_list[h]);
2286
2287        INIT_WORK(&vs->del_work, vxlan_del_work);
2288
2289        if (ipv6)
2290                rc = create_v6_sock(net, port, &sock);
2291        else
2292                rc = create_v4_sock(net, port, &sock);
2293        if (rc < 0) {
2294                kfree(vs);
2295                return ERR_PTR(rc);
2296        }
2297
2298        vs->sock = sock;
2299        sk = sock->sk;
2300        atomic_set(&vs->refcnt, 1);
2301        vs->rcv = rcv;
2302        vs->data = data;
2303        rcu_assign_sk_user_data(vs->sock->sk, vs);
2304
2305        spin_lock(&vn->sock_lock);
2306        hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2307        vxlan_notify_add_rx_port(sk);
2308        spin_unlock(&vn->sock_lock);
2309
2310        /* Mark socket as an encapsulation socket. */
2311        udp_sk(sk)->encap_type = 1;
2312        udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
2313#if IS_ENABLED(CONFIG_IPV6)
2314        if (ipv6)
2315                ipv6_stub->udpv6_encap_enable();
2316        else
2317#endif
2318                udp_encap_enable();
2319
2320        return vs;
2321}
2322
2323struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2324                                  vxlan_rcv_t *rcv, void *data,
2325                                  bool no_share, bool ipv6)
2326{
2327        struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2328        struct vxlan_sock *vs;
2329
2330        vs = vxlan_socket_create(net, port, rcv, data, ipv6);
2331        if (!IS_ERR(vs))
2332                return vs;
2333
2334        if (no_share)   /* Return error if sharing is not allowed. */
2335                return vs;
2336
2337        spin_lock(&vn->sock_lock);
2338        vs = vxlan_find_sock(net, port);
2339        if (vs) {
2340                if (vs->rcv == rcv)
2341                        atomic_inc(&vs->refcnt);
2342                else
2343                        vs = ERR_PTR(-EBUSY);
2344        }
2345        spin_unlock(&vn->sock_lock);
2346
2347        if (!vs)
2348                vs = ERR_PTR(-EINVAL);
2349
2350        return vs;
2351}
2352EXPORT_SYMBOL_GPL(vxlan_sock_add);
2353
2354/* Scheduled at device creation to bind to a socket */
2355static void vxlan_sock_work(struct work_struct *work)
2356{
2357        struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
2358        struct net *net = dev_net(vxlan->dev);
2359        struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2360        __be16 port = vxlan->dst_port;
2361        struct vxlan_sock *nvs;
2362
2363        nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags & VXLAN_F_IPV6);
2364        spin_lock(&vn->sock_lock);
2365        if (!IS_ERR(nvs))
2366                vxlan_vs_add_dev(nvs, vxlan);
2367        spin_unlock(&vn->sock_lock);
2368
2369        dev_put(vxlan->dev);
2370}
2371
2372static int vxlan_newlink(struct net *net, struct net_device *dev,
2373                         struct nlattr *tb[], struct nlattr *data[])
2374{
2375        struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2376        struct vxlan_dev *vxlan = netdev_priv(dev);
2377        struct vxlan_rdst *dst = &vxlan->default_dst;
2378        __u32 vni;
2379        int err;
2380        bool use_ipv6 = false;
2381
2382        if (!data[IFLA_VXLAN_ID])
2383                return -EINVAL;
2384
2385        vni = nla_get_u32(data[IFLA_VXLAN_ID]);
2386        dst->remote_vni = vni;
2387
2388        if (data[IFLA_VXLAN_GROUP]) {
2389                dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
2390                dst->remote_ip.sa.sa_family = AF_INET;
2391        } else if (data[IFLA_VXLAN_GROUP6]) {
2392                if (!IS_ENABLED(CONFIG_IPV6))
2393                        return -EPFNOSUPPORT;
2394
2395                nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2396                           sizeof(struct in6_addr));
2397                dst->remote_ip.sa.sa_family = AF_INET6;
2398                use_ipv6 = true;
2399        }
2400
2401        if (data[IFLA_VXLAN_LOCAL]) {
2402                vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2403                vxlan->saddr.sa.sa_family = AF_INET;
2404        } else if (data[IFLA_VXLAN_LOCAL6]) {
2405                if (!IS_ENABLED(CONFIG_IPV6))
2406                        return -EPFNOSUPPORT;
2407
2408                /* TODO: respect scope id */
2409                nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2410                           sizeof(struct in6_addr));
2411                vxlan->saddr.sa.sa_family = AF_INET6;
2412                use_ipv6 = true;
2413        }
2414
2415        if (data[IFLA_VXLAN_LINK] &&
2416            (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
2417                struct net_device *lowerdev
2418                         = __dev_get_by_index(net, dst->remote_ifindex);
2419
2420                if (!lowerdev) {
2421                        pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
2422                        return -ENODEV;
2423                }
2424
2425#if IS_ENABLED(CONFIG_IPV6)
2426                if (use_ipv6) {
2427                        struct inet6_dev *idev = __in6_dev_get(lowerdev);
2428                        if (idev && idev->cnf.disable_ipv6) {
2429                                pr_info("IPv6 is disabled via sysctl\n");
2430                                return -EPERM;
2431                        }
2432                        vxlan->flags |= VXLAN_F_IPV6;
2433                }
2434#endif
2435
2436                if (!tb[IFLA_MTU])
2437                        dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2438
2439                /* update header length based on lower device */
2440                dev->hard_header_len = lowerdev->hard_header_len +
2441                                       (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2442        }
2443
2444        if (data[IFLA_VXLAN_TOS])
2445                vxlan->tos  = nla_get_u8(data[IFLA_VXLAN_TOS]);
2446
2447        if (data[IFLA_VXLAN_TTL])
2448                vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2449
2450        if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
2451                vxlan->flags |= VXLAN_F_LEARN;
2452
2453        if (data[IFLA_VXLAN_AGEING])
2454                vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2455        else
2456                vxlan->age_interval = FDB_AGE_DEFAULT;
2457
2458        if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2459                vxlan->flags |= VXLAN_F_PROXY;
2460
2461        if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2462                vxlan->flags |= VXLAN_F_RSC;
2463
2464        if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2465                vxlan->flags |= VXLAN_F_L2MISS;
2466
2467        if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2468                vxlan->flags |= VXLAN_F_L3MISS;
2469
2470        if (data[IFLA_VXLAN_LIMIT])
2471                vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2472
2473        if (data[IFLA_VXLAN_PORT_RANGE]) {
2474                const struct ifla_vxlan_port_range *p
2475                        = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2476                vxlan->port_min = ntohs(p->low);
2477                vxlan->port_max = ntohs(p->high);
2478        }
2479
2480        if (data[IFLA_VXLAN_PORT])
2481                vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2482
2483        if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
2484                pr_info("duplicate VNI %u\n", vni);
2485                return -EEXIST;
2486        }
2487
2488        SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
2489
2490        /* create an fdb entry for a valid default destination */
2491        if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2492                err = vxlan_fdb_create(vxlan, all_zeros_mac,
2493                                       &vxlan->default_dst.remote_ip,
2494                                       NUD_REACHABLE|NUD_PERMANENT,
2495                                       NLM_F_EXCL|NLM_F_CREATE,
2496                                       vxlan->dst_port,
2497                                       vxlan->default_dst.remote_vni,
2498                                       vxlan->default_dst.remote_ifindex,
2499                                       NTF_SELF);
2500                if (err)
2501                        return err;
2502        }
2503
2504        err = register_netdevice(dev);
2505        if (err) {
2506                vxlan_fdb_delete_default(vxlan);
2507                return err;
2508        }
2509
2510        list_add(&vxlan->next, &vn->vxlan_list);
2511
2512        return 0;
2513}
2514
2515static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2516{
2517        struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
2518        struct vxlan_dev *vxlan = netdev_priv(dev);
2519
2520        spin_lock(&vn->sock_lock);
2521        if (!hlist_unhashed(&vxlan->hlist))
2522                hlist_del_rcu(&vxlan->hlist);
2523        spin_unlock(&vn->sock_lock);
2524
2525        list_del(&vxlan->next);
2526        unregister_netdevice_queue(dev, head);
2527}
2528
2529static size_t vxlan_get_size(const struct net_device *dev)
2530{
2531
2532        return nla_total_size(sizeof(__u32)) +  /* IFLA_VXLAN_ID */
2533                nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
2534                nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
2535                nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
2536                nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_TTL */
2537                nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_TOS */
2538                nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_LEARNING */
2539                nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_PROXY */
2540                nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_RSC */
2541                nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_L2MISS */
2542                nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_L3MISS */
2543                nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2544                nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
2545                nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
2546                nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */
2547                0;
2548}
2549
2550static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2551{
2552        const struct vxlan_dev *vxlan = netdev_priv(dev);
2553        const struct vxlan_rdst *dst = &vxlan->default_dst;
2554        struct ifla_vxlan_port_range ports = {
2555                .low =  htons(vxlan->port_min),
2556                .high = htons(vxlan->port_max),
2557        };
2558
2559        if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
2560                goto nla_put_failure;
2561
2562        if (!vxlan_addr_any(&dst->remote_ip)) {
2563                if (dst->remote_ip.sa.sa_family == AF_INET) {
2564                        if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2565                                         dst->remote_ip.sin.sin_addr.s_addr))
2566                                goto nla_put_failure;
2567#if IS_ENABLED(CONFIG_IPV6)
2568                } else {
2569                        if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2570                                    &dst->remote_ip.sin6.sin6_addr))
2571                                goto nla_put_failure;
2572#endif
2573                }
2574        }
2575
2576        if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
2577                goto nla_put_failure;
2578
2579        if (!vxlan_addr_any(&vxlan->saddr)) {
2580                if (vxlan->saddr.sa.sa_family == AF_INET) {
2581                        if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2582                                         vxlan->saddr.sin.sin_addr.s_addr))
2583                                goto nla_put_failure;
2584#if IS_ENABLED(CONFIG_IPV6)
2585                } else {
2586                        if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2587                                    &vxlan->saddr.sin6.sin6_addr))
2588                                goto nla_put_failure;
2589#endif
2590                }
2591        }
2592
2593        if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2594            nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
2595            nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2596                        !!(vxlan->flags & VXLAN_F_LEARN)) ||
2597            nla_put_u8(skb, IFLA_VXLAN_PROXY,
2598                        !!(vxlan->flags & VXLAN_F_PROXY)) ||
2599            nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2600            nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2601                        !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2602            nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2603                        !!(vxlan->flags & VXLAN_F_L3MISS)) ||
2604            nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
2605            nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
2606            nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port))
2607                goto nla_put_failure;
2608
2609        if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2610                goto nla_put_failure;
2611
2612        return 0;
2613
2614nla_put_failure:
2615        return -EMSGSIZE;
2616}
2617
2618static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2619        .kind           = "vxlan",
2620        .maxtype        = IFLA_VXLAN_MAX,
2621        .policy         = vxlan_policy,
2622        .priv_size      = sizeof(struct vxlan_dev),
2623        .setup          = vxlan_setup,
2624        .validate       = vxlan_validate,
2625        .newlink        = vxlan_newlink,
2626        .dellink        = vxlan_dellink,
2627        .get_size       = vxlan_get_size,
2628        .fill_info      = vxlan_fill_info,
2629};
2630
2631static __net_init int vxlan_init_net(struct net *net)
2632{
2633        struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2634        unsigned int h;
2635
2636        INIT_LIST_HEAD(&vn->vxlan_list);
2637        spin_lock_init(&vn->sock_lock);
2638
2639        for (h = 0; h < PORT_HASH_SIZE; ++h)
2640                INIT_HLIST_HEAD(&vn->sock_list[h]);
2641
2642        return 0;
2643}
2644
2645static __net_exit void vxlan_exit_net(struct net *net)
2646{
2647        struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2648        struct vxlan_dev *vxlan;
2649        LIST_HEAD(list);
2650
2651        rtnl_lock();
2652        list_for_each_entry(vxlan, &vn->vxlan_list, next)
2653                unregister_netdevice_queue(vxlan->dev, &list);
2654        unregister_netdevice_many(&list);
2655        rtnl_unlock();
2656}
2657
2658static struct pernet_operations vxlan_net_ops = {
2659        .init = vxlan_init_net,
2660        .exit = vxlan_exit_net,
2661        .id   = &vxlan_net_id,
2662        .size = sizeof(struct vxlan_net),
2663};
2664
2665static int __init vxlan_init_module(void)
2666{
2667        int rc;
2668
2669        vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2670        if (!vxlan_wq)
2671                return -ENOMEM;
2672
2673        get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2674
2675        rc = register_pernet_device(&vxlan_net_ops);
2676        if (rc)
2677                goto out1;
2678
2679        rc = rtnl_link_register(&vxlan_link_ops);
2680        if (rc)
2681                goto out2;
2682
2683        return 0;
2684
2685out2:
2686        unregister_pernet_device(&vxlan_net_ops);
2687out1:
2688        destroy_workqueue(vxlan_wq);
2689        return rc;
2690}
2691late_initcall(vxlan_init_module);
2692
2693static void __exit vxlan_cleanup_module(void)
2694{
2695        rtnl_link_unregister(&vxlan_link_ops);
2696        destroy_workqueue(vxlan_wq);
2697        unregister_pernet_device(&vxlan_net_ops);
2698        rcu_barrier();
2699}
2700module_exit(vxlan_cleanup_module);
2701
2702MODULE_LICENSE("GPL");
2703MODULE_VERSION(VXLAN_VERSION);
2704MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
2705MODULE_ALIAS_RTNL_LINK("vxlan");
2706
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.