Version:  2.0.40 2.2.26 2.4.37 3.0 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 3.13 3.14 3.15 3.16

Linux/drivers/net/vxlan.c

  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 #include <net/protocol.h>
 44 #if IS_ENABLED(CONFIG_IPV6)
 45 #include <net/ipv6.h>
 46 #include <net/addrconf.h>
 47 #include <net/ip6_tunnel.h>
 48 #include <net/ip6_checksum.h>
 49 #endif
 50 
 51 #define VXLAN_VERSION   "0.1"
 52 
 53 #define PORT_HASH_BITS  8
 54 #define PORT_HASH_SIZE  (1<<PORT_HASH_BITS)
 55 #define VNI_HASH_BITS   10
 56 #define VNI_HASH_SIZE   (1<<VNI_HASH_BITS)
 57 #define FDB_HASH_BITS   8
 58 #define FDB_HASH_SIZE   (1<<FDB_HASH_BITS)
 59 #define FDB_AGE_DEFAULT 300 /* 5 min */
 60 #define FDB_AGE_INTERVAL (10 * HZ)      /* rescan interval */
 61 
 62 #define VXLAN_N_VID     (1u << 24)
 63 #define VXLAN_VID_MASK  (VXLAN_N_VID - 1)
 64 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
 65 
 66 #define VXLAN_FLAGS 0x08000000  /* struct vxlanhdr.vx_flags required value. */
 67 
 68 /* VXLAN protocol header */
 69 struct vxlanhdr {
 70         __be32 vx_flags;
 71         __be32 vx_vni;
 72 };
 73 
 74 /* UDP port for VXLAN traffic.
 75  * The IANA assigned port is 4789, but the Linux default is 8472
 76  * for compatibility with early adopters.
 77  */
 78 static unsigned short vxlan_port __read_mostly = 8472;
 79 module_param_named(udp_port, vxlan_port, ushort, 0444);
 80 MODULE_PARM_DESC(udp_port, "Destination UDP port");
 81 
 82 static bool log_ecn_error = true;
 83 module_param(log_ecn_error, bool, 0644);
 84 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
 85 
 86 static int vxlan_net_id;
 87 
 88 static const u8 all_zeros_mac[ETH_ALEN];
 89 
 90 /* per-network namespace private data for this module */
 91 struct vxlan_net {
 92         struct list_head  vxlan_list;
 93         struct hlist_head sock_list[PORT_HASH_SIZE];
 94         spinlock_t        sock_lock;
 95 };
 96 
 97 union vxlan_addr {
 98         struct sockaddr_in sin;
 99         struct sockaddr_in6 sin6;
100         struct sockaddr sa;
101 };
102 
103 struct vxlan_rdst {
104         union vxlan_addr         remote_ip;
105         __be16                   remote_port;
106         u32                      remote_vni;
107         u32                      remote_ifindex;
108         struct list_head         list;
109         struct rcu_head          rcu;
110 };
111 
112 /* Forwarding table entry */
113 struct vxlan_fdb {
114         struct hlist_node hlist;        /* linked list of entries */
115         struct rcu_head   rcu;
116         unsigned long     updated;      /* jiffies */
117         unsigned long     used;
118         struct list_head  remotes;
119         u16               state;        /* see ndm_state */
120         u8                flags;        /* see ndm_flags */
121         u8                eth_addr[ETH_ALEN];
122 };
123 
124 /* Pseudo network device */
125 struct vxlan_dev {
126         struct hlist_node hlist;        /* vni hash table */
127         struct list_head  next;         /* vxlan's per namespace list */
128         struct vxlan_sock *vn_sock;     /* listening socket */
129         struct net_device *dev;
130         struct net        *net;         /* netns for packet i/o */
131         struct vxlan_rdst default_dst;  /* default destination */
132         union vxlan_addr  saddr;        /* source address */
133         __be16            dst_port;
134         __u16             port_min;     /* source port range */
135         __u16             port_max;
136         __u8              tos;          /* TOS override */
137         __u8              ttl;
138         u32               flags;        /* VXLAN_F_* in vxlan.h */
139 
140         struct work_struct sock_work;
141         struct work_struct igmp_join;
142         struct work_struct igmp_leave;
143 
144         unsigned long     age_interval;
145         struct timer_list age_timer;
146         spinlock_t        hash_lock;
147         unsigned int      addrcnt;
148         unsigned int      addrmax;
149 
150         struct hlist_head fdb_head[FDB_HASH_SIZE];
151 };
152 
153 /* salt for hash table */
154 static u32 vxlan_salt __read_mostly;
155 static struct workqueue_struct *vxlan_wq;
156 
157 static void vxlan_sock_work(struct work_struct *work);
158 
159 #if IS_ENABLED(CONFIG_IPV6)
160 static inline
161 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
162 {
163        if (a->sa.sa_family != b->sa.sa_family)
164                return false;
165        if (a->sa.sa_family == AF_INET6)
166                return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
167        else
168                return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
169 }
170 
171 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
172 {
173        if (ipa->sa.sa_family == AF_INET6)
174                return ipv6_addr_any(&ipa->sin6.sin6_addr);
175        else
176                return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
177 }
178 
179 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
180 {
181        if (ipa->sa.sa_family == AF_INET6)
182                return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
183        else
184                return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
185 }
186 
187 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
188 {
189        if (nla_len(nla) >= sizeof(struct in6_addr)) {
190                nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr));
191                ip->sa.sa_family = AF_INET6;
192                return 0;
193        } else if (nla_len(nla) >= sizeof(__be32)) {
194                ip->sin.sin_addr.s_addr = nla_get_be32(nla);
195                ip->sa.sa_family = AF_INET;
196                return 0;
197        } else {
198                return -EAFNOSUPPORT;
199        }
200 }
201 
202 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
203                              const union vxlan_addr *ip)
204 {
205        if (ip->sa.sa_family == AF_INET6)
206                return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr);
207        else
208                return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
209 }
210 
211 #else /* !CONFIG_IPV6 */
212 
213 static inline
214 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
215 {
216        return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
217 }
218 
219 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
220 {
221        return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
222 }
223 
224 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
225 {
226        return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
227 }
228 
229 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
230 {
231        if (nla_len(nla) >= sizeof(struct in6_addr)) {
232                return -EAFNOSUPPORT;
233        } else if (nla_len(nla) >= sizeof(__be32)) {
234                ip->sin.sin_addr.s_addr = nla_get_be32(nla);
235                ip->sa.sa_family = AF_INET;
236                return 0;
237        } else {
238                return -EAFNOSUPPORT;
239        }
240 }
241 
242 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
243                              const union vxlan_addr *ip)
244 {
245        return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
246 }
247 #endif
248 
249 /* Virtual Network hash table head */
250 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
251 {
252         return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
253 }
254 
255 /* Socket hash table head */
256 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
257 {
258         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
259 
260         return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
261 }
262 
263 /* First remote destination for a forwarding entry.
264  * Guaranteed to be non-NULL because remotes are never deleted.
265  */
266 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
267 {
268         return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
269 }
270 
271 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
272 {
273         return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
274 }
275 
276 /* Find VXLAN socket based on network namespace and UDP port */
277 static struct vxlan_sock *vxlan_find_sock(struct net *net, __be16 port)
278 {
279         struct vxlan_sock *vs;
280 
281         hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
282                 if (inet_sk(vs->sock->sk)->inet_sport == port)
283                         return vs;
284         }
285         return NULL;
286 }
287 
288 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
289 {
290         struct vxlan_dev *vxlan;
291 
292         hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
293                 if (vxlan->default_dst.remote_vni == id)
294                         return vxlan;
295         }
296 
297         return NULL;
298 }
299 
300 /* Look up VNI in a per net namespace table */
301 static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port)
302 {
303         struct vxlan_sock *vs;
304 
305         vs = vxlan_find_sock(net, port);
306         if (!vs)
307                 return NULL;
308 
309         return vxlan_vs_find_vni(vs, id);
310 }
311 
312 /* Fill in neighbour message in skbuff. */
313 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
314                           const struct vxlan_fdb *fdb,
315                           u32 portid, u32 seq, int type, unsigned int flags,
316                           const struct vxlan_rdst *rdst)
317 {
318         unsigned long now = jiffies;
319         struct nda_cacheinfo ci;
320         struct nlmsghdr *nlh;
321         struct ndmsg *ndm;
322         bool send_ip, send_eth;
323 
324         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
325         if (nlh == NULL)
326                 return -EMSGSIZE;
327 
328         ndm = nlmsg_data(nlh);
329         memset(ndm, 0, sizeof(*ndm));
330 
331         send_eth = send_ip = true;
332 
333         if (type == RTM_GETNEIGH) {
334                 ndm->ndm_family = AF_INET;
335                 send_ip = !vxlan_addr_any(&rdst->remote_ip);
336                 send_eth = !is_zero_ether_addr(fdb->eth_addr);
337         } else
338                 ndm->ndm_family = AF_BRIDGE;
339         ndm->ndm_state = fdb->state;
340         ndm->ndm_ifindex = vxlan->dev->ifindex;
341         ndm->ndm_flags = fdb->flags;
342         ndm->ndm_type = RTN_UNICAST;
343 
344         if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
345                 goto nla_put_failure;
346 
347         if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
348                 goto nla_put_failure;
349 
350         if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
351             nla_put_be16(skb, NDA_PORT, rdst->remote_port))
352                 goto nla_put_failure;
353         if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
354             nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
355                 goto nla_put_failure;
356         if (rdst->remote_ifindex &&
357             nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
358                 goto nla_put_failure;
359 
360         ci.ndm_used      = jiffies_to_clock_t(now - fdb->used);
361         ci.ndm_confirmed = 0;
362         ci.ndm_updated   = jiffies_to_clock_t(now - fdb->updated);
363         ci.ndm_refcnt    = 0;
364 
365         if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
366                 goto nla_put_failure;
367 
368         return nlmsg_end(skb, nlh);
369 
370 nla_put_failure:
371         nlmsg_cancel(skb, nlh);
372         return -EMSGSIZE;
373 }
374 
375 static inline size_t vxlan_nlmsg_size(void)
376 {
377         return NLMSG_ALIGN(sizeof(struct ndmsg))
378                 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
379                 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
380                 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
381                 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
382                 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
383                 + nla_total_size(sizeof(struct nda_cacheinfo));
384 }
385 
386 static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
387                              struct vxlan_rdst *rd, int type)
388 {
389         struct net *net = dev_net(vxlan->dev);
390         struct sk_buff *skb;
391         int err = -ENOBUFS;
392 
393         skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
394         if (skb == NULL)
395                 goto errout;
396 
397         err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
398         if (err < 0) {
399                 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
400                 WARN_ON(err == -EMSGSIZE);
401                 kfree_skb(skb);
402                 goto errout;
403         }
404 
405         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
406         return;
407 errout:
408         if (err < 0)
409                 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
410 }
411 
412 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
413 {
414         struct vxlan_dev *vxlan = netdev_priv(dev);
415         struct vxlan_fdb f = {
416                 .state = NUD_STALE,
417         };
418         struct vxlan_rdst remote = {
419                 .remote_ip = *ipa, /* goes to NDA_DST */
420                 .remote_vni = VXLAN_N_VID,
421         };
422 
423         vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
424 }
425 
426 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
427 {
428         struct vxlan_fdb f = {
429                 .state = NUD_STALE,
430         };
431         struct vxlan_rdst remote = { };
432 
433         memcpy(f.eth_addr, eth_addr, ETH_ALEN);
434 
435         vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
436 }
437 
438 /* Hash Ethernet address */
439 static u32 eth_hash(const unsigned char *addr)
440 {
441         u64 value = get_unaligned((u64 *)addr);
442 
443         /* only want 6 bytes */
444 #ifdef __BIG_ENDIAN
445         value >>= 16;
446 #else
447         value <<= 16;
448 #endif
449         return hash_64(value, FDB_HASH_BITS);
450 }
451 
452 /* Hash chain to use given mac address */
453 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
454                                                 const u8 *mac)
455 {
456         return &vxlan->fdb_head[eth_hash(mac)];
457 }
458 
459 /* Look up Ethernet address in forwarding table */
460 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
461                                         const u8 *mac)
462 {
463         struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
464         struct vxlan_fdb *f;
465 
466         hlist_for_each_entry_rcu(f, head, hlist) {
467                 if (ether_addr_equal(mac, f->eth_addr))
468                         return f;
469         }
470 
471         return NULL;
472 }
473 
474 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
475                                         const u8 *mac)
476 {
477         struct vxlan_fdb *f;
478 
479         f = __vxlan_find_mac(vxlan, mac);
480         if (f)
481                 f->used = jiffies;
482 
483         return f;
484 }
485 
486 /* caller should hold vxlan->hash_lock */
487 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
488                                               union vxlan_addr *ip, __be16 port,
489                                               __u32 vni, __u32 ifindex)
490 {
491         struct vxlan_rdst *rd;
492 
493         list_for_each_entry(rd, &f->remotes, list) {
494                 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
495                     rd->remote_port == port &&
496                     rd->remote_vni == vni &&
497                     rd->remote_ifindex == ifindex)
498                         return rd;
499         }
500 
501         return NULL;
502 }
503 
504 /* Replace destination of unicast mac */
505 static int vxlan_fdb_replace(struct vxlan_fdb *f,
506                              union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
507 {
508         struct vxlan_rdst *rd;
509 
510         rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
511         if (rd)
512                 return 0;
513 
514         rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
515         if (!rd)
516                 return 0;
517         rd->remote_ip = *ip;
518         rd->remote_port = port;
519         rd->remote_vni = vni;
520         rd->remote_ifindex = ifindex;
521         return 1;
522 }
523 
524 /* Add/update destinations for multicast */
525 static int vxlan_fdb_append(struct vxlan_fdb *f,
526                             union vxlan_addr *ip, __be16 port, __u32 vni,
527                             __u32 ifindex, struct vxlan_rdst **rdp)
528 {
529         struct vxlan_rdst *rd;
530 
531         rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
532         if (rd)
533                 return 0;
534 
535         rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
536         if (rd == NULL)
537                 return -ENOBUFS;
538         rd->remote_ip = *ip;
539         rd->remote_port = port;
540         rd->remote_vni = vni;
541         rd->remote_ifindex = ifindex;
542 
543         list_add_tail_rcu(&rd->list, &f->remotes);
544 
545         *rdp = rd;
546         return 1;
547 }
548 
549 static struct sk_buff **vxlan_gro_receive(struct sk_buff **head, struct sk_buff *skb)
550 {
551         struct sk_buff *p, **pp = NULL;
552         struct vxlanhdr *vh, *vh2;
553         struct ethhdr *eh, *eh2;
554         unsigned int hlen, off_vx, off_eth;
555         const struct packet_offload *ptype;
556         __be16 type;
557         int flush = 1;
558 
559         off_vx = skb_gro_offset(skb);
560         hlen = off_vx + sizeof(*vh);
561         vh   = skb_gro_header_fast(skb, off_vx);
562         if (skb_gro_header_hard(skb, hlen)) {
563                 vh = skb_gro_header_slow(skb, hlen, off_vx);
564                 if (unlikely(!vh))
565                         goto out;
566         }
567         skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
568         skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
569 
570         off_eth = skb_gro_offset(skb);
571         hlen = off_eth + sizeof(*eh);
572         eh   = skb_gro_header_fast(skb, off_eth);
573         if (skb_gro_header_hard(skb, hlen)) {
574                 eh = skb_gro_header_slow(skb, hlen, off_eth);
575                 if (unlikely(!eh))
576                         goto out;
577         }
578 
579         flush = 0;
580 
581         for (p = *head; p; p = p->next) {
582                 if (!NAPI_GRO_CB(p)->same_flow)
583                         continue;
584 
585                 vh2 = (struct vxlanhdr *)(p->data + off_vx);
586                 eh2 = (struct ethhdr   *)(p->data + off_eth);
587                 if (vh->vx_vni != vh2->vx_vni || compare_ether_header(eh, eh2)) {
588                         NAPI_GRO_CB(p)->same_flow = 0;
589                         continue;
590                 }
591         }
592 
593         type = eh->h_proto;
594 
595         rcu_read_lock();
596         ptype = gro_find_receive_by_type(type);
597         if (ptype == NULL) {
598                 flush = 1;
599                 goto out_unlock;
600         }
601 
602         skb_gro_pull(skb, sizeof(*eh)); /* pull inner eth header */
603         skb_gro_postpull_rcsum(skb, eh, sizeof(*eh));
604         pp = ptype->callbacks.gro_receive(head, skb);
605 
606 out_unlock:
607         rcu_read_unlock();
608 out:
609         NAPI_GRO_CB(skb)->flush |= flush;
610 
611         return pp;
612 }
613 
614 static int vxlan_gro_complete(struct sk_buff *skb, int nhoff)
615 {
616         struct ethhdr *eh;
617         struct packet_offload *ptype;
618         __be16 type;
619         int vxlan_len  = sizeof(struct vxlanhdr) + sizeof(struct ethhdr);
620         int err = -ENOSYS;
621 
622         eh = (struct ethhdr *)(skb->data + nhoff + sizeof(struct vxlanhdr));
623         type = eh->h_proto;
624 
625         rcu_read_lock();
626         ptype = gro_find_complete_by_type(type);
627         if (ptype != NULL)
628                 err = ptype->callbacks.gro_complete(skb, nhoff + vxlan_len);
629 
630         rcu_read_unlock();
631         return err;
632 }
633 
634 /* Notify netdevs that UDP port started listening */
635 static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
636 {
637         struct net_device *dev;
638         struct sock *sk = vs->sock->sk;
639         struct net *net = sock_net(sk);
640         sa_family_t sa_family = sk->sk_family;
641         __be16 port = inet_sk(sk)->inet_sport;
642         int err;
643 
644         if (sa_family == AF_INET) {
645                 err = udp_add_offload(&vs->udp_offloads);
646                 if (err)
647                         pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
648         }
649 
650         rcu_read_lock();
651         for_each_netdev_rcu(net, dev) {
652                 if (dev->netdev_ops->ndo_add_vxlan_port)
653                         dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
654                                                             port);
655         }
656         rcu_read_unlock();
657 }
658 
659 /* Notify netdevs that UDP port is no more listening */
660 static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
661 {
662         struct net_device *dev;
663         struct sock *sk = vs->sock->sk;
664         struct net *net = sock_net(sk);
665         sa_family_t sa_family = sk->sk_family;
666         __be16 port = inet_sk(sk)->inet_sport;
667 
668         rcu_read_lock();
669         for_each_netdev_rcu(net, dev) {
670                 if (dev->netdev_ops->ndo_del_vxlan_port)
671                         dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
672                                                             port);
673         }
674         rcu_read_unlock();
675 
676         if (sa_family == AF_INET)
677                 udp_del_offload(&vs->udp_offloads);
678 }
679 
680 /* Add new entry to forwarding table -- assumes lock held */
681 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
682                             const u8 *mac, union vxlan_addr *ip,
683                             __u16 state, __u16 flags,
684                             __be16 port, __u32 vni, __u32 ifindex,
685                             __u8 ndm_flags)
686 {
687         struct vxlan_rdst *rd = NULL;
688         struct vxlan_fdb *f;
689         int notify = 0;
690 
691         f = __vxlan_find_mac(vxlan, mac);
692         if (f) {
693                 if (flags & NLM_F_EXCL) {
694                         netdev_dbg(vxlan->dev,
695                                    "lost race to create %pM\n", mac);
696                         return -EEXIST;
697                 }
698                 if (f->state != state) {
699                         f->state = state;
700                         f->updated = jiffies;
701                         notify = 1;
702                 }
703                 if (f->flags != ndm_flags) {
704                         f->flags = ndm_flags;
705                         f->updated = jiffies;
706                         notify = 1;
707                 }
708                 if ((flags & NLM_F_REPLACE)) {
709                         /* Only change unicasts */
710                         if (!(is_multicast_ether_addr(f->eth_addr) ||
711                              is_zero_ether_addr(f->eth_addr))) {
712                                 int rc = vxlan_fdb_replace(f, ip, port, vni,
713                                                            ifindex);
714 
715                                 if (rc < 0)
716                                         return rc;
717                                 notify |= rc;
718                         } else
719                                 return -EOPNOTSUPP;
720                 }
721                 if ((flags & NLM_F_APPEND) &&
722                     (is_multicast_ether_addr(f->eth_addr) ||
723                      is_zero_ether_addr(f->eth_addr))) {
724                         int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
725                                                   &rd);
726 
727                         if (rc < 0)
728                                 return rc;
729                         notify |= rc;
730                 }
731         } else {
732                 if (!(flags & NLM_F_CREATE))
733                         return -ENOENT;
734 
735                 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
736                         return -ENOSPC;
737 
738                 /* Disallow replace to add a multicast entry */
739                 if ((flags & NLM_F_REPLACE) &&
740                     (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
741                         return -EOPNOTSUPP;
742 
743                 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
744                 f = kmalloc(sizeof(*f), GFP_ATOMIC);
745                 if (!f)
746                         return -ENOMEM;
747 
748                 notify = 1;
749                 f->state = state;
750                 f->flags = ndm_flags;
751                 f->updated = f->used = jiffies;
752                 INIT_LIST_HEAD(&f->remotes);
753                 memcpy(f->eth_addr, mac, ETH_ALEN);
754 
755                 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
756 
757                 ++vxlan->addrcnt;
758                 hlist_add_head_rcu(&f->hlist,
759                                    vxlan_fdb_head(vxlan, mac));
760         }
761 
762         if (notify) {
763                 if (rd == NULL)
764                         rd = first_remote_rtnl(f);
765                 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
766         }
767 
768         return 0;
769 }
770 
771 static void vxlan_fdb_free(struct rcu_head *head)
772 {
773         struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
774         struct vxlan_rdst *rd, *nd;
775 
776         list_for_each_entry_safe(rd, nd, &f->remotes, list)
777                 kfree(rd);
778         kfree(f);
779 }
780 
781 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
782 {
783         netdev_dbg(vxlan->dev,
784                     "delete %pM\n", f->eth_addr);
785 
786         --vxlan->addrcnt;
787         vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
788 
789         hlist_del_rcu(&f->hlist);
790         call_rcu(&f->rcu, vxlan_fdb_free);
791 }
792 
793 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
794                            union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
795 {
796         struct net *net = dev_net(vxlan->dev);
797         int err;
798 
799         if (tb[NDA_DST]) {
800                 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
801                 if (err)
802                         return err;
803         } else {
804                 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
805                 if (remote->sa.sa_family == AF_INET) {
806                         ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
807                         ip->sa.sa_family = AF_INET;
808 #if IS_ENABLED(CONFIG_IPV6)
809                 } else {
810                         ip->sin6.sin6_addr = in6addr_any;
811                         ip->sa.sa_family = AF_INET6;
812 #endif
813                 }
814         }
815 
816         if (tb[NDA_PORT]) {
817                 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
818                         return -EINVAL;
819                 *port = nla_get_be16(tb[NDA_PORT]);
820         } else {
821                 *port = vxlan->dst_port;
822         }
823 
824         if (tb[NDA_VNI]) {
825                 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
826                         return -EINVAL;
827                 *vni = nla_get_u32(tb[NDA_VNI]);
828         } else {
829                 *vni = vxlan->default_dst.remote_vni;
830         }
831 
832         if (tb[NDA_IFINDEX]) {
833                 struct net_device *tdev;
834 
835                 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
836                         return -EINVAL;
837                 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
838                 tdev = __dev_get_by_index(net, *ifindex);
839                 if (!tdev)
840                         return -EADDRNOTAVAIL;
841         } else {
842                 *ifindex = 0;
843         }
844 
845         return 0;
846 }
847 
848 /* Add static entry (via netlink) */
849 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
850                          struct net_device *dev,
851                          const unsigned char *addr, u16 flags)
852 {
853         struct vxlan_dev *vxlan = netdev_priv(dev);
854         /* struct net *net = dev_net(vxlan->dev); */
855         union vxlan_addr ip;
856         __be16 port;
857         u32 vni, ifindex;
858         int err;
859 
860         if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
861                 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
862                         ndm->ndm_state);
863                 return -EINVAL;
864         }
865 
866         if (tb[NDA_DST] == NULL)
867                 return -EINVAL;
868 
869         err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
870         if (err)
871                 return err;
872 
873         if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
874                 return -EAFNOSUPPORT;
875 
876         spin_lock_bh(&vxlan->hash_lock);
877         err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
878                                port, vni, ifindex, ndm->ndm_flags);
879         spin_unlock_bh(&vxlan->hash_lock);
880 
881         return err;
882 }
883 
884 /* Delete entry (via netlink) */
885 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
886                             struct net_device *dev,
887                             const unsigned char *addr)
888 {
889         struct vxlan_dev *vxlan = netdev_priv(dev);
890         struct vxlan_fdb *f;
891         struct vxlan_rdst *rd = NULL;
892         union vxlan_addr ip;
893         __be16 port;
894         u32 vni, ifindex;
895         int err;
896 
897         err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
898         if (err)
899                 return err;
900 
901         err = -ENOENT;
902 
903         spin_lock_bh(&vxlan->hash_lock);
904         f = vxlan_find_mac(vxlan, addr);
905         if (!f)
906                 goto out;
907 
908         if (!vxlan_addr_any(&ip)) {
909                 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
910                 if (!rd)
911                         goto out;
912         }
913 
914         err = 0;
915 
916         /* remove a destination if it's not the only one on the list,
917          * otherwise destroy the fdb entry
918          */
919         if (rd && !list_is_singular(&f->remotes)) {
920                 list_del_rcu(&rd->list);
921                 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
922                 kfree_rcu(rd, rcu);
923                 goto out;
924         }
925 
926         vxlan_fdb_destroy(vxlan, f);
927 
928 out:
929         spin_unlock_bh(&vxlan->hash_lock);
930 
931         return err;
932 }
933 
934 /* Dump forwarding table */
935 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
936                           struct net_device *dev, int idx)
937 {
938         struct vxlan_dev *vxlan = netdev_priv(dev);
939         unsigned int h;
940 
941         for (h = 0; h < FDB_HASH_SIZE; ++h) {
942                 struct vxlan_fdb *f;
943                 int err;
944 
945                 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
946                         struct vxlan_rdst *rd;
947 
948                         if (idx < cb->args[0])
949                                 goto skip;
950 
951                         list_for_each_entry_rcu(rd, &f->remotes, list) {
952                                 err = vxlan_fdb_info(skb, vxlan, f,
953                                                      NETLINK_CB(cb->skb).portid,
954                                                      cb->nlh->nlmsg_seq,
955                                                      RTM_NEWNEIGH,
956                                                      NLM_F_MULTI, rd);
957                                 if (err < 0)
958                                         goto out;
959                         }
960 skip:
961                         ++idx;
962                 }
963         }
964 out:
965         return idx;
966 }
967 
968 /* Watch incoming packets to learn mapping between Ethernet address
969  * and Tunnel endpoint.
970  * Return true if packet is bogus and should be droppped.
971  */
972 static bool vxlan_snoop(struct net_device *dev,
973                         union vxlan_addr *src_ip, const u8 *src_mac)
974 {
975         struct vxlan_dev *vxlan = netdev_priv(dev);
976         struct vxlan_fdb *f;
977 
978         f = vxlan_find_mac(vxlan, src_mac);
979         if (likely(f)) {
980                 struct vxlan_rdst *rdst = first_remote_rcu(f);
981 
982                 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
983                         return false;
984 
985                 /* Don't migrate static entries, drop packets */
986                 if (f->state & NUD_NOARP)
987                         return true;
988 
989                 if (net_ratelimit())
990                         netdev_info(dev,
991                                     "%pM migrated from %pIS to %pIS\n",
992                                     src_mac, &rdst->remote_ip, &src_ip);
993 
994                 rdst->remote_ip = *src_ip;
995                 f->updated = jiffies;
996                 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
997         } else {
998                 /* learned new entry */
999                 spin_lock(&vxlan->hash_lock);
1000 
1001                 /* close off race between vxlan_flush and incoming packets */
1002                 if (netif_running(dev))
1003                         vxlan_fdb_create(vxlan, src_mac, src_ip,
1004                                          NUD_REACHABLE,
1005                                          NLM_F_EXCL|NLM_F_CREATE,
1006                                          vxlan->dst_port,
1007                                          vxlan->default_dst.remote_vni,
1008                                          0, NTF_SELF);
1009                 spin_unlock(&vxlan->hash_lock);
1010         }
1011 
1012         return false;
1013 }
1014 
1015 /* See if multicast group is already in use by other ID */
1016 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1017 {
1018         struct vxlan_dev *vxlan;
1019 
1020         /* The vxlan_sock is only used by dev, leaving group has
1021          * no effect on other vxlan devices.
1022          */
1023         if (atomic_read(&dev->vn_sock->refcnt) == 1)
1024                 return false;
1025 
1026         list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1027                 if (!netif_running(vxlan->dev) || vxlan == dev)
1028                         continue;
1029 
1030                 if (vxlan->vn_sock != dev->vn_sock)
1031                         continue;
1032 
1033                 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1034                                       &dev->default_dst.remote_ip))
1035                         continue;
1036 
1037                 if (vxlan->default_dst.remote_ifindex !=
1038                     dev->default_dst.remote_ifindex)
1039                         continue;
1040 
1041                 return true;
1042         }
1043 
1044         return false;
1045 }
1046 
1047 static void vxlan_sock_hold(struct vxlan_sock *vs)
1048 {
1049         atomic_inc(&vs->refcnt);
1050 }
1051 
1052 void vxlan_sock_release(struct vxlan_sock *vs)
1053 {
1054         struct sock *sk = vs->sock->sk;
1055         struct net *net = sock_net(sk);
1056         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1057 
1058         if (!atomic_dec_and_test(&vs->refcnt))
1059                 return;
1060 
1061         spin_lock(&vn->sock_lock);
1062         hlist_del_rcu(&vs->hlist);
1063         rcu_assign_sk_user_data(vs->sock->sk, NULL);
1064         vxlan_notify_del_rx_port(vs);
1065         spin_unlock(&vn->sock_lock);
1066 
1067         queue_work(vxlan_wq, &vs->del_work);
1068 }
1069 EXPORT_SYMBOL_GPL(vxlan_sock_release);
1070 
1071 /* Callback to update multicast group membership when first VNI on
1072  * multicast asddress is brought up
1073  * Done as workqueue because ip_mc_join_group acquires RTNL.
1074  */
1075 static void vxlan_igmp_join(struct work_struct *work)
1076 {
1077         struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
1078         struct vxlan_sock *vs = vxlan->vn_sock;
1079         struct sock *sk = vs->sock->sk;
1080         union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1081         int ifindex = vxlan->default_dst.remote_ifindex;
1082 
1083         lock_sock(sk);
1084         if (ip->sa.sa_family == AF_INET) {
1085                 struct ip_mreqn mreq = {
1086                         .imr_multiaddr.s_addr   = ip->sin.sin_addr.s_addr,
1087                         .imr_ifindex            = ifindex,
1088                 };
1089 
1090                 ip_mc_join_group(sk, &mreq);
1091 #if IS_ENABLED(CONFIG_IPV6)
1092         } else {
1093                 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1094                                              &ip->sin6.sin6_addr);
1095 #endif
1096         }
1097         release_sock(sk);
1098 
1099         vxlan_sock_release(vs);
1100         dev_put(vxlan->dev);
1101 }
1102 
1103 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1104 static void vxlan_igmp_leave(struct work_struct *work)
1105 {
1106         struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
1107         struct vxlan_sock *vs = vxlan->vn_sock;
1108         struct sock *sk = vs->sock->sk;
1109         union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1110         int ifindex = vxlan->default_dst.remote_ifindex;
1111 
1112         lock_sock(sk);
1113         if (ip->sa.sa_family == AF_INET) {
1114                 struct ip_mreqn mreq = {
1115                         .imr_multiaddr.s_addr   = ip->sin.sin_addr.s_addr,
1116                         .imr_ifindex            = ifindex,
1117                 };
1118 
1119                 ip_mc_leave_group(sk, &mreq);
1120 #if IS_ENABLED(CONFIG_IPV6)
1121         } else {
1122                 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1123                                              &ip->sin6.sin6_addr);
1124 #endif
1125         }
1126 
1127         release_sock(sk);
1128 
1129         vxlan_sock_release(vs);
1130         dev_put(vxlan->dev);
1131 }
1132 
1133 /* Callback from net/ipv4/udp.c to receive packets */
1134 static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1135 {
1136         struct vxlan_sock *vs;
1137         struct vxlanhdr *vxh;
1138 
1139         /* Need Vxlan and inner Ethernet header to be present */
1140         if (!pskb_may_pull(skb, VXLAN_HLEN))
1141                 goto error;
1142 
1143         /* Return packets with reserved bits set */
1144         vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1145         if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
1146             (vxh->vx_vni & htonl(0xff))) {
1147                 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1148                            ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1149                 goto error;
1150         }
1151 
1152         if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1153                 goto drop;
1154 
1155         vs = rcu_dereference_sk_user_data(sk);
1156         if (!vs)
1157                 goto drop;
1158 
1159         skb_pop_rcv_encapsulation(skb);
1160 
1161         vs->rcv(vs, skb, vxh->vx_vni);
1162         return 0;
1163 
1164 drop:
1165         /* Consume bad packet */
1166         kfree_skb(skb);
1167         return 0;
1168 
1169 error:
1170         /* Return non vxlan pkt */
1171         return 1;
1172 }
1173 
1174 static void vxlan_rcv(struct vxlan_sock *vs,
1175                       struct sk_buff *skb, __be32 vx_vni)
1176 {
1177         struct iphdr *oip = NULL;
1178         struct ipv6hdr *oip6 = NULL;
1179         struct vxlan_dev *vxlan;
1180         struct pcpu_sw_netstats *stats;
1181         union vxlan_addr saddr;
1182         __u32 vni;
1183         int err = 0;
1184         union vxlan_addr *remote_ip;
1185 
1186         vni = ntohl(vx_vni) >> 8;
1187         /* Is this VNI defined? */
1188         vxlan = vxlan_vs_find_vni(vs, vni);
1189         if (!vxlan)
1190                 goto drop;
1191 
1192         remote_ip = &vxlan->default_dst.remote_ip;
1193         skb_reset_mac_header(skb);
1194         skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
1195         skb->protocol = eth_type_trans(skb, vxlan->dev);
1196         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1197 
1198         /* Ignore packet loops (and multicast echo) */
1199         if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1200                 goto drop;
1201 
1202         /* Re-examine inner Ethernet packet */
1203         if (remote_ip->sa.sa_family == AF_INET) {
1204                 oip = ip_hdr(skb);
1205                 saddr.sin.sin_addr.s_addr = oip->saddr;
1206                 saddr.sa.sa_family = AF_INET;
1207 #if IS_ENABLED(CONFIG_IPV6)
1208         } else {
1209                 oip6 = ipv6_hdr(skb);
1210                 saddr.sin6.sin6_addr = oip6->saddr;
1211                 saddr.sa.sa_family = AF_INET6;
1212 #endif
1213         }
1214 
1215         if ((vxlan->flags & VXLAN_F_LEARN) &&
1216             vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1217                 goto drop;
1218 
1219         skb_reset_network_header(skb);
1220 
1221         if (oip6)
1222                 err = IP6_ECN_decapsulate(oip6, skb);
1223         if (oip)
1224                 err = IP_ECN_decapsulate(oip, skb);
1225 
1226         if (unlikely(err)) {
1227                 if (log_ecn_error) {
1228                         if (oip6)
1229                                 net_info_ratelimited("non-ECT from %pI6\n",
1230                                                      &oip6->saddr);
1231                         if (oip)
1232                                 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1233                                                      &oip->saddr, oip->tos);
1234                 }
1235                 if (err > 1) {
1236                         ++vxlan->dev->stats.rx_frame_errors;
1237                         ++vxlan->dev->stats.rx_errors;
1238                         goto drop;
1239                 }
1240         }
1241 
1242         stats = this_cpu_ptr(vxlan->dev->tstats);
1243         u64_stats_update_begin(&stats->syncp);
1244         stats->rx_packets++;
1245         stats->rx_bytes += skb->len;
1246         u64_stats_update_end(&stats->syncp);
1247 
1248         netif_rx(skb);
1249 
1250         return;
1251 drop:
1252         /* Consume bad packet */
1253         kfree_skb(skb);
1254 }
1255 
1256 static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1257 {
1258         struct vxlan_dev *vxlan = netdev_priv(dev);
1259         struct arphdr *parp;
1260         u8 *arpptr, *sha;
1261         __be32 sip, tip;
1262         struct neighbour *n;
1263 
1264         if (dev->flags & IFF_NOARP)
1265                 goto out;
1266 
1267         if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1268                 dev->stats.tx_dropped++;
1269                 goto out;
1270         }
1271         parp = arp_hdr(skb);
1272 
1273         if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1274              parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1275             parp->ar_pro != htons(ETH_P_IP) ||
1276             parp->ar_op != htons(ARPOP_REQUEST) ||
1277             parp->ar_hln != dev->addr_len ||
1278             parp->ar_pln != 4)
1279                 goto out;
1280         arpptr = (u8 *)parp + sizeof(struct arphdr);
1281         sha = arpptr;
1282         arpptr += dev->addr_len;        /* sha */
1283         memcpy(&sip, arpptr, sizeof(sip));
1284         arpptr += sizeof(sip);
1285         arpptr += dev->addr_len;        /* tha */
1286         memcpy(&tip, arpptr, sizeof(tip));
1287 
1288         if (ipv4_is_loopback(tip) ||
1289             ipv4_is_multicast(tip))
1290                 goto out;
1291 
1292         n = neigh_lookup(&arp_tbl, &tip, dev);
1293 
1294         if (n) {
1295                 struct vxlan_fdb *f;
1296                 struct sk_buff  *reply;
1297 
1298                 if (!(n->nud_state & NUD_CONNECTED)) {
1299                         neigh_release(n);
1300                         goto out;
1301                 }
1302 
1303                 f = vxlan_find_mac(vxlan, n->ha);
1304                 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1305                         /* bridge-local neighbor */
1306                         neigh_release(n);
1307                         goto out;
1308                 }
1309 
1310                 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1311                                 n->ha, sha);
1312 
1313                 neigh_release(n);
1314 
1315                 if (reply == NULL)
1316                         goto out;
1317 
1318                 skb_reset_mac_header(reply);
1319                 __skb_pull(reply, skb_network_offset(reply));
1320                 reply->ip_summed = CHECKSUM_UNNECESSARY;
1321                 reply->pkt_type = PACKET_HOST;
1322 
1323                 if (netif_rx_ni(reply) == NET_RX_DROP)
1324                         dev->stats.rx_dropped++;
1325         } else if (vxlan->flags & VXLAN_F_L3MISS) {
1326                 union vxlan_addr ipa = {
1327                         .sin.sin_addr.s_addr = tip,
1328                         .sa.sa_family = AF_INET,
1329                 };
1330 
1331                 vxlan_ip_miss(dev, &ipa);
1332         }
1333 out:
1334         consume_skb(skb);
1335         return NETDEV_TX_OK;
1336 }
1337 
1338 #if IS_ENABLED(CONFIG_IPV6)
1339 
1340 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1341         struct neighbour *n, bool isrouter)
1342 {
1343         struct net_device *dev = request->dev;
1344         struct sk_buff *reply;
1345         struct nd_msg *ns, *na;
1346         struct ipv6hdr *pip6;
1347         u8 *daddr;
1348         int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1349         int ns_olen;
1350         int i, len;
1351 
1352         if (dev == NULL)
1353                 return NULL;
1354 
1355         len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1356                 sizeof(*na) + na_olen + dev->needed_tailroom;
1357         reply = alloc_skb(len, GFP_ATOMIC);
1358         if (reply == NULL)
1359                 return NULL;
1360 
1361         reply->protocol = htons(ETH_P_IPV6);
1362         reply->dev = dev;
1363         skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1364         skb_push(reply, sizeof(struct ethhdr));
1365         skb_set_mac_header(reply, 0);
1366 
1367         ns = (struct nd_msg *)skb_transport_header(request);
1368 
1369         daddr = eth_hdr(request)->h_source;
1370         ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1371         for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1372                 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1373                         daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1374                         break;
1375                 }
1376         }
1377 
1378         /* Ethernet header */
1379         ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1380         ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1381         eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1382         reply->protocol = htons(ETH_P_IPV6);
1383 
1384         skb_pull(reply, sizeof(struct ethhdr));
1385         skb_set_network_header(reply, 0);
1386         skb_put(reply, sizeof(struct ipv6hdr));
1387 
1388         /* IPv6 header */
1389 
1390         pip6 = ipv6_hdr(reply);
1391         memset(pip6, 0, sizeof(struct ipv6hdr));
1392         pip6->version = 6;
1393         pip6->priority = ipv6_hdr(request)->priority;
1394         pip6->nexthdr = IPPROTO_ICMPV6;
1395         pip6->hop_limit = 255;
1396         pip6->daddr = ipv6_hdr(request)->saddr;
1397         pip6->saddr = *(struct in6_addr *)n->primary_key;
1398 
1399         skb_pull(reply, sizeof(struct ipv6hdr));
1400         skb_set_transport_header(reply, 0);
1401 
1402         na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1403 
1404         /* Neighbor Advertisement */
1405         memset(na, 0, sizeof(*na)+na_olen);
1406         na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1407         na->icmph.icmp6_router = isrouter;
1408         na->icmph.icmp6_override = 1;
1409         na->icmph.icmp6_solicited = 1;
1410         na->target = ns->target;
1411         ether_addr_copy(&na->opt[2], n->ha);
1412         na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1413         na->opt[1] = na_olen >> 3;
1414 
1415         na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1416                 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1417                 csum_partial(na, sizeof(*na)+na_olen, 0));
1418 
1419         pip6->payload_len = htons(sizeof(*na)+na_olen);
1420 
1421         skb_push(reply, sizeof(struct ipv6hdr));
1422 
1423         reply->ip_summed = CHECKSUM_UNNECESSARY;
1424 
1425         return reply;
1426 }
1427 
1428 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1429 {
1430         struct vxlan_dev *vxlan = netdev_priv(dev);
1431         struct nd_msg *msg;
1432         const struct ipv6hdr *iphdr;
1433         const struct in6_addr *saddr, *daddr;
1434         struct neighbour *n;
1435         struct inet6_dev *in6_dev;
1436 
1437         in6_dev = __in6_dev_get(dev);
1438         if (!in6_dev)
1439                 goto out;
1440 
1441         if (!pskb_may_pull(skb, skb->len))
1442                 goto out;
1443 
1444         iphdr = ipv6_hdr(skb);
1445         saddr = &iphdr->saddr;
1446         daddr = &iphdr->daddr;
1447 
1448         msg = (struct nd_msg *)skb_transport_header(skb);
1449         if (msg->icmph.icmp6_code != 0 ||
1450             msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1451                 goto out;
1452 
1453         if (ipv6_addr_loopback(daddr) ||
1454             ipv6_addr_is_multicast(&msg->target))
1455                 goto out;
1456 
1457         n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
1458 
1459         if (n) {
1460                 struct vxlan_fdb *f;
1461                 struct sk_buff *reply;
1462 
1463                 if (!(n->nud_state & NUD_CONNECTED)) {
1464                         neigh_release(n);
1465                         goto out;
1466                 }
1467 
1468                 f = vxlan_find_mac(vxlan, n->ha);
1469                 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1470                         /* bridge-local neighbor */
1471                         neigh_release(n);
1472                         goto out;
1473                 }
1474 
1475                 reply = vxlan_na_create(skb, n,
1476                                         !!(f ? f->flags & NTF_ROUTER : 0));
1477 
1478                 neigh_release(n);
1479 
1480                 if (reply == NULL)
1481                         goto out;
1482 
1483                 if (netif_rx_ni(reply) == NET_RX_DROP)
1484                         dev->stats.rx_dropped++;
1485 
1486         } else if (vxlan->flags & VXLAN_F_L3MISS) {
1487                 union vxlan_addr ipa = {
1488                         .sin6.sin6_addr = msg->target,
1489                         .sa.sa_family = AF_INET6,
1490                 };
1491 
1492                 vxlan_ip_miss(dev, &ipa);
1493         }
1494 
1495 out:
1496         consume_skb(skb);
1497         return NETDEV_TX_OK;
1498 }
1499 #endif
1500 
1501 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1502 {
1503         struct vxlan_dev *vxlan = netdev_priv(dev);
1504         struct neighbour *n;
1505 
1506         if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1507                 return false;
1508 
1509         n = NULL;
1510         switch (ntohs(eth_hdr(skb)->h_proto)) {
1511         case ETH_P_IP:
1512         {
1513                 struct iphdr *pip;
1514 
1515                 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1516                         return false;
1517                 pip = ip_hdr(skb);
1518                 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1519                 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1520                         union vxlan_addr ipa = {
1521                                 .sin.sin_addr.s_addr = pip->daddr,
1522                                 .sa.sa_family = AF_INET,
1523                         };
1524 
1525                         vxlan_ip_miss(dev, &ipa);
1526                         return false;
1527                 }
1528 
1529                 break;
1530         }
1531 #if IS_ENABLED(CONFIG_IPV6)
1532         case ETH_P_IPV6:
1533         {
1534                 struct ipv6hdr *pip6;
1535 
1536                 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1537                         return false;
1538                 pip6 = ipv6_hdr(skb);
1539                 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1540                 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1541                         union vxlan_addr ipa = {
1542                                 .sin6.sin6_addr = pip6->daddr,
1543                                 .sa.sa_family = AF_INET6,
1544                         };
1545 
1546                         vxlan_ip_miss(dev, &ipa);
1547                         return false;
1548                 }
1549 
1550                 break;
1551         }
1552 #endif
1553         default:
1554                 return false;
1555         }
1556 
1557         if (n) {
1558                 bool diff;
1559 
1560                 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1561                 if (diff) {
1562                         memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1563                                 dev->addr_len);
1564                         memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1565                 }
1566                 neigh_release(n);
1567                 return diff;
1568         }
1569 
1570         return false;
1571 }
1572 
1573 /* Compute source port for outgoing packet
1574  *   first choice to use L4 flow hash since it will spread
1575  *     better and maybe available from hardware
1576  *   secondary choice is to use jhash on the Ethernet header
1577  */
1578 __be16 vxlan_src_port(__u16 port_min, __u16 port_max, struct sk_buff *skb)
1579 {
1580         unsigned int range = (port_max - port_min) + 1;
1581         u32 hash;
1582 
1583         hash = skb_get_hash(skb);
1584         if (!hash)
1585                 hash = jhash(skb->data, 2 * ETH_ALEN,
1586                              (__force u32) skb->protocol);
1587 
1588         return htons((((u64) hash * range) >> 32) + port_min);
1589 }
1590 EXPORT_SYMBOL_GPL(vxlan_src_port);
1591 
1592 static inline struct sk_buff *vxlan_handle_offloads(struct sk_buff *skb,
1593                                                     bool udp_csum)
1594 {
1595         int type = udp_csum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1596         return iptunnel_handle_offloads(skb, udp_csum, type);
1597 }
1598 
1599 #if IS_ENABLED(CONFIG_IPV6)
1600 static int vxlan6_xmit_skb(struct vxlan_sock *vs,
1601                            struct dst_entry *dst, struct sk_buff *skb,
1602                            struct net_device *dev, struct in6_addr *saddr,
1603                            struct in6_addr *daddr, __u8 prio, __u8 ttl,
1604                            __be16 src_port, __be16 dst_port, __be32 vni,
1605                            bool xnet)
1606 {
1607         struct ipv6hdr *ip6h;
1608         struct vxlanhdr *vxh;
1609         struct udphdr *uh;
1610         int min_headroom;
1611         int err;
1612 
1613         skb = vxlan_handle_offloads(skb, !udp_get_no_check6_tx(vs->sock->sk));
1614         if (IS_ERR(skb))
1615                 return -EINVAL;
1616 
1617         skb_scrub_packet(skb, xnet);
1618 
1619         min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1620                         + VXLAN_HLEN + sizeof(struct ipv6hdr)
1621                         + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1622 
1623         /* Need space for new headers (invalidates iph ptr) */
1624         err = skb_cow_head(skb, min_headroom);
1625         if (unlikely(err))
1626                 return err;
1627 
1628         if (vlan_tx_tag_present(skb)) {
1629                 if (WARN_ON(!__vlan_put_tag(skb,
1630                                             skb->vlan_proto,
1631                                             vlan_tx_tag_get(skb))))
1632                         return -ENOMEM;
1633 
1634                 skb->vlan_tci = 0;
1635         }
1636 
1637         vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1638         vxh->vx_flags = htonl(VXLAN_FLAGS);
1639         vxh->vx_vni = vni;
1640 
1641         __skb_push(skb, sizeof(*uh));
1642         skb_reset_transport_header(skb);
1643         uh = udp_hdr(skb);
1644 
1645         uh->dest = dst_port;
1646         uh->source = src_port;
1647 
1648         uh->len = htons(skb->len);
1649 
1650         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1651         IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1652                               IPSKB_REROUTED);
1653         skb_dst_set(skb, dst);
1654 
1655         udp6_set_csum(udp_get_no_check6_tx(vs->sock->sk), skb,
1656                       saddr, daddr, skb->len);
1657 
1658         __skb_push(skb, sizeof(*ip6h));
1659         skb_reset_network_header(skb);
1660         ip6h              = ipv6_hdr(skb);
1661         ip6h->version     = 6;
1662         ip6h->priority    = prio;
1663         ip6h->flow_lbl[0] = 0;
1664         ip6h->flow_lbl[1] = 0;
1665         ip6h->flow_lbl[2] = 0;
1666         ip6h->payload_len = htons(skb->len);
1667         ip6h->nexthdr     = IPPROTO_UDP;
1668         ip6h->hop_limit   = ttl;
1669         ip6h->daddr       = *daddr;
1670         ip6h->saddr       = *saddr;
1671 
1672         ip6tunnel_xmit(skb, dev);
1673         return 0;
1674 }
1675 #endif
1676 
1677 int vxlan_xmit_skb(struct vxlan_sock *vs,
1678                    struct rtable *rt, struct sk_buff *skb,
1679                    __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
1680                    __be16 src_port, __be16 dst_port, __be32 vni, bool xnet)
1681 {
1682         struct vxlanhdr *vxh;
1683         struct udphdr *uh;
1684         int min_headroom;
1685         int err;
1686 
1687         skb = vxlan_handle_offloads(skb, !vs->sock->sk->sk_no_check_tx);
1688         if (IS_ERR(skb))
1689                 return -EINVAL;
1690 
1691         min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1692                         + VXLAN_HLEN + sizeof(struct iphdr)
1693                         + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1694 
1695         /* Need space for new headers (invalidates iph ptr) */
1696         err = skb_cow_head(skb, min_headroom);
1697         if (unlikely(err))
1698                 return err;
1699 
1700         if (vlan_tx_tag_present(skb)) {
1701                 if (WARN_ON(!__vlan_put_tag(skb,
1702                                             skb->vlan_proto,
1703                                             vlan_tx_tag_get(skb))))
1704                         return -ENOMEM;
1705 
1706                 skb->vlan_tci = 0;
1707         }
1708 
1709         vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1710         vxh->vx_flags = htonl(VXLAN_FLAGS);
1711         vxh->vx_vni = vni;
1712 
1713         __skb_push(skb, sizeof(*uh));
1714         skb_reset_transport_header(skb);
1715         uh = udp_hdr(skb);
1716 
1717         uh->dest = dst_port;
1718         uh->source = src_port;
1719 
1720         uh->len = htons(skb->len);
1721 
1722         udp_set_csum(vs->sock->sk->sk_no_check_tx, skb,
1723                      src, dst, skb->len);
1724 
1725         return iptunnel_xmit(vs->sock->sk, rt, skb, src, dst, IPPROTO_UDP,
1726                              tos, ttl, df, xnet);
1727 }
1728 EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1729 
1730 /* Bypass encapsulation if the destination is local */
1731 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1732                                struct vxlan_dev *dst_vxlan)
1733 {
1734         struct pcpu_sw_netstats *tx_stats, *rx_stats;
1735         union vxlan_addr loopback;
1736         union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
1737 
1738         tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1739         rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1740         skb->pkt_type = PACKET_HOST;
1741         skb->encapsulation = 0;
1742         skb->dev = dst_vxlan->dev;
1743         __skb_pull(skb, skb_network_offset(skb));
1744 
1745         if (remote_ip->sa.sa_family == AF_INET) {
1746                 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1747                 loopback.sa.sa_family =  AF_INET;
1748 #if IS_ENABLED(CONFIG_IPV6)
1749         } else {
1750                 loopback.sin6.sin6_addr = in6addr_loopback;
1751                 loopback.sa.sa_family =  AF_INET6;
1752 #endif
1753         }
1754 
1755         if (dst_vxlan->flags & VXLAN_F_LEARN)
1756                 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
1757 
1758         u64_stats_update_begin(&tx_stats->syncp);
1759         tx_stats->tx_packets++;
1760         tx_stats->tx_bytes += skb->len;
1761         u64_stats_update_end(&tx_stats->syncp);
1762 
1763         if (netif_rx(skb) == NET_RX_SUCCESS) {
1764                 u64_stats_update_begin(&rx_stats->syncp);
1765                 rx_stats->rx_packets++;
1766                 rx_stats->rx_bytes += skb->len;
1767                 u64_stats_update_end(&rx_stats->syncp);
1768         } else {
1769                 skb->dev->stats.rx_dropped++;
1770         }
1771 }
1772 
1773 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1774                            struct vxlan_rdst *rdst, bool did_rsc)
1775 {
1776         struct vxlan_dev *vxlan = netdev_priv(dev);
1777         struct rtable *rt = NULL;
1778         const struct iphdr *old_iph;
1779         struct flowi4 fl4;
1780         union vxlan_addr *dst;
1781         __be16 src_port = 0, dst_port;
1782         u32 vni;
1783         __be16 df = 0;
1784         __u8 tos, ttl;
1785         int err;
1786 
1787         dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
1788         vni = rdst->remote_vni;
1789         dst = &rdst->remote_ip;
1790 
1791         if (vxlan_addr_any(dst)) {
1792                 if (did_rsc) {
1793                         /* short-circuited back to local bridge */
1794                         vxlan_encap_bypass(skb, vxlan, vxlan);
1795                         return;
1796                 }
1797                 goto drop;
1798         }
1799 
1800         old_iph = ip_hdr(skb);
1801 
1802         ttl = vxlan->ttl;
1803         if (!ttl && vxlan_addr_multicast(dst))
1804                 ttl = 1;
1805 
1806         tos = vxlan->tos;
1807         if (tos == 1)
1808                 tos = ip_tunnel_get_dsfield(old_iph, skb);
1809 
1810         src_port = vxlan_src_port(vxlan->port_min, vxlan->port_max, skb);
1811 
1812         if (dst->sa.sa_family == AF_INET) {
1813                 memset(&fl4, 0, sizeof(fl4));
1814                 fl4.flowi4_oif = rdst->remote_ifindex;
1815                 fl4.flowi4_tos = RT_TOS(tos);
1816                 fl4.daddr = dst->sin.sin_addr.s_addr;
1817                 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1818 
1819                 rt = ip_route_output_key(vxlan->net, &fl4);
1820                 if (IS_ERR(rt)) {
1821                         netdev_dbg(dev, "no route to %pI4\n",
1822                                    &dst->sin.sin_addr.s_addr);
1823                         dev->stats.tx_carrier_errors++;
1824                         goto tx_error;
1825                 }
1826 
1827                 if (rt->dst.dev == dev) {
1828                         netdev_dbg(dev, "circular route to %pI4\n",
1829                                    &dst->sin.sin_addr.s_addr);
1830                         dev->stats.collisions++;
1831                         goto rt_tx_error;
1832                 }
1833 
1834                 /* Bypass encapsulation if the destination is local */
1835                 if (rt->rt_flags & RTCF_LOCAL &&
1836                     !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1837                         struct vxlan_dev *dst_vxlan;
1838 
1839                         ip_rt_put(rt);
1840                         dst_vxlan = vxlan_find_vni(vxlan->net, vni, dst_port);
1841                         if (!dst_vxlan)
1842                                 goto tx_error;
1843                         vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1844                         return;
1845                 }
1846 
1847                 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1848                 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
1849 
1850                 err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
1851                                      fl4.saddr, dst->sin.sin_addr.s_addr,
1852                                      tos, ttl, df, src_port, dst_port,
1853                                      htonl(vni << 8),
1854                                      !net_eq(vxlan->net, dev_net(vxlan->dev)));
1855 
1856                 if (err < 0)
1857                         goto rt_tx_error;
1858                 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1859 #if IS_ENABLED(CONFIG_IPV6)
1860         } else {
1861                 struct sock *sk = vxlan->vn_sock->sock->sk;
1862                 struct dst_entry *ndst;
1863                 struct flowi6 fl6;
1864                 u32 flags;
1865 
1866                 memset(&fl6, 0, sizeof(fl6));
1867                 fl6.flowi6_oif = rdst->remote_ifindex;
1868                 fl6.daddr = dst->sin6.sin6_addr;
1869                 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
1870                 fl6.flowi6_proto = IPPROTO_UDP;
1871 
1872                 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1873                         netdev_dbg(dev, "no route to %pI6\n",
1874                                    &dst->sin6.sin6_addr);
1875                         dev->stats.tx_carrier_errors++;
1876                         goto tx_error;
1877                 }
1878 
1879                 if (ndst->dev == dev) {
1880                         netdev_dbg(dev, "circular route to %pI6\n",
1881                                    &dst->sin6.sin6_addr);
1882                         dst_release(ndst);
1883                         dev->stats.collisions++;
1884                         goto tx_error;
1885                 }
1886 
1887                 /* Bypass encapsulation if the destination is local */
1888                 flags = ((struct rt6_info *)ndst)->rt6i_flags;
1889                 if (flags & RTF_LOCAL &&
1890                     !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1891                         struct vxlan_dev *dst_vxlan;
1892 
1893                         dst_release(ndst);
1894                         dst_vxlan = vxlan_find_vni(vxlan->net, vni, dst_port);
1895                         if (!dst_vxlan)
1896                                 goto tx_error;
1897                         vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1898                         return;
1899                 }
1900 
1901                 ttl = ttl ? : ip6_dst_hoplimit(ndst);
1902 
1903                 err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
1904                                       dev, &fl6.saddr, &fl6.daddr, 0, ttl,
1905                                       src_port, dst_port, htonl(vni << 8),
1906                                       !net_eq(vxlan->net, dev_net(vxlan->dev)));
1907 #endif
1908         }
1909 
1910         return;
1911 
1912 drop:
1913         dev->stats.tx_dropped++;
1914         goto tx_free;
1915 
1916 rt_tx_error:
1917         ip_rt_put(rt);
1918 tx_error:
1919         dev->stats.tx_errors++;
1920 tx_free:
1921         dev_kfree_skb(skb);
1922 }
1923 
1924 /* Transmit local packets over Vxlan
1925  *
1926  * Outer IP header inherits ECN and DF from inner header.
1927  * Outer UDP destination is the VXLAN assigned port.
1928  *           source port is based on hash of flow
1929  */
1930 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1931 {
1932         struct vxlan_dev *vxlan = netdev_priv(dev);
1933         struct ethhdr *eth;
1934         bool did_rsc = false;
1935         struct vxlan_rdst *rdst, *fdst = NULL;
1936         struct vxlan_fdb *f;
1937 
1938         skb_reset_mac_header(skb);
1939         eth = eth_hdr(skb);
1940 
1941         if ((vxlan->flags & VXLAN_F_PROXY)) {
1942                 if (ntohs(eth->h_proto) == ETH_P_ARP)
1943                         return arp_reduce(dev, skb);
1944 #if IS_ENABLED(CONFIG_IPV6)
1945                 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
1946                          skb->len >= sizeof(struct ipv6hdr) + sizeof(struct nd_msg) &&
1947                          ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
1948                                 struct nd_msg *msg;
1949 
1950                                 msg = (struct nd_msg *)skb_transport_header(skb);
1951                                 if (msg->icmph.icmp6_code == 0 &&
1952                                     msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
1953                                         return neigh_reduce(dev, skb);
1954                 }
1955 #endif
1956         }
1957 
1958         f = vxlan_find_mac(vxlan, eth->h_dest);
1959         did_rsc = false;
1960 
1961         if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
1962             (ntohs(eth->h_proto) == ETH_P_IP ||
1963              ntohs(eth->h_proto) == ETH_P_IPV6)) {
1964                 did_rsc = route_shortcircuit(dev, skb);
1965                 if (did_rsc)
1966                         f = vxlan_find_mac(vxlan, eth->h_dest);
1967         }
1968 
1969         if (f == NULL) {
1970                 f = vxlan_find_mac(vxlan, all_zeros_mac);
1971                 if (f == NULL) {
1972                         if ((vxlan->flags & VXLAN_F_L2MISS) &&
1973                             !is_multicast_ether_addr(eth->h_dest))
1974                                 vxlan_fdb_miss(vxlan, eth->h_dest);
1975 
1976                         dev->stats.tx_dropped++;
1977                         kfree_skb(skb);
1978                         return NETDEV_TX_OK;
1979                 }
1980         }
1981 
1982         list_for_each_entry_rcu(rdst, &f->remotes, list) {
1983                 struct sk_buff *skb1;
1984 
1985                 if (!fdst) {
1986                         fdst = rdst;
1987                         continue;
1988                 }
1989                 skb1 = skb_clone(skb, GFP_ATOMIC);
1990                 if (skb1)
1991                         vxlan_xmit_one(skb1, dev, rdst, did_rsc);
1992         }
1993 
1994         if (fdst)
1995                 vxlan_xmit_one(skb, dev, fdst, did_rsc);
1996         else
1997                 kfree_skb(skb);
1998         return NETDEV_TX_OK;
1999 }
2000 
2001 /* Walk the forwarding table and purge stale entries */
2002 static void vxlan_cleanup(unsigned long arg)
2003 {
2004         struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2005         unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2006         unsigned int h;
2007 
2008         if (!netif_running(vxlan->dev))
2009                 return;
2010 
2011         spin_lock_bh(&vxlan->hash_lock);
2012         for (h = 0; h < FDB_HASH_SIZE; ++h) {
2013                 struct hlist_node *p, *n;
2014                 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2015                         struct vxlan_fdb *f
2016                                 = container_of(p, struct vxlan_fdb, hlist);
2017                         unsigned long timeout;
2018 
2019                         if (f->state & NUD_PERMANENT)
2020                                 continue;
2021 
2022                         timeout = f->used + vxlan->age_interval * HZ;
2023                         if (time_before_eq(timeout, jiffies)) {
2024                                 netdev_dbg(vxlan->dev,
2025                                            "garbage collect %pM\n",
2026                                            f->eth_addr);
2027                                 f->state = NUD_STALE;
2028                                 vxlan_fdb_destroy(vxlan, f);
2029                         } else if (time_before(timeout, next_timer))
2030                                 next_timer = timeout;
2031                 }
2032         }
2033         spin_unlock_bh(&vxlan->hash_lock);
2034 
2035         mod_timer(&vxlan->age_timer, next_timer);
2036 }
2037 
2038 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2039 {
2040         __u32 vni = vxlan->default_dst.remote_vni;
2041 
2042         vxlan->vn_sock = vs;
2043         hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
2044 }
2045 
2046 /* Setup stats when device is created */
2047 static int vxlan_init(struct net_device *dev)
2048 {
2049         struct vxlan_dev *vxlan = netdev_priv(dev);
2050         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2051         struct vxlan_sock *vs;
2052 
2053         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2054         if (!dev->tstats)
2055                 return -ENOMEM;
2056 
2057         spin_lock(&vn->sock_lock);
2058         vs = vxlan_find_sock(vxlan->net, vxlan->dst_port);
2059         if (vs) {
2060                 /* If we have a socket with same port already, reuse it */
2061                 atomic_inc(&vs->refcnt);
2062                 vxlan_vs_add_dev(vs, vxlan);
2063         } else {
2064                 /* otherwise make new socket outside of RTNL */
2065                 dev_hold(dev);
2066                 queue_work(vxlan_wq, &vxlan->sock_work);
2067         }
2068         spin_unlock(&vn->sock_lock);
2069 
2070         return 0;
2071 }
2072 
2073 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
2074 {
2075         struct vxlan_fdb *f;
2076 
2077         spin_lock_bh(&vxlan->hash_lock);
2078         f = __vxlan_find_mac(vxlan, all_zeros_mac);
2079         if (f)
2080                 vxlan_fdb_destroy(vxlan, f);
2081         spin_unlock_bh(&vxlan->hash_lock);
2082 }
2083 
2084 static void vxlan_uninit(struct net_device *dev)
2085 {
2086         struct vxlan_dev *vxlan = netdev_priv(dev);
2087         struct vxlan_sock *vs = vxlan->vn_sock;
2088 
2089         vxlan_fdb_delete_default(vxlan);
2090 
2091         if (vs)
2092                 vxlan_sock_release(vs);
2093         free_percpu(dev->tstats);
2094 }
2095 
2096 /* Start ageing timer and join group when device is brought up */
2097 static int vxlan_open(struct net_device *dev)
2098 {
2099         struct vxlan_dev *vxlan = netdev_priv(dev);
2100         struct vxlan_sock *vs = vxlan->vn_sock;
2101 
2102         /* socket hasn't been created */
2103         if (!vs)
2104                 return -ENOTCONN;
2105 
2106         if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
2107                 vxlan_sock_hold(vs);
2108                 dev_hold(dev);
2109                 queue_work(vxlan_wq, &vxlan->igmp_join);
2110         }
2111 
2112         if (vxlan->age_interval)
2113                 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2114 
2115         return 0;
2116 }
2117 
2118 /* Purge the forwarding table */
2119 static void vxlan_flush(struct vxlan_dev *vxlan)
2120 {
2121         unsigned int h;
2122 
2123         spin_lock_bh(&vxlan->hash_lock);
2124         for (h = 0; h < FDB_HASH_SIZE; ++h) {
2125                 struct hlist_node *p, *n;
2126                 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2127                         struct vxlan_fdb *f
2128                                 = container_of(p, struct vxlan_fdb, hlist);
2129                         /* the all_zeros_mac entry is deleted at vxlan_uninit */
2130                         if (!is_zero_ether_addr(f->eth_addr))
2131                                 vxlan_fdb_destroy(vxlan, f);
2132                 }
2133         }
2134         spin_unlock_bh(&vxlan->hash_lock);
2135 }
2136 
2137 /* Cleanup timer and forwarding table on shutdown */
2138 static int vxlan_stop(struct net_device *dev)
2139 {
2140         struct vxlan_dev *vxlan = netdev_priv(dev);
2141         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2142         struct vxlan_sock *vs = vxlan->vn_sock;
2143 
2144         if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2145             !vxlan_group_used(vn, vxlan)) {
2146                 vxlan_sock_hold(vs);
2147                 dev_hold(dev);
2148                 queue_work(vxlan_wq, &vxlan->igmp_leave);
2149         }
2150 
2151         del_timer_sync(&vxlan->age_timer);
2152 
2153         vxlan_flush(vxlan);
2154 
2155         return 0;
2156 }
2157 
2158 /* Stub, nothing needs to be done. */
2159 static void vxlan_set_multicast_list(struct net_device *dev)
2160 {
2161 }
2162 
2163 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2164 {
2165         struct vxlan_dev *vxlan = netdev_priv(dev);
2166         struct vxlan_rdst *dst = &vxlan->default_dst;
2167         struct net_device *lowerdev;
2168         int max_mtu;
2169 
2170         lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
2171         if (lowerdev == NULL)
2172                 return eth_change_mtu(dev, new_mtu);
2173 
2174         if (dst->remote_ip.sa.sa_family == AF_INET6)
2175                 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2176         else
2177                 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2178 
2179         if (new_mtu < 68 || new_mtu > max_mtu)
2180                 return -EINVAL;
2181 
2182         dev->mtu = new_mtu;
2183         return 0;
2184 }
2185 
2186 static const struct net_device_ops vxlan_netdev_ops = {
2187         .ndo_init               = vxlan_init,
2188         .ndo_uninit             = vxlan_uninit,
2189         .ndo_open               = vxlan_open,
2190         .ndo_stop               = vxlan_stop,
2191         .ndo_start_xmit         = vxlan_xmit,
2192         .ndo_get_stats64        = ip_tunnel_get_stats64,
2193         .ndo_set_rx_mode        = vxlan_set_multicast_list,
2194         .ndo_change_mtu         = vxlan_change_mtu,
2195         .ndo_validate_addr      = eth_validate_addr,
2196         .ndo_set_mac_address    = eth_mac_addr,
2197         .ndo_fdb_add            = vxlan_fdb_add,
2198         .ndo_fdb_del            = vxlan_fdb_delete,
2199         .ndo_fdb_dump           = vxlan_fdb_dump,
2200 };
2201 
2202 /* Info for udev, that this is a virtual tunnel endpoint */
2203 static struct device_type vxlan_type = {
2204         .name = "vxlan",
2205 };
2206 
2207 /* Calls the ndo_add_vxlan_port of the caller in order to
2208  * supply the listening VXLAN udp ports. Callers are expected
2209  * to implement the ndo_add_vxlan_port.
2210  */
2211 void vxlan_get_rx_port(struct net_device *dev)
2212 {
2213         struct vxlan_sock *vs;
2214         struct net *net = dev_net(dev);
2215         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2216         sa_family_t sa_family;
2217         __be16 port;
2218         unsigned int i;
2219 
2220         spin_lock(&vn->sock_lock);
2221         for (i = 0; i < PORT_HASH_SIZE; ++i) {
2222                 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2223                         port = inet_sk(vs->sock->sk)->inet_sport;
2224                         sa_family = vs->sock->sk->sk_family;
2225                         dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2226                                                             port);
2227                 }
2228         }
2229         spin_unlock(&vn->sock_lock);
2230 }
2231 EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2232 
2233 /* Initialize the device structure. */
2234 static void vxlan_setup(struct net_device *dev)
2235 {
2236         struct vxlan_dev *vxlan = netdev_priv(dev);
2237         unsigned int h;
2238         int low, high;
2239 
2240         eth_hw_addr_random(dev);
2241         ether_setup(dev);
2242         if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2243                 dev->needed_headroom = ETH_HLEN + VXLAN6_HEADROOM;
2244         else
2245                 dev->needed_headroom = ETH_HLEN + VXLAN_HEADROOM;
2246 
2247         dev->netdev_ops = &vxlan_netdev_ops;
2248         dev->destructor = free_netdev;
2249         SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2250 
2251         dev->tx_queue_len = 0;
2252         dev->features   |= NETIF_F_LLTX;
2253         dev->features   |= NETIF_F_SG | NETIF_F_HW_CSUM;
2254         dev->features   |= NETIF_F_RXCSUM;
2255         dev->features   |= NETIF_F_GSO_SOFTWARE;
2256 
2257         dev->vlan_features = dev->features;
2258         dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2259         dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2260         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2261         dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2262         dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
2263         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2264 
2265         INIT_LIST_HEAD(&vxlan->next);
2266         spin_lock_init(&vxlan->hash_lock);
2267         INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2268         INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
2269         INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
2270 
2271         init_timer_deferrable(&vxlan->age_timer);
2272         vxlan->age_timer.function = vxlan_cleanup;
2273         vxlan->age_timer.data = (unsigned long) vxlan;
2274 
2275         inet_get_local_port_range(dev_net(dev), &low, &high);
2276         vxlan->port_min = low;
2277         vxlan->port_max = high;
2278         vxlan->dst_port = htons(vxlan_port);
2279 
2280         vxlan->dev = dev;
2281 
2282         for (h = 0; h < FDB_HASH_SIZE; ++h)
2283                 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2284 }
2285 
2286 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2287         [IFLA_VXLAN_ID]         = { .type = NLA_U32 },
2288         [IFLA_VXLAN_GROUP]      = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2289         [IFLA_VXLAN_GROUP6]     = { .len = sizeof(struct in6_addr) },
2290         [IFLA_VXLAN_LINK]       = { .type = NLA_U32 },
2291         [IFLA_VXLAN_LOCAL]      = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2292         [IFLA_VXLAN_LOCAL6]     = { .len = sizeof(struct in6_addr) },
2293         [IFLA_VXLAN_TOS]        = { .type = NLA_U8 },
2294         [IFLA_VXLAN_TTL]        = { .type = NLA_U8 },
2295         [IFLA_VXLAN_LEARNING]   = { .type = NLA_U8 },
2296         [IFLA_VXLAN_AGEING]     = { .type = NLA_U32 },
2297         [IFLA_VXLAN_LIMIT]      = { .type = NLA_U32 },
2298         [IFLA_VXLAN_PORT_RANGE] = { .len  = sizeof(struct ifla_vxlan_port_range) },
2299         [IFLA_VXLAN_PROXY]      = { .type = NLA_U8 },
2300         [IFLA_VXLAN_RSC]        = { .type = NLA_U8 },
2301         [IFLA_VXLAN_L2MISS]     = { .type = NLA_U8 },
2302         [IFLA_VXLAN_L3MISS]     = { .type = NLA_U8 },
2303         [IFLA_VXLAN_PORT]       = { .type = NLA_U16 },
2304 };
2305 
2306 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2307 {
2308         if (tb[IFLA_ADDRESS]) {
2309                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2310                         pr_debug("invalid link address (not ethernet)\n");
2311                         return -EINVAL;
2312                 }
2313 
2314                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2315                         pr_debug("invalid all zero ethernet address\n");
2316                         return -EADDRNOTAVAIL;
2317                 }
2318         }
2319 
2320         if (!data)
2321                 return -EINVAL;
2322 
2323         if (data[IFLA_VXLAN_ID]) {
2324                 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2325                 if (id >= VXLAN_VID_MASK)
2326                         return -ERANGE;
2327         }
2328 
2329         if (data[IFLA_VXLAN_PORT_RANGE]) {
2330                 const struct ifla_vxlan_port_range *p
2331                         = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2332 
2333                 if (ntohs(p->high) < ntohs(p->low)) {
2334                         pr_debug("port range %u .. %u not valid\n",
2335                                  ntohs(p->low), ntohs(p->high));
2336                         return -EINVAL;
2337                 }
2338         }
2339 
2340         return 0;
2341 }
2342 
2343 static void vxlan_get_drvinfo(struct net_device *netdev,
2344                               struct ethtool_drvinfo *drvinfo)
2345 {
2346         strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2347         strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2348 }
2349 
2350 static const struct ethtool_ops vxlan_ethtool_ops = {
2351         .get_drvinfo    = vxlan_get_drvinfo,
2352         .get_link       = ethtool_op_get_link,
2353 };
2354 
2355 static void vxlan_del_work(struct work_struct *work)
2356 {
2357         struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2358 
2359         sk_release_kernel(vs->sock->sk);
2360         kfree_rcu(vs, rcu);
2361 }
2362 
2363 #if IS_ENABLED(CONFIG_IPV6)
2364 /* Create UDP socket for encapsulation receive. AF_INET6 socket
2365  * could be used for both IPv4 and IPv6 communications, but
2366  * users may set bindv6only=1.
2367  */
2368 static struct socket *create_v6_sock(struct net *net, __be16 port, u32 flags)
2369 {
2370         struct sock *sk;
2371         struct socket *sock;
2372         struct sockaddr_in6 vxlan_addr = {
2373                 .sin6_family = AF_INET6,
2374                 .sin6_port = port,
2375         };
2376         int rc, val = 1;
2377 
2378         rc = sock_create_kern(AF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
2379         if (rc < 0) {
2380                 pr_debug("UDPv6 socket create failed\n");
2381                 return ERR_PTR(rc);
2382         }
2383 
2384         /* Put in proper namespace */
2385         sk = sock->sk;
2386         sk_change_net(sk, net);
2387 
2388         kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
2389                           (char *)&val, sizeof(val));
2390         rc = kernel_bind(sock, (struct sockaddr *)&vxlan_addr,
2391                          sizeof(struct sockaddr_in6));
2392         if (rc < 0) {
2393                 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2394                          &vxlan_addr.sin6_addr, ntohs(vxlan_addr.sin6_port), rc);
2395                 sk_release_kernel(sk);
2396                 return ERR_PTR(rc);
2397         }
2398         /* At this point, IPv6 module should have been loaded in
2399          * sock_create_kern().
2400          */
2401         BUG_ON(!ipv6_stub);
2402 
2403         /* Disable multicast loopback */
2404         inet_sk(sk)->mc_loop = 0;
2405 
2406         if (flags & VXLAN_F_UDP_ZERO_CSUM6_TX)
2407                 udp_set_no_check6_tx(sk, true);
2408 
2409         if (flags & VXLAN_F_UDP_ZERO_CSUM6_RX)
2410                 udp_set_no_check6_rx(sk, true);
2411 
2412         return sock;
2413 }
2414 
2415 #else
2416 
2417 static struct socket *create_v6_sock(struct net *net, __be16 port, u32 flags)
2418 {
2419                 return ERR_PTR(-EPFNOSUPPORT);
2420 }
2421 #endif
2422 
2423 static struct socket *create_v4_sock(struct net *net, __be16 port, u32 flags)
2424 {
2425         struct sock *sk;
2426         struct socket *sock;
2427         struct sockaddr_in vxlan_addr = {
2428                 .sin_family = AF_INET,
2429                 .sin_addr.s_addr = htonl(INADDR_ANY),
2430                 .sin_port = port,
2431         };
2432         int rc;
2433 
2434         /* Create UDP socket for encapsulation receive. */
2435         rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
2436         if (rc < 0) {
2437                 pr_debug("UDP socket create failed\n");
2438                 return ERR_PTR(rc);
2439         }
2440 
2441         /* Put in proper namespace */
2442         sk = sock->sk;
2443         sk_change_net(sk, net);
2444 
2445         rc = kernel_bind(sock, (struct sockaddr *) &vxlan_addr,
2446                          sizeof(vxlan_addr));
2447         if (rc < 0) {
2448                 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2449                          &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
2450                 sk_release_kernel(sk);
2451                 return ERR_PTR(rc);
2452         }
2453 
2454         /* Disable multicast loopback */
2455         inet_sk(sk)->mc_loop = 0;
2456 
2457         if (!(flags & VXLAN_F_UDP_CSUM))
2458                 sock->sk->sk_no_check_tx = 1;
2459 
2460         return sock;
2461 }
2462 
2463 /* Create new listen socket if needed */
2464 static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
2465                                               vxlan_rcv_t *rcv, void *data,
2466                                               u32 flags)
2467 {
2468         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2469         struct vxlan_sock *vs;
2470         struct socket *sock;
2471         struct sock *sk;
2472         unsigned int h;
2473         bool ipv6 = !!(flags & VXLAN_F_IPV6);
2474 
2475         vs = kzalloc(sizeof(*vs), GFP_KERNEL);
2476         if (!vs)
2477                 return ERR_PTR(-ENOMEM);
2478 
2479         for (h = 0; h < VNI_HASH_SIZE; ++h)
2480                 INIT_HLIST_HEAD(&vs->vni_list[h]);
2481 
2482         INIT_WORK(&vs->del_work, vxlan_del_work);
2483 
2484         if (ipv6)
2485                 sock = create_v6_sock(net, port, flags);
2486         else
2487                 sock = create_v4_sock(net, port, flags);
2488         if (IS_ERR(sock)) {
2489                 kfree(vs);
2490                 return ERR_CAST(sock);
2491         }
2492 
2493         vs->sock = sock;
2494         sk = sock->sk;
2495         atomic_set(&vs->refcnt, 1);
2496         vs->rcv = rcv;
2497         vs->data = data;
2498         rcu_assign_sk_user_data(vs->sock->sk, vs);
2499 
2500         /* Initialize the vxlan udp offloads structure */
2501         vs->udp_offloads.port = port;
2502         vs->udp_offloads.callbacks.gro_receive  = vxlan_gro_receive;
2503         vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2504 
2505         spin_lock(&vn->sock_lock);
2506         hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2507         vxlan_notify_add_rx_port(vs);
2508         spin_unlock(&vn->sock_lock);
2509 
2510         /* Mark socket as an encapsulation socket. */
2511         udp_sk(sk)->encap_type = 1;
2512         udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
2513 #if IS_ENABLED(CONFIG_IPV6)
2514         if (ipv6)
2515                 ipv6_stub->udpv6_encap_enable();
2516         else
2517 #endif
2518                 udp_encap_enable();
2519 
2520         return vs;
2521 }
2522 
2523 struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2524                                   vxlan_rcv_t *rcv, void *data,
2525                                   bool no_share, u32 flags)
2526 {
2527         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2528         struct vxlan_sock *vs;
2529 
2530         vs = vxlan_socket_create(net, port, rcv, data, flags);
2531         if (!IS_ERR(vs))
2532                 return vs;
2533 
2534         if (no_share)   /* Return error if sharing is not allowed. */
2535                 return vs;
2536 
2537         spin_lock(&vn->sock_lock);
2538         vs = vxlan_find_sock(net, port);
2539         if (vs) {
2540                 if (vs->rcv == rcv)
2541                         atomic_inc(&vs->refcnt);
2542                 else
2543                         vs = ERR_PTR(-EBUSY);
2544         }
2545         spin_unlock(&vn->sock_lock);
2546 
2547         if (!vs)
2548                 vs = ERR_PTR(-EINVAL);
2549 
2550         return vs;
2551 }
2552 EXPORT_SYMBOL_GPL(vxlan_sock_add);
2553 
2554 /* Scheduled at device creation to bind to a socket */
2555 static void vxlan_sock_work(struct work_struct *work)
2556 {
2557         struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
2558         struct net *net = vxlan->net;
2559         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2560         __be16 port = vxlan->dst_port;
2561         struct vxlan_sock *nvs;
2562 
2563         nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags);
2564         spin_lock(&vn->sock_lock);
2565         if (!IS_ERR(nvs))
2566                 vxlan_vs_add_dev(nvs, vxlan);
2567         spin_unlock(&vn->sock_lock);
2568 
2569         dev_put(vxlan->dev);
2570 }
2571 
2572 static int vxlan_newlink(struct net *net, struct net_device *dev,
2573                          struct nlattr *tb[], struct nlattr *data[])
2574 {
2575         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2576         struct vxlan_dev *vxlan = netdev_priv(dev);
2577         struct vxlan_rdst *dst = &vxlan->default_dst;
2578         __u32 vni;
2579         int err;
2580         bool use_ipv6 = false;
2581 
2582         if (!data[IFLA_VXLAN_ID])
2583                 return -EINVAL;
2584 
2585         vxlan->net = dev_net(dev);
2586 
2587         vni = nla_get_u32(data[IFLA_VXLAN_ID]);
2588         dst->remote_vni = vni;
2589 
2590         /* Unless IPv6 is explicitly requested, assume IPv4 */
2591         dst->remote_ip.sa.sa_family = AF_INET;
2592         if (data[IFLA_VXLAN_GROUP]) {
2593                 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
2594         } else if (data[IFLA_VXLAN_GROUP6]) {
2595                 if (!IS_ENABLED(CONFIG_IPV6))
2596                         return -EPFNOSUPPORT;
2597 
2598                 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2599                            sizeof(struct in6_addr));
2600                 dst->remote_ip.sa.sa_family = AF_INET6;
2601                 use_ipv6 = true;
2602         }
2603 
2604         if (data[IFLA_VXLAN_LOCAL]) {
2605                 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2606                 vxlan->saddr.sa.sa_family = AF_INET;
2607         } else if (data[IFLA_VXLAN_LOCAL6]) {
2608                 if (!IS_ENABLED(CONFIG_IPV6))
2609                         return -EPFNOSUPPORT;
2610 
2611                 /* TODO: respect scope id */
2612                 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2613                            sizeof(struct in6_addr));
2614                 vxlan->saddr.sa.sa_family = AF_INET6;
2615                 use_ipv6 = true;
2616         }
2617 
2618         if (data[IFLA_VXLAN_LINK] &&
2619             (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
2620                 struct net_device *lowerdev
2621                          = __dev_get_by_index(net, dst->remote_ifindex);
2622 
2623                 if (!lowerdev) {
2624                         pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
2625                         return -ENODEV;
2626                 }
2627 
2628 #if IS_ENABLED(CONFIG_IPV6)
2629                 if (use_ipv6) {
2630                         struct inet6_dev *idev = __in6_dev_get(lowerdev);
2631                         if (idev && idev->cnf.disable_ipv6) {
2632                                 pr_info("IPv6 is disabled via sysctl\n");
2633                                 return -EPERM;
2634                         }
2635                         vxlan->flags |= VXLAN_F_IPV6;
2636                 }
2637 #endif
2638 
2639                 if (!tb[IFLA_MTU])
2640                         dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2641 
2642                 dev->needed_headroom = lowerdev->hard_header_len +
2643                                        (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2644         } else if (use_ipv6)
2645                 vxlan->flags |= VXLAN_F_IPV6;
2646 
2647         if (data[IFLA_VXLAN_TOS])
2648                 vxlan->tos  = nla_get_u8(data[IFLA_VXLAN_TOS]);
2649 
2650         if (data[IFLA_VXLAN_TTL])
2651                 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2652 
2653         if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
2654                 vxlan->flags |= VXLAN_F_LEARN;
2655 
2656         if (data[IFLA_VXLAN_AGEING])
2657                 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2658         else
2659                 vxlan->age_interval = FDB_AGE_DEFAULT;
2660 
2661         if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2662                 vxlan->flags |= VXLAN_F_PROXY;
2663 
2664         if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2665                 vxlan->flags |= VXLAN_F_RSC;
2666 
2667         if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2668                 vxlan->flags |= VXLAN_F_L2MISS;
2669 
2670         if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2671                 vxlan->flags |= VXLAN_F_L3MISS;
2672 
2673         if (data[IFLA_VXLAN_LIMIT])
2674                 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2675 
2676         if (data[IFLA_VXLAN_PORT_RANGE]) {
2677                 const struct ifla_vxlan_port_range *p
2678                         = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2679                 vxlan->port_min = ntohs(p->low);
2680                 vxlan->port_max = ntohs(p->high);
2681         }
2682 
2683         if (data[IFLA_VXLAN_PORT])
2684                 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2685 
2686         if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2687                 vxlan->flags |= VXLAN_F_UDP_CSUM;
2688 
2689         if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2690             nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2691                 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2692 
2693         if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2694             nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2695                 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2696 
2697         if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
2698                 pr_info("duplicate VNI %u\n", vni);
2699                 return -EEXIST;
2700         }
2701 
2702         dev->ethtool_ops = &vxlan_ethtool_ops;
2703 
2704         /* create an fdb entry for a valid default destination */
2705         if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2706                 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2707                                        &vxlan->default_dst.remote_ip,
2708                                        NUD_REACHABLE|NUD_PERMANENT,
2709                                        NLM_F_EXCL|NLM_F_CREATE,
2710                                        vxlan->dst_port,
2711                                        vxlan->default_dst.remote_vni,
2712                                        vxlan->default_dst.remote_ifindex,
2713                                        NTF_SELF);
2714                 if (err)
2715                         return err;
2716         }
2717 
2718         err = register_netdevice(dev);
2719         if (err) {
2720                 vxlan_fdb_delete_default(vxlan);
2721                 return err;
2722         }
2723 
2724         list_add(&vxlan->next, &vn->vxlan_list);
2725 
2726         return 0;
2727 }
2728 
2729 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2730 {
2731         struct vxlan_dev *vxlan = netdev_priv(dev);
2732         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2733 
2734         spin_lock(&vn->sock_lock);
2735         if (!hlist_unhashed(&vxlan->hlist))
2736                 hlist_del_rcu(&vxlan->hlist);
2737         spin_unlock(&vn->sock_lock);
2738 
2739         list_del(&vxlan->next);
2740         unregister_netdevice_queue(dev, head);
2741 }
2742 
2743 static size_t vxlan_get_size(const struct net_device *dev)
2744 {
2745 
2746         return nla_total_size(sizeof(__u32)) +  /* IFLA_VXLAN_ID */
2747                 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
2748                 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
2749                 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
2750                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_TTL */
2751                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_TOS */
2752                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_LEARNING */
2753                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_PROXY */
2754                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_RSC */
2755                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_L2MISS */
2756                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_L3MISS */
2757                 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2758                 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
2759                 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
2760                 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2761                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2762                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2763                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
2764                 0;
2765 }
2766 
2767 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2768 {
2769         const struct vxlan_dev *vxlan = netdev_priv(dev);
2770         const struct vxlan_rdst *dst = &vxlan->default_dst;
2771         struct ifla_vxlan_port_range ports = {
2772                 .low =  htons(vxlan->port_min),
2773                 .high = htons(vxlan->port_max),
2774         };
2775 
2776         if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
2777                 goto nla_put_failure;
2778 
2779         if (!vxlan_addr_any(&dst->remote_ip)) {
2780                 if (dst->remote_ip.sa.sa_family == AF_INET) {
2781                         if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2782                                          dst->remote_ip.sin.sin_addr.s_addr))
2783                                 goto nla_put_failure;
2784 #if IS_ENABLED(CONFIG_IPV6)
2785                 } else {
2786                         if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2787                                     &dst->remote_ip.sin6.sin6_addr))
2788                                 goto nla_put_failure;
2789 #endif
2790                 }
2791         }
2792 
2793         if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
2794                 goto nla_put_failure;
2795 
2796         if (!vxlan_addr_any(&vxlan->saddr)) {
2797                 if (vxlan->saddr.sa.sa_family == AF_INET) {
2798                         if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2799                                          vxlan->saddr.sin.sin_addr.s_addr))
2800                                 goto nla_put_failure;
2801 #if IS_ENABLED(CONFIG_IPV6)
2802                 } else {
2803                         if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2804                                     &vxlan->saddr.sin6.sin6_addr))
2805                                 goto nla_put_failure;
2806 #endif
2807                 }
2808         }
2809 
2810         if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2811             nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
2812             nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2813                         !!(vxlan->flags & VXLAN_F_LEARN)) ||
2814             nla_put_u8(skb, IFLA_VXLAN_PROXY,
2815                         !!(vxlan->flags & VXLAN_F_PROXY)) ||
2816             nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2817             nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2818                         !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2819             nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2820                         !!(vxlan->flags & VXLAN_F_L3MISS)) ||
2821             nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
2822             nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
2823             nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port) ||
2824             nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
2825                         !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
2826             nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
2827                         !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
2828             nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
2829                         !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)))
2830                 goto nla_put_failure;
2831 
2832         if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2833                 goto nla_put_failure;
2834 
2835         return 0;
2836 
2837 nla_put_failure:
2838         return -EMSGSIZE;
2839 }
2840 
2841 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2842         .kind           = "vxlan",
2843         .maxtype        = IFLA_VXLAN_MAX,
2844         .policy         = vxlan_policy,
2845         .priv_size      = sizeof(struct vxlan_dev),
2846         .setup          = vxlan_setup,
2847         .validate       = vxlan_validate,
2848         .newlink        = vxlan_newlink,
2849         .dellink        = vxlan_dellink,
2850         .get_size       = vxlan_get_size,
2851         .fill_info      = vxlan_fill_info,
2852 };
2853 
2854 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
2855                                              struct net_device *dev)
2856 {
2857         struct vxlan_dev *vxlan, *next;
2858         LIST_HEAD(list_kill);
2859 
2860         list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2861                 struct vxlan_rdst *dst = &vxlan->default_dst;
2862 
2863                 /* In case we created vxlan device with carrier
2864                  * and we loose the carrier due to module unload
2865                  * we also need to remove vxlan device. In other
2866                  * cases, it's not necessary and remote_ifindex
2867                  * is 0 here, so no matches.
2868                  */
2869                 if (dst->remote_ifindex == dev->ifindex)
2870                         vxlan_dellink(vxlan->dev, &list_kill);
2871         }
2872 
2873         unregister_netdevice_many(&list_kill);
2874 }
2875 
2876 static int vxlan_lowerdev_event(struct notifier_block *unused,
2877                                 unsigned long event, void *ptr)
2878 {
2879         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2880         struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
2881 
2882         if (event == NETDEV_UNREGISTER)
2883                 vxlan_handle_lowerdev_unregister(vn, dev);
2884 
2885         return NOTIFY_DONE;
2886 }
2887 
2888 static struct notifier_block vxlan_notifier_block __read_mostly = {
2889         .notifier_call = vxlan_lowerdev_event,
2890 };
2891 
2892 static __net_init int vxlan_init_net(struct net *net)
2893 {
2894         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2895         unsigned int h;
2896 
2897         INIT_LIST_HEAD(&vn->vxlan_list);
2898         spin_lock_init(&vn->sock_lock);
2899 
2900         for (h = 0; h < PORT_HASH_SIZE; ++h)
2901                 INIT_HLIST_HEAD(&vn->sock_list[h]);
2902 
2903         return 0;
2904 }
2905 
2906 static void __net_exit vxlan_exit_net(struct net *net)
2907 {
2908         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2909         struct vxlan_dev *vxlan, *next;
2910         struct net_device *dev, *aux;
2911         LIST_HEAD(list);
2912 
2913         rtnl_lock();
2914         for_each_netdev_safe(net, dev, aux)
2915                 if (dev->rtnl_link_ops == &vxlan_link_ops)
2916                         unregister_netdevice_queue(dev, &list);
2917 
2918         list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2919                 /* If vxlan->dev is in the same netns, it has already been added
2920                  * to the list by the previous loop.
2921                  */
2922                 if (!net_eq(dev_net(vxlan->dev), net))
2923                         unregister_netdevice_queue(dev, &list);
2924         }
2925 
2926         unregister_netdevice_many(&list);
2927         rtnl_unlock();
2928 }
2929 
2930 static struct pernet_operations vxlan_net_ops = {
2931         .init = vxlan_init_net,
2932         .exit = vxlan_exit_net,
2933         .id   = &vxlan_net_id,
2934         .size = sizeof(struct vxlan_net),
2935 };
2936 
2937 static int __init vxlan_init_module(void)
2938 {
2939         int rc;
2940 
2941         vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2942         if (!vxlan_wq)
2943                 return -ENOMEM;
2944 
2945         get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2946 
2947         rc = register_pernet_subsys(&vxlan_net_ops);
2948         if (rc)
2949                 goto out1;
2950 
2951         rc = register_netdevice_notifier(&vxlan_notifier_block);
2952         if (rc)
2953                 goto out2;
2954 
2955         rc = rtnl_link_register(&vxlan_link_ops);
2956         if (rc)
2957                 goto out3;
2958 
2959         return 0;
2960 out3:
2961         unregister_netdevice_notifier(&vxlan_notifier_block);
2962 out2:
2963         unregister_pernet_subsys(&vxlan_net_ops);
2964 out1:
2965         destroy_workqueue(vxlan_wq);
2966         return rc;
2967 }
2968 late_initcall(vxlan_init_module);
2969 
2970 static void __exit vxlan_cleanup_module(void)
2971 {
2972         rtnl_link_unregister(&vxlan_link_ops);
2973         unregister_netdevice_notifier(&vxlan_notifier_block);
2974         destroy_workqueue(vxlan_wq);
2975         unregister_pernet_subsys(&vxlan_net_ops);
2976         /* rcu_barrier() is called by netns */
2977 }
2978 module_exit(vxlan_cleanup_module);
2979 
2980 MODULE_LICENSE("GPL");
2981 MODULE_VERSION(VXLAN_VERSION);
2982 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
2983 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
2984 MODULE_ALIAS_RTNL_LINK("vxlan");
2985 

This page was automatically generated by LXR 0.3.1 (source).  •  Linux is a registered trademark of Linus Torvalds  •  Contact us