Version:  2.0.40 2.2.26 2.4.37 3.12 3.13 3.14 3.15 3.16 3.17 3.18 3.19 4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9

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/module.h>
 15 #include <linux/errno.h>
 16 #include <linux/slab.h>
 17 #include <linux/udp.h>
 18 #include <linux/igmp.h>
 19 #include <linux/if_ether.h>
 20 #include <linux/ethtool.h>
 21 #include <net/arp.h>
 22 #include <net/ndisc.h>
 23 #include <net/ip.h>
 24 #include <net/icmp.h>
 25 #include <net/rtnetlink.h>
 26 #include <net/inet_ecn.h>
 27 #include <net/net_namespace.h>
 28 #include <net/netns/generic.h>
 29 #include <net/vxlan.h>
 30 
 31 #if IS_ENABLED(CONFIG_IPV6)
 32 #include <net/ip6_tunnel.h>
 33 #include <net/ip6_checksum.h>
 34 #endif
 35 
 36 #define VXLAN_VERSION   "0.1"
 37 
 38 #define PORT_HASH_BITS  8
 39 #define PORT_HASH_SIZE  (1<<PORT_HASH_BITS)
 40 #define FDB_AGE_DEFAULT 300 /* 5 min */
 41 #define FDB_AGE_INTERVAL (10 * HZ)      /* rescan interval */
 42 
 43 /* UDP port for VXLAN traffic.
 44  * The IANA assigned port is 4789, but the Linux default is 8472
 45  * for compatibility with early adopters.
 46  */
 47 static unsigned short vxlan_port __read_mostly = 8472;
 48 module_param_named(udp_port, vxlan_port, ushort, 0444);
 49 MODULE_PARM_DESC(udp_port, "Destination UDP port");
 50 
 51 static bool log_ecn_error = true;
 52 module_param(log_ecn_error, bool, 0644);
 53 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
 54 
 55 static int vxlan_net_id;
 56 static struct rtnl_link_ops vxlan_link_ops;
 57 
 58 static const u8 all_zeros_mac[ETH_ALEN + 2];
 59 
 60 static int vxlan_sock_add(struct vxlan_dev *vxlan);
 61 
 62 /* per-network namespace private data for this module */
 63 struct vxlan_net {
 64         struct list_head  vxlan_list;
 65         struct hlist_head sock_list[PORT_HASH_SIZE];
 66         spinlock_t        sock_lock;
 67 };
 68 
 69 /* Forwarding table entry */
 70 struct vxlan_fdb {
 71         struct hlist_node hlist;        /* linked list of entries */
 72         struct rcu_head   rcu;
 73         unsigned long     updated;      /* jiffies */
 74         unsigned long     used;
 75         struct list_head  remotes;
 76         u8                eth_addr[ETH_ALEN];
 77         u16               state;        /* see ndm_state */
 78         u8                flags;        /* see ndm_flags */
 79 };
 80 
 81 /* salt for hash table */
 82 static u32 vxlan_salt __read_mostly;
 83 
 84 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
 85 {
 86         return vs->flags & VXLAN_F_COLLECT_METADATA ||
 87                ip_tunnel_collect_metadata();
 88 }
 89 
 90 #if IS_ENABLED(CONFIG_IPV6)
 91 static inline
 92 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
 93 {
 94         if (a->sa.sa_family != b->sa.sa_family)
 95                 return false;
 96         if (a->sa.sa_family == AF_INET6)
 97                 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
 98         else
 99                 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
100 }
101 
102 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
103 {
104         if (ipa->sa.sa_family == AF_INET6)
105                 return ipv6_addr_any(&ipa->sin6.sin6_addr);
106         else
107                 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
108 }
109 
110 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
111 {
112         if (ipa->sa.sa_family == AF_INET6)
113                 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
114         else
115                 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
116 }
117 
118 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
119 {
120         if (nla_len(nla) >= sizeof(struct in6_addr)) {
121                 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
122                 ip->sa.sa_family = AF_INET6;
123                 return 0;
124         } else if (nla_len(nla) >= sizeof(__be32)) {
125                 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
126                 ip->sa.sa_family = AF_INET;
127                 return 0;
128         } else {
129                 return -EAFNOSUPPORT;
130         }
131 }
132 
133 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
134                               const union vxlan_addr *ip)
135 {
136         if (ip->sa.sa_family == AF_INET6)
137                 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
138         else
139                 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
140 }
141 
142 #else /* !CONFIG_IPV6 */
143 
144 static inline
145 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
146 {
147         return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
148 }
149 
150 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
151 {
152         return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
153 }
154 
155 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
156 {
157         return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
158 }
159 
160 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
161 {
162         if (nla_len(nla) >= sizeof(struct in6_addr)) {
163                 return -EAFNOSUPPORT;
164         } else if (nla_len(nla) >= sizeof(__be32)) {
165                 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
166                 ip->sa.sa_family = AF_INET;
167                 return 0;
168         } else {
169                 return -EAFNOSUPPORT;
170         }
171 }
172 
173 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
174                               const union vxlan_addr *ip)
175 {
176         return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
177 }
178 #endif
179 
180 /* Virtual Network hash table head */
181 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
182 {
183         return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
184 }
185 
186 /* Socket hash table head */
187 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
188 {
189         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
190 
191         return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
192 }
193 
194 /* First remote destination for a forwarding entry.
195  * Guaranteed to be non-NULL because remotes are never deleted.
196  */
197 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
198 {
199         return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
200 }
201 
202 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
203 {
204         return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
205 }
206 
207 /* Find VXLAN socket based on network namespace, address family and UDP port
208  * and enabled unshareable flags.
209  */
210 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
211                                           __be16 port, u32 flags)
212 {
213         struct vxlan_sock *vs;
214 
215         flags &= VXLAN_F_RCV_FLAGS;
216 
217         hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
218                 if (inet_sk(vs->sock->sk)->inet_sport == port &&
219                     vxlan_get_sk_family(vs) == family &&
220                     vs->flags == flags)
221                         return vs;
222         }
223         return NULL;
224 }
225 
226 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni)
227 {
228         struct vxlan_dev *vxlan;
229 
230         /* For flow based devices, map all packets to VNI 0 */
231         if (vs->flags & VXLAN_F_COLLECT_METADATA)
232                 vni = 0;
233 
234         hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) {
235                 if (vxlan->default_dst.remote_vni == vni)
236                         return vxlan;
237         }
238 
239         return NULL;
240 }
241 
242 /* Look up VNI in a per net namespace table */
243 static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni,
244                                         sa_family_t family, __be16 port,
245                                         u32 flags)
246 {
247         struct vxlan_sock *vs;
248 
249         vs = vxlan_find_sock(net, family, port, flags);
250         if (!vs)
251                 return NULL;
252 
253         return vxlan_vs_find_vni(vs, vni);
254 }
255 
256 /* Fill in neighbour message in skbuff. */
257 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
258                           const struct vxlan_fdb *fdb,
259                           u32 portid, u32 seq, int type, unsigned int flags,
260                           const struct vxlan_rdst *rdst)
261 {
262         unsigned long now = jiffies;
263         struct nda_cacheinfo ci;
264         struct nlmsghdr *nlh;
265         struct ndmsg *ndm;
266         bool send_ip, send_eth;
267 
268         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
269         if (nlh == NULL)
270                 return -EMSGSIZE;
271 
272         ndm = nlmsg_data(nlh);
273         memset(ndm, 0, sizeof(*ndm));
274 
275         send_eth = send_ip = true;
276 
277         if (type == RTM_GETNEIGH) {
278                 ndm->ndm_family = AF_INET;
279                 send_ip = !vxlan_addr_any(&rdst->remote_ip);
280                 send_eth = !is_zero_ether_addr(fdb->eth_addr);
281         } else
282                 ndm->ndm_family = AF_BRIDGE;
283         ndm->ndm_state = fdb->state;
284         ndm->ndm_ifindex = vxlan->dev->ifindex;
285         ndm->ndm_flags = fdb->flags;
286         ndm->ndm_type = RTN_UNICAST;
287 
288         if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
289             nla_put_s32(skb, NDA_LINK_NETNSID,
290                         peernet2id(dev_net(vxlan->dev), vxlan->net)))
291                 goto nla_put_failure;
292 
293         if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
294                 goto nla_put_failure;
295 
296         if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
297                 goto nla_put_failure;
298 
299         if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
300             nla_put_be16(skb, NDA_PORT, rdst->remote_port))
301                 goto nla_put_failure;
302         if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
303             nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
304                 goto nla_put_failure;
305         if (rdst->remote_ifindex &&
306             nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
307                 goto nla_put_failure;
308 
309         ci.ndm_used      = jiffies_to_clock_t(now - fdb->used);
310         ci.ndm_confirmed = 0;
311         ci.ndm_updated   = jiffies_to_clock_t(now - fdb->updated);
312         ci.ndm_refcnt    = 0;
313 
314         if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
315                 goto nla_put_failure;
316 
317         nlmsg_end(skb, nlh);
318         return 0;
319 
320 nla_put_failure:
321         nlmsg_cancel(skb, nlh);
322         return -EMSGSIZE;
323 }
324 
325 static inline size_t vxlan_nlmsg_size(void)
326 {
327         return NLMSG_ALIGN(sizeof(struct ndmsg))
328                 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
329                 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
330                 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
331                 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
332                 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
333                 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
334                 + nla_total_size(sizeof(struct nda_cacheinfo));
335 }
336 
337 static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
338                              struct vxlan_rdst *rd, int type)
339 {
340         struct net *net = dev_net(vxlan->dev);
341         struct sk_buff *skb;
342         int err = -ENOBUFS;
343 
344         skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
345         if (skb == NULL)
346                 goto errout;
347 
348         err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
349         if (err < 0) {
350                 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
351                 WARN_ON(err == -EMSGSIZE);
352                 kfree_skb(skb);
353                 goto errout;
354         }
355 
356         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
357         return;
358 errout:
359         if (err < 0)
360                 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
361 }
362 
363 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
364 {
365         struct vxlan_dev *vxlan = netdev_priv(dev);
366         struct vxlan_fdb f = {
367                 .state = NUD_STALE,
368         };
369         struct vxlan_rdst remote = {
370                 .remote_ip = *ipa, /* goes to NDA_DST */
371                 .remote_vni = cpu_to_be32(VXLAN_N_VID),
372         };
373 
374         vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
375 }
376 
377 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
378 {
379         struct vxlan_fdb f = {
380                 .state = NUD_STALE,
381         };
382         struct vxlan_rdst remote = { };
383 
384         memcpy(f.eth_addr, eth_addr, ETH_ALEN);
385 
386         vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
387 }
388 
389 /* Hash Ethernet address */
390 static u32 eth_hash(const unsigned char *addr)
391 {
392         u64 value = get_unaligned((u64 *)addr);
393 
394         /* only want 6 bytes */
395 #ifdef __BIG_ENDIAN
396         value >>= 16;
397 #else
398         value <<= 16;
399 #endif
400         return hash_64(value, FDB_HASH_BITS);
401 }
402 
403 /* Hash chain to use given mac address */
404 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
405                                                 const u8 *mac)
406 {
407         return &vxlan->fdb_head[eth_hash(mac)];
408 }
409 
410 /* Look up Ethernet address in forwarding table */
411 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
412                                         const u8 *mac)
413 {
414         struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
415         struct vxlan_fdb *f;
416 
417         hlist_for_each_entry_rcu(f, head, hlist) {
418                 if (ether_addr_equal(mac, f->eth_addr))
419                         return f;
420         }
421 
422         return NULL;
423 }
424 
425 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
426                                         const u8 *mac)
427 {
428         struct vxlan_fdb *f;
429 
430         f = __vxlan_find_mac(vxlan, mac);
431         if (f)
432                 f->used = jiffies;
433 
434         return f;
435 }
436 
437 /* caller should hold vxlan->hash_lock */
438 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
439                                               union vxlan_addr *ip, __be16 port,
440                                               __be32 vni, __u32 ifindex)
441 {
442         struct vxlan_rdst *rd;
443 
444         list_for_each_entry(rd, &f->remotes, list) {
445                 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
446                     rd->remote_port == port &&
447                     rd->remote_vni == vni &&
448                     rd->remote_ifindex == ifindex)
449                         return rd;
450         }
451 
452         return NULL;
453 }
454 
455 /* Replace destination of unicast mac */
456 static int vxlan_fdb_replace(struct vxlan_fdb *f,
457                              union vxlan_addr *ip, __be16 port, __be32 vni,
458                              __u32 ifindex)
459 {
460         struct vxlan_rdst *rd;
461 
462         rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
463         if (rd)
464                 return 0;
465 
466         rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
467         if (!rd)
468                 return 0;
469 
470         dst_cache_reset(&rd->dst_cache);
471         rd->remote_ip = *ip;
472         rd->remote_port = port;
473         rd->remote_vni = vni;
474         rd->remote_ifindex = ifindex;
475         return 1;
476 }
477 
478 /* Add/update destinations for multicast */
479 static int vxlan_fdb_append(struct vxlan_fdb *f,
480                             union vxlan_addr *ip, __be16 port, __be32 vni,
481                             __u32 ifindex, struct vxlan_rdst **rdp)
482 {
483         struct vxlan_rdst *rd;
484 
485         rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
486         if (rd)
487                 return 0;
488 
489         rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
490         if (rd == NULL)
491                 return -ENOBUFS;
492 
493         if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
494                 kfree(rd);
495                 return -ENOBUFS;
496         }
497 
498         rd->remote_ip = *ip;
499         rd->remote_port = port;
500         rd->remote_vni = vni;
501         rd->remote_ifindex = ifindex;
502 
503         list_add_tail_rcu(&rd->list, &f->remotes);
504 
505         *rdp = rd;
506         return 1;
507 }
508 
509 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
510                                           unsigned int off,
511                                           struct vxlanhdr *vh, size_t hdrlen,
512                                           __be32 vni_field,
513                                           struct gro_remcsum *grc,
514                                           bool nopartial)
515 {
516         size_t start, offset;
517 
518         if (skb->remcsum_offload)
519                 return vh;
520 
521         if (!NAPI_GRO_CB(skb)->csum_valid)
522                 return NULL;
523 
524         start = vxlan_rco_start(vni_field);
525         offset = start + vxlan_rco_offset(vni_field);
526 
527         vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
528                                      start, offset, grc, nopartial);
529 
530         skb->remcsum_offload = 1;
531 
532         return vh;
533 }
534 
535 static struct sk_buff **vxlan_gro_receive(struct sock *sk,
536                                           struct sk_buff **head,
537                                           struct sk_buff *skb)
538 {
539         struct sk_buff *p, **pp = NULL;
540         struct vxlanhdr *vh, *vh2;
541         unsigned int hlen, off_vx;
542         int flush = 1;
543         struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
544         __be32 flags;
545         struct gro_remcsum grc;
546 
547         skb_gro_remcsum_init(&grc);
548 
549         off_vx = skb_gro_offset(skb);
550         hlen = off_vx + sizeof(*vh);
551         vh   = skb_gro_header_fast(skb, off_vx);
552         if (skb_gro_header_hard(skb, hlen)) {
553                 vh = skb_gro_header_slow(skb, hlen, off_vx);
554                 if (unlikely(!vh))
555                         goto out;
556         }
557 
558         skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
559 
560         flags = vh->vx_flags;
561 
562         if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
563                 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
564                                        vh->vx_vni, &grc,
565                                        !!(vs->flags &
566                                           VXLAN_F_REMCSUM_NOPARTIAL));
567 
568                 if (!vh)
569                         goto out;
570         }
571 
572         skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
573 
574         for (p = *head; p; p = p->next) {
575                 if (!NAPI_GRO_CB(p)->same_flow)
576                         continue;
577 
578                 vh2 = (struct vxlanhdr *)(p->data + off_vx);
579                 if (vh->vx_flags != vh2->vx_flags ||
580                     vh->vx_vni != vh2->vx_vni) {
581                         NAPI_GRO_CB(p)->same_flow = 0;
582                         continue;
583                 }
584         }
585 
586         pp = call_gro_receive(eth_gro_receive, head, skb);
587         flush = 0;
588 
589 out:
590         skb_gro_remcsum_cleanup(skb, &grc);
591         NAPI_GRO_CB(skb)->flush |= flush;
592 
593         return pp;
594 }
595 
596 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
597 {
598         /* Sets 'skb->inner_mac_header' since we are always called with
599          * 'skb->encapsulation' set.
600          */
601         return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
602 }
603 
604 /* Add new entry to forwarding table -- assumes lock held */
605 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
606                             const u8 *mac, union vxlan_addr *ip,
607                             __u16 state, __u16 flags,
608                             __be16 port, __be32 vni, __u32 ifindex,
609                             __u8 ndm_flags)
610 {
611         struct vxlan_rdst *rd = NULL;
612         struct vxlan_fdb *f;
613         int notify = 0;
614         int rc;
615 
616         f = __vxlan_find_mac(vxlan, mac);
617         if (f) {
618                 if (flags & NLM_F_EXCL) {
619                         netdev_dbg(vxlan->dev,
620                                    "lost race to create %pM\n", mac);
621                         return -EEXIST;
622                 }
623                 if (f->state != state) {
624                         f->state = state;
625                         f->updated = jiffies;
626                         notify = 1;
627                 }
628                 if (f->flags != ndm_flags) {
629                         f->flags = ndm_flags;
630                         f->updated = jiffies;
631                         notify = 1;
632                 }
633                 if ((flags & NLM_F_REPLACE)) {
634                         /* Only change unicasts */
635                         if (!(is_multicast_ether_addr(f->eth_addr) ||
636                              is_zero_ether_addr(f->eth_addr))) {
637                                 notify |= vxlan_fdb_replace(f, ip, port, vni,
638                                                            ifindex);
639                         } else
640                                 return -EOPNOTSUPP;
641                 }
642                 if ((flags & NLM_F_APPEND) &&
643                     (is_multicast_ether_addr(f->eth_addr) ||
644                      is_zero_ether_addr(f->eth_addr))) {
645                         rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
646 
647                         if (rc < 0)
648                                 return rc;
649                         notify |= rc;
650                 }
651         } else {
652                 if (!(flags & NLM_F_CREATE))
653                         return -ENOENT;
654 
655                 if (vxlan->cfg.addrmax &&
656                     vxlan->addrcnt >= vxlan->cfg.addrmax)
657                         return -ENOSPC;
658 
659                 /* Disallow replace to add a multicast entry */
660                 if ((flags & NLM_F_REPLACE) &&
661                     (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
662                         return -EOPNOTSUPP;
663 
664                 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
665                 f = kmalloc(sizeof(*f), GFP_ATOMIC);
666                 if (!f)
667                         return -ENOMEM;
668 
669                 notify = 1;
670                 f->state = state;
671                 f->flags = ndm_flags;
672                 f->updated = f->used = jiffies;
673                 INIT_LIST_HEAD(&f->remotes);
674                 memcpy(f->eth_addr, mac, ETH_ALEN);
675 
676                 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
677                 if (rc < 0) {
678                         kfree(f);
679                         return rc;
680                 }
681 
682                 ++vxlan->addrcnt;
683                 hlist_add_head_rcu(&f->hlist,
684                                    vxlan_fdb_head(vxlan, mac));
685         }
686 
687         if (notify) {
688                 if (rd == NULL)
689                         rd = first_remote_rtnl(f);
690                 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
691         }
692 
693         return 0;
694 }
695 
696 static void vxlan_fdb_free(struct rcu_head *head)
697 {
698         struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
699         struct vxlan_rdst *rd, *nd;
700 
701         list_for_each_entry_safe(rd, nd, &f->remotes, list) {
702                 dst_cache_destroy(&rd->dst_cache);
703                 kfree(rd);
704         }
705         kfree(f);
706 }
707 
708 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
709 {
710         netdev_dbg(vxlan->dev,
711                     "delete %pM\n", f->eth_addr);
712 
713         --vxlan->addrcnt;
714         vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
715 
716         hlist_del_rcu(&f->hlist);
717         call_rcu(&f->rcu, vxlan_fdb_free);
718 }
719 
720 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
721                            union vxlan_addr *ip, __be16 *port, __be32 *vni,
722                            u32 *ifindex)
723 {
724         struct net *net = dev_net(vxlan->dev);
725         int err;
726 
727         if (tb[NDA_DST]) {
728                 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
729                 if (err)
730                         return err;
731         } else {
732                 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
733                 if (remote->sa.sa_family == AF_INET) {
734                         ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
735                         ip->sa.sa_family = AF_INET;
736 #if IS_ENABLED(CONFIG_IPV6)
737                 } else {
738                         ip->sin6.sin6_addr = in6addr_any;
739                         ip->sa.sa_family = AF_INET6;
740 #endif
741                 }
742         }
743 
744         if (tb[NDA_PORT]) {
745                 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
746                         return -EINVAL;
747                 *port = nla_get_be16(tb[NDA_PORT]);
748         } else {
749                 *port = vxlan->cfg.dst_port;
750         }
751 
752         if (tb[NDA_VNI]) {
753                 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
754                         return -EINVAL;
755                 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
756         } else {
757                 *vni = vxlan->default_dst.remote_vni;
758         }
759 
760         if (tb[NDA_IFINDEX]) {
761                 struct net_device *tdev;
762 
763                 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
764                         return -EINVAL;
765                 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
766                 tdev = __dev_get_by_index(net, *ifindex);
767                 if (!tdev)
768                         return -EADDRNOTAVAIL;
769         } else {
770                 *ifindex = 0;
771         }
772 
773         return 0;
774 }
775 
776 /* Add static entry (via netlink) */
777 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
778                          struct net_device *dev,
779                          const unsigned char *addr, u16 vid, u16 flags)
780 {
781         struct vxlan_dev *vxlan = netdev_priv(dev);
782         /* struct net *net = dev_net(vxlan->dev); */
783         union vxlan_addr ip;
784         __be16 port;
785         __be32 vni;
786         u32 ifindex;
787         int err;
788 
789         if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
790                 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
791                         ndm->ndm_state);
792                 return -EINVAL;
793         }
794 
795         if (tb[NDA_DST] == NULL)
796                 return -EINVAL;
797 
798         err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
799         if (err)
800                 return err;
801 
802         if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
803                 return -EAFNOSUPPORT;
804 
805         spin_lock_bh(&vxlan->hash_lock);
806         err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
807                                port, vni, ifindex, ndm->ndm_flags);
808         spin_unlock_bh(&vxlan->hash_lock);
809 
810         return err;
811 }
812 
813 /* Delete entry (via netlink) */
814 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
815                             struct net_device *dev,
816                             const unsigned char *addr, u16 vid)
817 {
818         struct vxlan_dev *vxlan = netdev_priv(dev);
819         struct vxlan_fdb *f;
820         struct vxlan_rdst *rd = NULL;
821         union vxlan_addr ip;
822         __be16 port;
823         __be32 vni;
824         u32 ifindex;
825         int err;
826 
827         err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
828         if (err)
829                 return err;
830 
831         err = -ENOENT;
832 
833         spin_lock_bh(&vxlan->hash_lock);
834         f = vxlan_find_mac(vxlan, addr);
835         if (!f)
836                 goto out;
837 
838         if (!vxlan_addr_any(&ip)) {
839                 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
840                 if (!rd)
841                         goto out;
842         }
843 
844         err = 0;
845 
846         /* remove a destination if it's not the only one on the list,
847          * otherwise destroy the fdb entry
848          */
849         if (rd && !list_is_singular(&f->remotes)) {
850                 list_del_rcu(&rd->list);
851                 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
852                 kfree_rcu(rd, rcu);
853                 goto out;
854         }
855 
856         vxlan_fdb_destroy(vxlan, f);
857 
858 out:
859         spin_unlock_bh(&vxlan->hash_lock);
860 
861         return err;
862 }
863 
864 /* Dump forwarding table */
865 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
866                           struct net_device *dev,
867                           struct net_device *filter_dev, int *idx)
868 {
869         struct vxlan_dev *vxlan = netdev_priv(dev);
870         unsigned int h;
871         int err = 0;
872 
873         for (h = 0; h < FDB_HASH_SIZE; ++h) {
874                 struct vxlan_fdb *f;
875 
876                 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
877                         struct vxlan_rdst *rd;
878 
879                         list_for_each_entry_rcu(rd, &f->remotes, list) {
880                                 if (*idx < cb->args[2])
881                                         goto skip;
882 
883                                 err = vxlan_fdb_info(skb, vxlan, f,
884                                                      NETLINK_CB(cb->skb).portid,
885                                                      cb->nlh->nlmsg_seq,
886                                                      RTM_NEWNEIGH,
887                                                      NLM_F_MULTI, rd);
888                                 if (err < 0)
889                                         goto out;
890 skip:
891                                 *idx += 1;
892                         }
893                 }
894         }
895 out:
896         return err;
897 }
898 
899 /* Watch incoming packets to learn mapping between Ethernet address
900  * and Tunnel endpoint.
901  * Return true if packet is bogus and should be dropped.
902  */
903 static bool vxlan_snoop(struct net_device *dev,
904                         union vxlan_addr *src_ip, const u8 *src_mac)
905 {
906         struct vxlan_dev *vxlan = netdev_priv(dev);
907         struct vxlan_fdb *f;
908 
909         f = vxlan_find_mac(vxlan, src_mac);
910         if (likely(f)) {
911                 struct vxlan_rdst *rdst = first_remote_rcu(f);
912 
913                 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
914                         return false;
915 
916                 /* Don't migrate static entries, drop packets */
917                 if (f->state & NUD_NOARP)
918                         return true;
919 
920                 if (net_ratelimit())
921                         netdev_info(dev,
922                                     "%pM migrated from %pIS to %pIS\n",
923                                     src_mac, &rdst->remote_ip.sa, &src_ip->sa);
924 
925                 rdst->remote_ip = *src_ip;
926                 f->updated = jiffies;
927                 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
928         } else {
929                 /* learned new entry */
930                 spin_lock(&vxlan->hash_lock);
931 
932                 /* close off race between vxlan_flush and incoming packets */
933                 if (netif_running(dev))
934                         vxlan_fdb_create(vxlan, src_mac, src_ip,
935                                          NUD_REACHABLE,
936                                          NLM_F_EXCL|NLM_F_CREATE,
937                                          vxlan->cfg.dst_port,
938                                          vxlan->default_dst.remote_vni,
939                                          0, NTF_SELF);
940                 spin_unlock(&vxlan->hash_lock);
941         }
942 
943         return false;
944 }
945 
946 /* See if multicast group is already in use by other ID */
947 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
948 {
949         struct vxlan_dev *vxlan;
950         struct vxlan_sock *sock4;
951 #if IS_ENABLED(CONFIG_IPV6)
952         struct vxlan_sock *sock6;
953 #endif
954         unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
955 
956         sock4 = rtnl_dereference(dev->vn4_sock);
957 
958         /* The vxlan_sock is only used by dev, leaving group has
959          * no effect on other vxlan devices.
960          */
961         if (family == AF_INET && sock4 && atomic_read(&sock4->refcnt) == 1)
962                 return false;
963 #if IS_ENABLED(CONFIG_IPV6)
964         sock6 = rtnl_dereference(dev->vn6_sock);
965         if (family == AF_INET6 && sock6 && atomic_read(&sock6->refcnt) == 1)
966                 return false;
967 #endif
968 
969         list_for_each_entry(vxlan, &vn->vxlan_list, next) {
970                 if (!netif_running(vxlan->dev) || vxlan == dev)
971                         continue;
972 
973                 if (family == AF_INET &&
974                     rtnl_dereference(vxlan->vn4_sock) != sock4)
975                         continue;
976 #if IS_ENABLED(CONFIG_IPV6)
977                 if (family == AF_INET6 &&
978                     rtnl_dereference(vxlan->vn6_sock) != sock6)
979                         continue;
980 #endif
981 
982                 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
983                                       &dev->default_dst.remote_ip))
984                         continue;
985 
986                 if (vxlan->default_dst.remote_ifindex !=
987                     dev->default_dst.remote_ifindex)
988                         continue;
989 
990                 return true;
991         }
992 
993         return false;
994 }
995 
996 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
997 {
998         struct vxlan_net *vn;
999 
1000         if (!vs)
1001                 return false;
1002         if (!atomic_dec_and_test(&vs->refcnt))
1003                 return false;
1004 
1005         vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1006         spin_lock(&vn->sock_lock);
1007         hlist_del_rcu(&vs->hlist);
1008         udp_tunnel_notify_del_rx_port(vs->sock,
1009                                       (vs->flags & VXLAN_F_GPE) ?
1010                                       UDP_TUNNEL_TYPE_VXLAN_GPE :
1011                                       UDP_TUNNEL_TYPE_VXLAN);
1012         spin_unlock(&vn->sock_lock);
1013 
1014         return true;
1015 }
1016 
1017 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1018 {
1019         struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1020 #if IS_ENABLED(CONFIG_IPV6)
1021         struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1022 
1023         rcu_assign_pointer(vxlan->vn6_sock, NULL);
1024 #endif
1025 
1026         rcu_assign_pointer(vxlan->vn4_sock, NULL);
1027         synchronize_net();
1028 
1029         if (__vxlan_sock_release_prep(sock4)) {
1030                 udp_tunnel_sock_release(sock4->sock);
1031                 kfree(sock4);
1032         }
1033 
1034 #if IS_ENABLED(CONFIG_IPV6)
1035         if (__vxlan_sock_release_prep(sock6)) {
1036                 udp_tunnel_sock_release(sock6->sock);
1037                 kfree(sock6);
1038         }
1039 #endif
1040 }
1041 
1042 /* Update multicast group membership when first VNI on
1043  * multicast address is brought up
1044  */
1045 static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1046 {
1047         struct sock *sk;
1048         union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1049         int ifindex = vxlan->default_dst.remote_ifindex;
1050         int ret = -EINVAL;
1051 
1052         if (ip->sa.sa_family == AF_INET) {
1053                 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1054                 struct ip_mreqn mreq = {
1055                         .imr_multiaddr.s_addr   = ip->sin.sin_addr.s_addr,
1056                         .imr_ifindex            = ifindex,
1057                 };
1058 
1059                 sk = sock4->sock->sk;
1060                 lock_sock(sk);
1061                 ret = ip_mc_join_group(sk, &mreq);
1062                 release_sock(sk);
1063 #if IS_ENABLED(CONFIG_IPV6)
1064         } else {
1065                 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1066 
1067                 sk = sock6->sock->sk;
1068                 lock_sock(sk);
1069                 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1070                                                    &ip->sin6.sin6_addr);
1071                 release_sock(sk);
1072 #endif
1073         }
1074 
1075         return ret;
1076 }
1077 
1078 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1079 static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1080 {
1081         struct sock *sk;
1082         union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1083         int ifindex = vxlan->default_dst.remote_ifindex;
1084         int ret = -EINVAL;
1085 
1086         if (ip->sa.sa_family == AF_INET) {
1087                 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1088                 struct ip_mreqn mreq = {
1089                         .imr_multiaddr.s_addr   = ip->sin.sin_addr.s_addr,
1090                         .imr_ifindex            = ifindex,
1091                 };
1092 
1093                 sk = sock4->sock->sk;
1094                 lock_sock(sk);
1095                 ret = ip_mc_leave_group(sk, &mreq);
1096                 release_sock(sk);
1097 #if IS_ENABLED(CONFIG_IPV6)
1098         } else {
1099                 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1100 
1101                 sk = sock6->sock->sk;
1102                 lock_sock(sk);
1103                 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1104                                                    &ip->sin6.sin6_addr);
1105                 release_sock(sk);
1106 #endif
1107         }
1108 
1109         return ret;
1110 }
1111 
1112 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1113                           struct sk_buff *skb, u32 vxflags)
1114 {
1115         size_t start, offset;
1116 
1117         if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1118                 goto out;
1119 
1120         start = vxlan_rco_start(unparsed->vx_vni);
1121         offset = start + vxlan_rco_offset(unparsed->vx_vni);
1122 
1123         if (!pskb_may_pull(skb, offset + sizeof(u16)))
1124                 return false;
1125 
1126         skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1127                             !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1128 out:
1129         unparsed->vx_flags &= ~VXLAN_HF_RCO;
1130         unparsed->vx_vni &= VXLAN_VNI_MASK;
1131         return true;
1132 }
1133 
1134 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1135                                 struct sk_buff *skb, u32 vxflags,
1136                                 struct vxlan_metadata *md)
1137 {
1138         struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1139         struct metadata_dst *tun_dst;
1140 
1141         if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1142                 goto out;
1143 
1144         md->gbp = ntohs(gbp->policy_id);
1145 
1146         tun_dst = (struct metadata_dst *)skb_dst(skb);
1147         if (tun_dst) {
1148                 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1149                 tun_dst->u.tun_info.options_len = sizeof(*md);
1150         }
1151         if (gbp->dont_learn)
1152                 md->gbp |= VXLAN_GBP_DONT_LEARN;
1153 
1154         if (gbp->policy_applied)
1155                 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1156 
1157         /* In flow-based mode, GBP is carried in dst_metadata */
1158         if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1159                 skb->mark = md->gbp;
1160 out:
1161         unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1162 }
1163 
1164 static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1165                                 __be16 *protocol,
1166                                 struct sk_buff *skb, u32 vxflags)
1167 {
1168         struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1169 
1170         /* Need to have Next Protocol set for interfaces in GPE mode. */
1171         if (!gpe->np_applied)
1172                 return false;
1173         /* "The initial version is 0. If a receiver does not support the
1174          * version indicated it MUST drop the packet.
1175          */
1176         if (gpe->version != 0)
1177                 return false;
1178         /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1179          * processing MUST occur." However, we don't implement OAM
1180          * processing, thus drop the packet.
1181          */
1182         if (gpe->oam_flag)
1183                 return false;
1184 
1185         switch (gpe->next_protocol) {
1186         case VXLAN_GPE_NP_IPV4:
1187                 *protocol = htons(ETH_P_IP);
1188                 break;
1189         case VXLAN_GPE_NP_IPV6:
1190                 *protocol = htons(ETH_P_IPV6);
1191                 break;
1192         case VXLAN_GPE_NP_ETHERNET:
1193                 *protocol = htons(ETH_P_TEB);
1194                 break;
1195         default:
1196                 return false;
1197         }
1198 
1199         unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1200         return true;
1201 }
1202 
1203 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1204                           struct vxlan_sock *vs,
1205                           struct sk_buff *skb)
1206 {
1207         union vxlan_addr saddr;
1208 
1209         skb_reset_mac_header(skb);
1210         skb->protocol = eth_type_trans(skb, vxlan->dev);
1211         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1212 
1213         /* Ignore packet loops (and multicast echo) */
1214         if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1215                 return false;
1216 
1217         /* Get address from the outer IP header */
1218         if (vxlan_get_sk_family(vs) == AF_INET) {
1219                 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1220                 saddr.sa.sa_family = AF_INET;
1221 #if IS_ENABLED(CONFIG_IPV6)
1222         } else {
1223                 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1224                 saddr.sa.sa_family = AF_INET6;
1225 #endif
1226         }
1227 
1228         if ((vxlan->flags & VXLAN_F_LEARN) &&
1229             vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1230                 return false;
1231 
1232         return true;
1233 }
1234 
1235 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1236                                   struct sk_buff *skb)
1237 {
1238         int err = 0;
1239 
1240         if (vxlan_get_sk_family(vs) == AF_INET)
1241                 err = IP_ECN_decapsulate(oiph, skb);
1242 #if IS_ENABLED(CONFIG_IPV6)
1243         else
1244                 err = IP6_ECN_decapsulate(oiph, skb);
1245 #endif
1246 
1247         if (unlikely(err) && log_ecn_error) {
1248                 if (vxlan_get_sk_family(vs) == AF_INET)
1249                         net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1250                                              &((struct iphdr *)oiph)->saddr,
1251                                              ((struct iphdr *)oiph)->tos);
1252                 else
1253                         net_info_ratelimited("non-ECT from %pI6\n",
1254                                              &((struct ipv6hdr *)oiph)->saddr);
1255         }
1256         return err <= 1;
1257 }
1258 
1259 /* Callback from net/ipv4/udp.c to receive packets */
1260 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1261 {
1262         struct pcpu_sw_netstats *stats;
1263         struct vxlan_dev *vxlan;
1264         struct vxlan_sock *vs;
1265         struct vxlanhdr unparsed;
1266         struct vxlan_metadata _md;
1267         struct vxlan_metadata *md = &_md;
1268         __be16 protocol = htons(ETH_P_TEB);
1269         bool raw_proto = false;
1270         void *oiph;
1271 
1272         /* Need UDP and VXLAN header to be present */
1273         if (!pskb_may_pull(skb, VXLAN_HLEN))
1274                 goto drop;
1275 
1276         unparsed = *vxlan_hdr(skb);
1277         /* VNI flag always required to be set */
1278         if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1279                 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1280                            ntohl(vxlan_hdr(skb)->vx_flags),
1281                            ntohl(vxlan_hdr(skb)->vx_vni));
1282                 /* Return non vxlan pkt */
1283                 goto drop;
1284         }
1285         unparsed.vx_flags &= ~VXLAN_HF_VNI;
1286         unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1287 
1288         vs = rcu_dereference_sk_user_data(sk);
1289         if (!vs)
1290                 goto drop;
1291 
1292         vxlan = vxlan_vs_find_vni(vs, vxlan_vni(vxlan_hdr(skb)->vx_vni));
1293         if (!vxlan)
1294                 goto drop;
1295 
1296         /* For backwards compatibility, only allow reserved fields to be
1297          * used by VXLAN extensions if explicitly requested.
1298          */
1299         if (vs->flags & VXLAN_F_GPE) {
1300                 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1301                         goto drop;
1302                 raw_proto = true;
1303         }
1304 
1305         if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1306                                    !net_eq(vxlan->net, dev_net(vxlan->dev))))
1307                         goto drop;
1308 
1309         if (vxlan_collect_metadata(vs)) {
1310                 __be32 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1311                 struct metadata_dst *tun_dst;
1312 
1313                 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1314                                          key32_to_tunnel_id(vni), sizeof(*md));
1315 
1316                 if (!tun_dst)
1317                         goto drop;
1318 
1319                 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1320 
1321                 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1322         } else {
1323                 memset(md, 0, sizeof(*md));
1324         }
1325 
1326         if (vs->flags & VXLAN_F_REMCSUM_RX)
1327                 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1328                         goto drop;
1329         if (vs->flags & VXLAN_F_GBP)
1330                 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1331         /* Note that GBP and GPE can never be active together. This is
1332          * ensured in vxlan_dev_configure.
1333          */
1334 
1335         if (unparsed.vx_flags || unparsed.vx_vni) {
1336                 /* If there are any unprocessed flags remaining treat
1337                  * this as a malformed packet. This behavior diverges from
1338                  * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1339                  * in reserved fields are to be ignored. The approach here
1340                  * maintains compatibility with previous stack code, and also
1341                  * is more robust and provides a little more security in
1342                  * adding extensions to VXLAN.
1343                  */
1344                 goto drop;
1345         }
1346 
1347         if (!raw_proto) {
1348                 if (!vxlan_set_mac(vxlan, vs, skb))
1349                         goto drop;
1350         } else {
1351                 skb_reset_mac_header(skb);
1352                 skb->dev = vxlan->dev;
1353                 skb->pkt_type = PACKET_HOST;
1354         }
1355 
1356         oiph = skb_network_header(skb);
1357         skb_reset_network_header(skb);
1358 
1359         if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1360                 ++vxlan->dev->stats.rx_frame_errors;
1361                 ++vxlan->dev->stats.rx_errors;
1362                 goto drop;
1363         }
1364 
1365         stats = this_cpu_ptr(vxlan->dev->tstats);
1366         u64_stats_update_begin(&stats->syncp);
1367         stats->rx_packets++;
1368         stats->rx_bytes += skb->len;
1369         u64_stats_update_end(&stats->syncp);
1370 
1371         gro_cells_receive(&vxlan->gro_cells, skb);
1372         return 0;
1373 
1374 drop:
1375         /* Consume bad packet */
1376         kfree_skb(skb);
1377         return 0;
1378 }
1379 
1380 static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1381 {
1382         struct vxlan_dev *vxlan = netdev_priv(dev);
1383         struct arphdr *parp;
1384         u8 *arpptr, *sha;
1385         __be32 sip, tip;
1386         struct neighbour *n;
1387 
1388         if (dev->flags & IFF_NOARP)
1389                 goto out;
1390 
1391         if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1392                 dev->stats.tx_dropped++;
1393                 goto out;
1394         }
1395         parp = arp_hdr(skb);
1396 
1397         if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1398              parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1399             parp->ar_pro != htons(ETH_P_IP) ||
1400             parp->ar_op != htons(ARPOP_REQUEST) ||
1401             parp->ar_hln != dev->addr_len ||
1402             parp->ar_pln != 4)
1403                 goto out;
1404         arpptr = (u8 *)parp + sizeof(struct arphdr);
1405         sha = arpptr;
1406         arpptr += dev->addr_len;        /* sha */
1407         memcpy(&sip, arpptr, sizeof(sip));
1408         arpptr += sizeof(sip);
1409         arpptr += dev->addr_len;        /* tha */
1410         memcpy(&tip, arpptr, sizeof(tip));
1411 
1412         if (ipv4_is_loopback(tip) ||
1413             ipv4_is_multicast(tip))
1414                 goto out;
1415 
1416         n = neigh_lookup(&arp_tbl, &tip, dev);
1417 
1418         if (n) {
1419                 struct vxlan_fdb *f;
1420                 struct sk_buff  *reply;
1421 
1422                 if (!(n->nud_state & NUD_CONNECTED)) {
1423                         neigh_release(n);
1424                         goto out;
1425                 }
1426 
1427                 f = vxlan_find_mac(vxlan, n->ha);
1428                 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1429                         /* bridge-local neighbor */
1430                         neigh_release(n);
1431                         goto out;
1432                 }
1433 
1434                 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1435                                 n->ha, sha);
1436 
1437                 neigh_release(n);
1438 
1439                 if (reply == NULL)
1440                         goto out;
1441 
1442                 skb_reset_mac_header(reply);
1443                 __skb_pull(reply, skb_network_offset(reply));
1444                 reply->ip_summed = CHECKSUM_UNNECESSARY;
1445                 reply->pkt_type = PACKET_HOST;
1446 
1447                 if (netif_rx_ni(reply) == NET_RX_DROP)
1448                         dev->stats.rx_dropped++;
1449         } else if (vxlan->flags & VXLAN_F_L3MISS) {
1450                 union vxlan_addr ipa = {
1451                         .sin.sin_addr.s_addr = tip,
1452                         .sin.sin_family = AF_INET,
1453                 };
1454 
1455                 vxlan_ip_miss(dev, &ipa);
1456         }
1457 out:
1458         consume_skb(skb);
1459         return NETDEV_TX_OK;
1460 }
1461 
1462 #if IS_ENABLED(CONFIG_IPV6)
1463 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1464         struct neighbour *n, bool isrouter)
1465 {
1466         struct net_device *dev = request->dev;
1467         struct sk_buff *reply;
1468         struct nd_msg *ns, *na;
1469         struct ipv6hdr *pip6;
1470         u8 *daddr;
1471         int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1472         int ns_olen;
1473         int i, len;
1474 
1475         if (dev == NULL)
1476                 return NULL;
1477 
1478         len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1479                 sizeof(*na) + na_olen + dev->needed_tailroom;
1480         reply = alloc_skb(len, GFP_ATOMIC);
1481         if (reply == NULL)
1482                 return NULL;
1483 
1484         reply->protocol = htons(ETH_P_IPV6);
1485         reply->dev = dev;
1486         skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1487         skb_push(reply, sizeof(struct ethhdr));
1488         skb_reset_mac_header(reply);
1489 
1490         ns = (struct nd_msg *)skb_transport_header(request);
1491 
1492         daddr = eth_hdr(request)->h_source;
1493         ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1494         for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1495                 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1496                         daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1497                         break;
1498                 }
1499         }
1500 
1501         /* Ethernet header */
1502         ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1503         ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1504         eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1505         reply->protocol = htons(ETH_P_IPV6);
1506 
1507         skb_pull(reply, sizeof(struct ethhdr));
1508         skb_reset_network_header(reply);
1509         skb_put(reply, sizeof(struct ipv6hdr));
1510 
1511         /* IPv6 header */
1512 
1513         pip6 = ipv6_hdr(reply);
1514         memset(pip6, 0, sizeof(struct ipv6hdr));
1515         pip6->version = 6;
1516         pip6->priority = ipv6_hdr(request)->priority;
1517         pip6->nexthdr = IPPROTO_ICMPV6;
1518         pip6->hop_limit = 255;
1519         pip6->daddr = ipv6_hdr(request)->saddr;
1520         pip6->saddr = *(struct in6_addr *)n->primary_key;
1521 
1522         skb_pull(reply, sizeof(struct ipv6hdr));
1523         skb_reset_transport_header(reply);
1524 
1525         na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1526 
1527         /* Neighbor Advertisement */
1528         memset(na, 0, sizeof(*na)+na_olen);
1529         na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1530         na->icmph.icmp6_router = isrouter;
1531         na->icmph.icmp6_override = 1;
1532         na->icmph.icmp6_solicited = 1;
1533         na->target = ns->target;
1534         ether_addr_copy(&na->opt[2], n->ha);
1535         na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1536         na->opt[1] = na_olen >> 3;
1537 
1538         na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1539                 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1540                 csum_partial(na, sizeof(*na)+na_olen, 0));
1541 
1542         pip6->payload_len = htons(sizeof(*na)+na_olen);
1543 
1544         skb_push(reply, sizeof(struct ipv6hdr));
1545 
1546         reply->ip_summed = CHECKSUM_UNNECESSARY;
1547 
1548         return reply;
1549 }
1550 
1551 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1552 {
1553         struct vxlan_dev *vxlan = netdev_priv(dev);
1554         struct nd_msg *msg;
1555         const struct ipv6hdr *iphdr;
1556         const struct in6_addr *saddr, *daddr;
1557         struct neighbour *n;
1558         struct inet6_dev *in6_dev;
1559 
1560         in6_dev = __in6_dev_get(dev);
1561         if (!in6_dev)
1562                 goto out;
1563 
1564         iphdr = ipv6_hdr(skb);
1565         saddr = &iphdr->saddr;
1566         daddr = &iphdr->daddr;
1567 
1568         msg = (struct nd_msg *)skb_transport_header(skb);
1569         if (msg->icmph.icmp6_code != 0 ||
1570             msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1571                 goto out;
1572 
1573         if (ipv6_addr_loopback(daddr) ||
1574             ipv6_addr_is_multicast(&msg->target))
1575                 goto out;
1576 
1577         n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
1578 
1579         if (n) {
1580                 struct vxlan_fdb *f;
1581                 struct sk_buff *reply;
1582 
1583                 if (!(n->nud_state & NUD_CONNECTED)) {
1584                         neigh_release(n);
1585                         goto out;
1586                 }
1587 
1588                 f = vxlan_find_mac(vxlan, n->ha);
1589                 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1590                         /* bridge-local neighbor */
1591                         neigh_release(n);
1592                         goto out;
1593                 }
1594 
1595                 reply = vxlan_na_create(skb, n,
1596                                         !!(f ? f->flags & NTF_ROUTER : 0));
1597 
1598                 neigh_release(n);
1599 
1600                 if (reply == NULL)
1601                         goto out;
1602 
1603                 if (netif_rx_ni(reply) == NET_RX_DROP)
1604                         dev->stats.rx_dropped++;
1605 
1606         } else if (vxlan->flags & VXLAN_F_L3MISS) {
1607                 union vxlan_addr ipa = {
1608                         .sin6.sin6_addr = msg->target,
1609                         .sin6.sin6_family = AF_INET6,
1610                 };
1611 
1612                 vxlan_ip_miss(dev, &ipa);
1613         }
1614 
1615 out:
1616         consume_skb(skb);
1617         return NETDEV_TX_OK;
1618 }
1619 #endif
1620 
1621 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1622 {
1623         struct vxlan_dev *vxlan = netdev_priv(dev);
1624         struct neighbour *n;
1625 
1626         if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1627                 return false;
1628 
1629         n = NULL;
1630         switch (ntohs(eth_hdr(skb)->h_proto)) {
1631         case ETH_P_IP:
1632         {
1633                 struct iphdr *pip;
1634 
1635                 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1636                         return false;
1637                 pip = ip_hdr(skb);
1638                 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1639                 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1640                         union vxlan_addr ipa = {
1641                                 .sin.sin_addr.s_addr = pip->daddr,
1642                                 .sin.sin_family = AF_INET,
1643                         };
1644 
1645                         vxlan_ip_miss(dev, &ipa);
1646                         return false;
1647                 }
1648 
1649                 break;
1650         }
1651 #if IS_ENABLED(CONFIG_IPV6)
1652         case ETH_P_IPV6:
1653         {
1654                 struct ipv6hdr *pip6;
1655 
1656                 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1657                         return false;
1658                 pip6 = ipv6_hdr(skb);
1659                 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1660                 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1661                         union vxlan_addr ipa = {
1662                                 .sin6.sin6_addr = pip6->daddr,
1663                                 .sin6.sin6_family = AF_INET6,
1664                         };
1665 
1666                         vxlan_ip_miss(dev, &ipa);
1667                         return false;
1668                 }
1669 
1670                 break;
1671         }
1672 #endif
1673         default:
1674                 return false;
1675         }
1676 
1677         if (n) {
1678                 bool diff;
1679 
1680                 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1681                 if (diff) {
1682                         memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1683                                 dev->addr_len);
1684                         memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1685                 }
1686                 neigh_release(n);
1687                 return diff;
1688         }
1689 
1690         return false;
1691 }
1692 
1693 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
1694                                 struct vxlan_metadata *md)
1695 {
1696         struct vxlanhdr_gbp *gbp;
1697 
1698         if (!md->gbp)
1699                 return;
1700 
1701         gbp = (struct vxlanhdr_gbp *)vxh;
1702         vxh->vx_flags |= VXLAN_HF_GBP;
1703 
1704         if (md->gbp & VXLAN_GBP_DONT_LEARN)
1705                 gbp->dont_learn = 1;
1706 
1707         if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1708                 gbp->policy_applied = 1;
1709 
1710         gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1711 }
1712 
1713 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1714                                __be16 protocol)
1715 {
1716         struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1717 
1718         gpe->np_applied = 1;
1719 
1720         switch (protocol) {
1721         case htons(ETH_P_IP):
1722                 gpe->next_protocol = VXLAN_GPE_NP_IPV4;
1723                 return 0;
1724         case htons(ETH_P_IPV6):
1725                 gpe->next_protocol = VXLAN_GPE_NP_IPV6;
1726                 return 0;
1727         case htons(ETH_P_TEB):
1728                 gpe->next_protocol = VXLAN_GPE_NP_ETHERNET;
1729                 return 0;
1730         }
1731         return -EPFNOSUPPORT;
1732 }
1733 
1734 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1735                            int iphdr_len, __be32 vni,
1736                            struct vxlan_metadata *md, u32 vxflags,
1737                            bool udp_sum)
1738 {
1739         struct vxlanhdr *vxh;
1740         int min_headroom;
1741         int err;
1742         int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1743         __be16 inner_protocol = htons(ETH_P_TEB);
1744 
1745         if ((vxflags & VXLAN_F_REMCSUM_TX) &&
1746             skb->ip_summed == CHECKSUM_PARTIAL) {
1747                 int csum_start = skb_checksum_start_offset(skb);
1748 
1749                 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1750                     !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1751                     (skb->csum_offset == offsetof(struct udphdr, check) ||
1752                      skb->csum_offset == offsetof(struct tcphdr, check)))
1753                         type |= SKB_GSO_TUNNEL_REMCSUM;
1754         }
1755 
1756         min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1757                         + VXLAN_HLEN + iphdr_len
1758                         + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
1759 
1760         /* Need space for new headers (invalidates iph ptr) */
1761         err = skb_cow_head(skb, min_headroom);
1762         if (unlikely(err))
1763                 goto out_free;
1764 
1765         skb = vlan_hwaccel_push_inside(skb);
1766         if (WARN_ON(!skb))
1767                 return -ENOMEM;
1768 
1769         err = iptunnel_handle_offloads(skb, type);
1770         if (err)
1771                 goto out_free;
1772 
1773         vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1774         vxh->vx_flags = VXLAN_HF_VNI;
1775         vxh->vx_vni = vxlan_vni_field(vni);
1776 
1777         if (type & SKB_GSO_TUNNEL_REMCSUM) {
1778                 unsigned int start;
1779 
1780                 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1781                 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1782                 vxh->vx_flags |= VXLAN_HF_RCO;
1783 
1784                 if (!skb_is_gso(skb)) {
1785                         skb->ip_summed = CHECKSUM_NONE;
1786                         skb->encapsulation = 0;
1787                 }
1788         }
1789 
1790         if (vxflags & VXLAN_F_GBP)
1791                 vxlan_build_gbp_hdr(vxh, vxflags, md);
1792         if (vxflags & VXLAN_F_GPE) {
1793                 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1794                 if (err < 0)
1795                         goto out_free;
1796                 inner_protocol = skb->protocol;
1797         }
1798 
1799         skb_set_inner_protocol(skb, inner_protocol);
1800         return 0;
1801 
1802 out_free:
1803         kfree_skb(skb);
1804         return err;
1805 }
1806 
1807 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan,
1808                                       struct sk_buff *skb, int oif, u8 tos,
1809                                       __be32 daddr, __be32 *saddr,
1810                                       struct dst_cache *dst_cache,
1811                                       const struct ip_tunnel_info *info)
1812 {
1813         bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1814         struct rtable *rt = NULL;
1815         struct flowi4 fl4;
1816 
1817         if (tos && !info)
1818                 use_cache = false;
1819         if (use_cache) {
1820                 rt = dst_cache_get_ip4(dst_cache, saddr);
1821                 if (rt)
1822                         return rt;
1823         }
1824 
1825         memset(&fl4, 0, sizeof(fl4));
1826         fl4.flowi4_oif = oif;
1827         fl4.flowi4_tos = RT_TOS(tos);
1828         fl4.flowi4_mark = skb->mark;
1829         fl4.flowi4_proto = IPPROTO_UDP;
1830         fl4.daddr = daddr;
1831         fl4.saddr = *saddr;
1832 
1833         rt = ip_route_output_key(vxlan->net, &fl4);
1834         if (!IS_ERR(rt)) {
1835                 *saddr = fl4.saddr;
1836                 if (use_cache)
1837                         dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
1838         }
1839         return rt;
1840 }
1841 
1842 #if IS_ENABLED(CONFIG_IPV6)
1843 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1844                                           struct sk_buff *skb, int oif, u8 tos,
1845                                           __be32 label,
1846                                           const struct in6_addr *daddr,
1847                                           struct in6_addr *saddr,
1848                                           struct dst_cache *dst_cache,
1849                                           const struct ip_tunnel_info *info)
1850 {
1851         struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
1852         bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1853         struct dst_entry *ndst;
1854         struct flowi6 fl6;
1855         int err;
1856 
1857         if (!sock6)
1858                 return ERR_PTR(-EIO);
1859 
1860         if (tos && !info)
1861                 use_cache = false;
1862         if (use_cache) {
1863                 ndst = dst_cache_get_ip6(dst_cache, saddr);
1864                 if (ndst)
1865                         return ndst;
1866         }
1867 
1868         memset(&fl6, 0, sizeof(fl6));
1869         fl6.flowi6_oif = oif;
1870         fl6.daddr = *daddr;
1871         fl6.saddr = *saddr;
1872         fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
1873         fl6.flowi6_mark = skb->mark;
1874         fl6.flowi6_proto = IPPROTO_UDP;
1875 
1876         err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1877                                          sock6->sock->sk,
1878                                          &ndst, &fl6);
1879         if (err < 0)
1880                 return ERR_PTR(err);
1881 
1882         *saddr = fl6.saddr;
1883         if (use_cache)
1884                 dst_cache_set_ip6(dst_cache, ndst, saddr);
1885         return ndst;
1886 }
1887 #endif
1888 
1889 /* Bypass encapsulation if the destination is local */
1890 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1891                                struct vxlan_dev *dst_vxlan)
1892 {
1893         struct pcpu_sw_netstats *tx_stats, *rx_stats;
1894         union vxlan_addr loopback;
1895         union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
1896         struct net_device *dev = skb->dev;
1897         int len = skb->len;
1898 
1899         tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1900         rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1901         skb->pkt_type = PACKET_HOST;
1902         skb->encapsulation = 0;
1903         skb->dev = dst_vxlan->dev;
1904         __skb_pull(skb, skb_network_offset(skb));
1905 
1906         if (remote_ip->sa.sa_family == AF_INET) {
1907                 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1908                 loopback.sa.sa_family =  AF_INET;
1909 #if IS_ENABLED(CONFIG_IPV6)
1910         } else {
1911                 loopback.sin6.sin6_addr = in6addr_loopback;
1912                 loopback.sa.sa_family =  AF_INET6;
1913 #endif
1914         }
1915 
1916         if (dst_vxlan->flags & VXLAN_F_LEARN)
1917                 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
1918 
1919         u64_stats_update_begin(&tx_stats->syncp);
1920         tx_stats->tx_packets++;
1921         tx_stats->tx_bytes += len;
1922         u64_stats_update_end(&tx_stats->syncp);
1923 
1924         if (netif_rx(skb) == NET_RX_SUCCESS) {
1925                 u64_stats_update_begin(&rx_stats->syncp);
1926                 rx_stats->rx_packets++;
1927                 rx_stats->rx_bytes += len;
1928                 u64_stats_update_end(&rx_stats->syncp);
1929         } else {
1930                 dev->stats.rx_dropped++;
1931         }
1932 }
1933 
1934 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1935                            struct vxlan_rdst *rdst, bool did_rsc)
1936 {
1937         struct dst_cache *dst_cache;
1938         struct ip_tunnel_info *info;
1939         struct vxlan_dev *vxlan = netdev_priv(dev);
1940         struct sock *sk;
1941         struct rtable *rt = NULL;
1942         const struct iphdr *old_iph;
1943         union vxlan_addr *dst;
1944         union vxlan_addr remote_ip, local_ip;
1945         union vxlan_addr *src;
1946         struct vxlan_metadata _md;
1947         struct vxlan_metadata *md = &_md;
1948         __be16 src_port = 0, dst_port;
1949         __be32 vni, label;
1950         __be16 df = 0;
1951         __u8 tos, ttl;
1952         int err;
1953         u32 flags = vxlan->flags;
1954         bool udp_sum = false;
1955         bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
1956 
1957         info = skb_tunnel_info(skb);
1958 
1959         if (rdst) {
1960                 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
1961                 vni = rdst->remote_vni;
1962                 dst = &rdst->remote_ip;
1963                 src = &vxlan->cfg.saddr;
1964                 dst_cache = &rdst->dst_cache;
1965         } else {
1966                 if (!info) {
1967                         WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1968                                   dev->name);
1969                         goto drop;
1970                 }
1971                 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
1972                 vni = tunnel_id_to_key32(info->key.tun_id);
1973                 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
1974                 if (remote_ip.sa.sa_family == AF_INET) {
1975                         remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1976                         local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
1977                 } else {
1978                         remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
1979                         local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
1980                 }
1981                 dst = &remote_ip;
1982                 src = &local_ip;
1983                 dst_cache = &info->dst_cache;
1984         }
1985 
1986         if (vxlan_addr_any(dst)) {
1987                 if (did_rsc) {
1988                         /* short-circuited back to local bridge */
1989                         vxlan_encap_bypass(skb, vxlan, vxlan);
1990                         return;
1991                 }
1992                 goto drop;
1993         }
1994 
1995         old_iph = ip_hdr(skb);
1996 
1997         ttl = vxlan->cfg.ttl;
1998         if (!ttl && vxlan_addr_multicast(dst))
1999                 ttl = 1;
2000 
2001         tos = vxlan->cfg.tos;
2002         if (tos == 1)
2003                 tos = ip_tunnel_get_dsfield(old_iph, skb);
2004 
2005         label = vxlan->cfg.label;
2006         src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2007                                      vxlan->cfg.port_max, true);
2008 
2009         if (info) {
2010                 ttl = info->key.ttl;
2011                 tos = info->key.tos;
2012                 label = info->key.label;
2013                 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2014 
2015                 if (info->options_len)
2016                         md = ip_tunnel_info_opts(info);
2017         } else {
2018                 md->gbp = skb->mark;
2019         }
2020 
2021         if (dst->sa.sa_family == AF_INET) {
2022                 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2023 
2024                 if (!sock4)
2025                         goto drop;
2026                 sk = sock4->sock->sk;
2027 
2028                 rt = vxlan_get_route(vxlan, skb,
2029                                      rdst ? rdst->remote_ifindex : 0, tos,
2030                                      dst->sin.sin_addr.s_addr,
2031                                      &src->sin.sin_addr.s_addr,
2032                                      dst_cache, info);
2033                 if (IS_ERR(rt)) {
2034                         netdev_dbg(dev, "no route to %pI4\n",
2035                                    &dst->sin.sin_addr.s_addr);
2036                         dev->stats.tx_carrier_errors++;
2037                         goto tx_error;
2038                 }
2039 
2040                 if (rt->dst.dev == dev) {
2041                         netdev_dbg(dev, "circular route to %pI4\n",
2042                                    &dst->sin.sin_addr.s_addr);
2043                         dev->stats.collisions++;
2044                         goto rt_tx_error;
2045                 }
2046 
2047                 /* Bypass encapsulation if the destination is local */
2048                 if (!info && rt->rt_flags & RTCF_LOCAL &&
2049                     !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2050                         struct vxlan_dev *dst_vxlan;
2051 
2052                         ip_rt_put(rt);
2053                         dst_vxlan = vxlan_find_vni(vxlan->net, vni,
2054                                                    dst->sa.sa_family, dst_port,
2055                                                    vxlan->flags);
2056                         if (!dst_vxlan)
2057                                 goto tx_error;
2058                         vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2059                         return;
2060                 }
2061 
2062                 if (!info)
2063                         udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2064                 else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
2065                         df = htons(IP_DF);
2066 
2067                 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2068                 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2069                 err = vxlan_build_skb(skb, &rt->dst, sizeof(struct iphdr),
2070                                       vni, md, flags, udp_sum);
2071                 if (err < 0)
2072                         goto xmit_tx_error;
2073 
2074                 udp_tunnel_xmit_skb(rt, sk, skb, src->sin.sin_addr.s_addr,
2075                                     dst->sin.sin_addr.s_addr, tos, ttl, df,
2076                                     src_port, dst_port, xnet, !udp_sum);
2077 #if IS_ENABLED(CONFIG_IPV6)
2078         } else {
2079                 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2080                 struct dst_entry *ndst;
2081                 u32 rt6i_flags;
2082 
2083                 if (!sock6)
2084                         goto drop;
2085                 sk = sock6->sock->sk;
2086 
2087                 ndst = vxlan6_get_route(vxlan, skb,
2088                                         rdst ? rdst->remote_ifindex : 0, tos,
2089                                         label, &dst->sin6.sin6_addr,
2090                                         &src->sin6.sin6_addr,
2091                                         dst_cache, info);
2092                 if (IS_ERR(ndst)) {
2093                         netdev_dbg(dev, "no route to %pI6\n",
2094                                    &dst->sin6.sin6_addr);
2095                         dev->stats.tx_carrier_errors++;
2096                         goto tx_error;
2097                 }
2098 
2099                 if (ndst->dev == dev) {
2100                         netdev_dbg(dev, "circular route to %pI6\n",
2101                                    &dst->sin6.sin6_addr);
2102                         dst_release(ndst);
2103                         dev->stats.collisions++;
2104                         goto tx_error;
2105                 }
2106 
2107                 /* Bypass encapsulation if the destination is local */
2108                 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2109                 if (!info && rt6i_flags & RTF_LOCAL &&
2110                     !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2111                         struct vxlan_dev *dst_vxlan;
2112 
2113                         dst_release(ndst);
2114                         dst_vxlan = vxlan_find_vni(vxlan->net, vni,
2115                                                    dst->sa.sa_family, dst_port,
2116                                                    vxlan->flags);
2117                         if (!dst_vxlan)
2118                                 goto tx_error;
2119                         vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2120                         return;
2121                 }
2122 
2123                 if (!info)
2124                         udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2125 
2126                 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2127                 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2128                 skb_scrub_packet(skb, xnet);
2129                 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2130                                       vni, md, flags, udp_sum);
2131                 if (err < 0) {
2132                         dst_release(ndst);
2133                         dev->stats.tx_errors++;
2134                         return;
2135                 }
2136                 udp_tunnel6_xmit_skb(ndst, sk, skb, dev,
2137                                      &src->sin6.sin6_addr,
2138                                      &dst->sin6.sin6_addr, tos, ttl,
2139                                      label, src_port, dst_port, !udp_sum);
2140 #endif
2141         }
2142 
2143         return;
2144 
2145 drop:
2146         dev->stats.tx_dropped++;
2147         goto tx_free;
2148 
2149 xmit_tx_error:
2150         /* skb is already freed. */
2151         skb = NULL;
2152 rt_tx_error:
2153         ip_rt_put(rt);
2154 tx_error:
2155         dev->stats.tx_errors++;
2156 tx_free:
2157         dev_kfree_skb(skb);
2158 }
2159 
2160 /* Transmit local packets over Vxlan
2161  *
2162  * Outer IP header inherits ECN and DF from inner header.
2163  * Outer UDP destination is the VXLAN assigned port.
2164  *           source port is based on hash of flow
2165  */
2166 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2167 {
2168         struct vxlan_dev *vxlan = netdev_priv(dev);
2169         const struct ip_tunnel_info *info;
2170         struct ethhdr *eth;
2171         bool did_rsc = false;
2172         struct vxlan_rdst *rdst, *fdst = NULL;
2173         struct vxlan_fdb *f;
2174 
2175         info = skb_tunnel_info(skb);
2176 
2177         skb_reset_mac_header(skb);
2178 
2179         if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
2180                 if (info && info->mode & IP_TUNNEL_INFO_TX)
2181                         vxlan_xmit_one(skb, dev, NULL, false);
2182                 else
2183                         kfree_skb(skb);
2184                 return NETDEV_TX_OK;
2185         }
2186 
2187         if (vxlan->flags & VXLAN_F_PROXY) {
2188                 eth = eth_hdr(skb);
2189                 if (ntohs(eth->h_proto) == ETH_P_ARP)
2190                         return arp_reduce(dev, skb);
2191 #if IS_ENABLED(CONFIG_IPV6)
2192                 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2193                          pskb_may_pull(skb, sizeof(struct ipv6hdr)
2194                                        + sizeof(struct nd_msg)) &&
2195                          ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2196                                 struct nd_msg *msg;
2197 
2198                                 msg = (struct nd_msg *)skb_transport_header(skb);
2199                                 if (msg->icmph.icmp6_code == 0 &&
2200                                     msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2201                                         return neigh_reduce(dev, skb);
2202                 }
2203 #endif
2204         }
2205 
2206         eth = eth_hdr(skb);
2207         f = vxlan_find_mac(vxlan, eth->h_dest);
2208         did_rsc = false;
2209 
2210         if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
2211             (ntohs(eth->h_proto) == ETH_P_IP ||
2212              ntohs(eth->h_proto) == ETH_P_IPV6)) {
2213                 did_rsc = route_shortcircuit(dev, skb);
2214                 if (did_rsc)
2215                         f = vxlan_find_mac(vxlan, eth->h_dest);
2216         }
2217 
2218         if (f == NULL) {
2219                 f = vxlan_find_mac(vxlan, all_zeros_mac);
2220                 if (f == NULL) {
2221                         if ((vxlan->flags & VXLAN_F_L2MISS) &&
2222                             !is_multicast_ether_addr(eth->h_dest))
2223                                 vxlan_fdb_miss(vxlan, eth->h_dest);
2224 
2225                         dev->stats.tx_dropped++;
2226                         kfree_skb(skb);
2227                         return NETDEV_TX_OK;
2228                 }
2229         }
2230 
2231         list_for_each_entry_rcu(rdst, &f->remotes, list) {
2232                 struct sk_buff *skb1;
2233 
2234                 if (!fdst) {
2235                         fdst = rdst;
2236                         continue;
2237                 }
2238                 skb1 = skb_clone(skb, GFP_ATOMIC);
2239                 if (skb1)
2240                         vxlan_xmit_one(skb1, dev, rdst, did_rsc);
2241         }
2242 
2243         if (fdst)
2244                 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2245         else
2246                 kfree_skb(skb);
2247         return NETDEV_TX_OK;
2248 }
2249 
2250 /* Walk the forwarding table and purge stale entries */
2251 static void vxlan_cleanup(unsigned long arg)
2252 {
2253         struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2254         unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2255         unsigned int h;
2256 
2257         if (!netif_running(vxlan->dev))
2258                 return;
2259 
2260         for (h = 0; h < FDB_HASH_SIZE; ++h) {
2261                 struct hlist_node *p, *n;
2262 
2263                 spin_lock_bh(&vxlan->hash_lock);
2264                 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2265                         struct vxlan_fdb *f
2266                                 = container_of(p, struct vxlan_fdb, hlist);
2267                         unsigned long timeout;
2268 
2269                         if (f->state & NUD_PERMANENT)
2270                                 continue;
2271 
2272                         timeout = f->used + vxlan->cfg.age_interval * HZ;
2273                         if (time_before_eq(timeout, jiffies)) {
2274                                 netdev_dbg(vxlan->dev,
2275                                            "garbage collect %pM\n",
2276                                            f->eth_addr);
2277                                 f->state = NUD_STALE;
2278                                 vxlan_fdb_destroy(vxlan, f);
2279                         } else if (time_before(timeout, next_timer))
2280                                 next_timer = timeout;
2281                 }
2282                 spin_unlock_bh(&vxlan->hash_lock);
2283         }
2284 
2285         mod_timer(&vxlan->age_timer, next_timer);
2286 }
2287 
2288 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2289 {
2290         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2291         __be32 vni = vxlan->default_dst.remote_vni;
2292 
2293         spin_lock(&vn->sock_lock);
2294         hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
2295         spin_unlock(&vn->sock_lock);
2296 }
2297 
2298 /* Setup stats when device is created */
2299 static int vxlan_init(struct net_device *dev)
2300 {
2301         dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2302         if (!dev->tstats)
2303                 return -ENOMEM;
2304 
2305         return 0;
2306 }
2307 
2308 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
2309 {
2310         struct vxlan_fdb *f;
2311 
2312         spin_lock_bh(&vxlan->hash_lock);
2313         f = __vxlan_find_mac(vxlan, all_zeros_mac);
2314         if (f)
2315                 vxlan_fdb_destroy(vxlan, f);
2316         spin_unlock_bh(&vxlan->hash_lock);
2317 }
2318 
2319 static void vxlan_uninit(struct net_device *dev)
2320 {
2321         struct vxlan_dev *vxlan = netdev_priv(dev);
2322 
2323         vxlan_fdb_delete_default(vxlan);
2324 
2325         free_percpu(dev->tstats);
2326 }
2327 
2328 /* Start ageing timer and join group when device is brought up */
2329 static int vxlan_open(struct net_device *dev)
2330 {
2331         struct vxlan_dev *vxlan = netdev_priv(dev);
2332         int ret;
2333 
2334         ret = vxlan_sock_add(vxlan);
2335         if (ret < 0)
2336                 return ret;
2337 
2338         if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
2339                 ret = vxlan_igmp_join(vxlan);
2340                 if (ret == -EADDRINUSE)
2341                         ret = 0;
2342                 if (ret) {
2343                         vxlan_sock_release(vxlan);
2344                         return ret;
2345                 }
2346         }
2347 
2348         if (vxlan->cfg.age_interval)
2349                 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2350 
2351         return ret;
2352 }
2353 
2354 /* Purge the forwarding table */
2355 static void vxlan_flush(struct vxlan_dev *vxlan)
2356 {
2357         unsigned int h;
2358 
2359         spin_lock_bh(&vxlan->hash_lock);
2360         for (h = 0; h < FDB_HASH_SIZE; ++h) {
2361                 struct hlist_node *p, *n;
2362                 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2363                         struct vxlan_fdb *f
2364                                 = container_of(p, struct vxlan_fdb, hlist);
2365                         /* the all_zeros_mac entry is deleted at vxlan_uninit */
2366                         if (!is_zero_ether_addr(f->eth_addr))
2367                                 vxlan_fdb_destroy(vxlan, f);
2368                 }
2369         }
2370         spin_unlock_bh(&vxlan->hash_lock);
2371 }
2372 
2373 /* Cleanup timer and forwarding table on shutdown */
2374 static int vxlan_stop(struct net_device *dev)
2375 {
2376         struct vxlan_dev *vxlan = netdev_priv(dev);
2377         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2378         int ret = 0;
2379 
2380         if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2381             !vxlan_group_used(vn, vxlan))
2382                 ret = vxlan_igmp_leave(vxlan);
2383 
2384         del_timer_sync(&vxlan->age_timer);
2385 
2386         vxlan_flush(vxlan);
2387         vxlan_sock_release(vxlan);
2388 
2389         return ret;
2390 }
2391 
2392 /* Stub, nothing needs to be done. */
2393 static void vxlan_set_multicast_list(struct net_device *dev)
2394 {
2395 }
2396 
2397 static int __vxlan_change_mtu(struct net_device *dev,
2398                               struct net_device *lowerdev,
2399                               struct vxlan_rdst *dst, int new_mtu, bool strict)
2400 {
2401         int max_mtu = IP_MAX_MTU;
2402 
2403         if (lowerdev)
2404                 max_mtu = lowerdev->mtu;
2405 
2406         if (dst->remote_ip.sa.sa_family == AF_INET6)
2407                 max_mtu -= VXLAN6_HEADROOM;
2408         else
2409                 max_mtu -= VXLAN_HEADROOM;
2410 
2411         if (new_mtu < 68)
2412                 return -EINVAL;
2413 
2414         if (new_mtu > max_mtu) {
2415                 if (strict)
2416                         return -EINVAL;
2417 
2418                 new_mtu = max_mtu;
2419         }
2420 
2421         dev->mtu = new_mtu;
2422         return 0;
2423 }
2424 
2425 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2426 {
2427         struct vxlan_dev *vxlan = netdev_priv(dev);
2428         struct vxlan_rdst *dst = &vxlan->default_dst;
2429         struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2430                                                          dst->remote_ifindex);
2431         return __vxlan_change_mtu(dev, lowerdev, dst, new_mtu, true);
2432 }
2433 
2434 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2435 {
2436         struct vxlan_dev *vxlan = netdev_priv(dev);
2437         struct ip_tunnel_info *info = skb_tunnel_info(skb);
2438         __be16 sport, dport;
2439 
2440         sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2441                                   vxlan->cfg.port_max, true);
2442         dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2443 
2444         if (ip_tunnel_info_af(info) == AF_INET) {
2445                 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2446                 struct rtable *rt;
2447 
2448                 if (!sock4)
2449                         return -EINVAL;
2450                 rt = vxlan_get_route(vxlan, skb, 0, info->key.tos,
2451                                      info->key.u.ipv4.dst,
2452                                      &info->key.u.ipv4.src, NULL, info);
2453                 if (IS_ERR(rt))
2454                         return PTR_ERR(rt);
2455                 ip_rt_put(rt);
2456         } else {
2457 #if IS_ENABLED(CONFIG_IPV6)
2458                 struct dst_entry *ndst;
2459 
2460                 ndst = vxlan6_get_route(vxlan, skb, 0, info->key.tos,
2461                                         info->key.label, &info->key.u.ipv6.dst,
2462                                         &info->key.u.ipv6.src, NULL, info);
2463                 if (IS_ERR(ndst))
2464                         return PTR_ERR(ndst);
2465                 dst_release(ndst);
2466 #else /* !CONFIG_IPV6 */
2467                 return -EPFNOSUPPORT;
2468 #endif
2469         }
2470         info->key.tp_src = sport;
2471         info->key.tp_dst = dport;
2472         return 0;
2473 }
2474 
2475 static const struct net_device_ops vxlan_netdev_ether_ops = {
2476         .ndo_init               = vxlan_init,
2477         .ndo_uninit             = vxlan_uninit,
2478         .ndo_open               = vxlan_open,
2479         .ndo_stop               = vxlan_stop,
2480         .ndo_start_xmit         = vxlan_xmit,
2481         .ndo_get_stats64        = ip_tunnel_get_stats64,
2482         .ndo_set_rx_mode        = vxlan_set_multicast_list,
2483         .ndo_change_mtu         = vxlan_change_mtu,
2484         .ndo_validate_addr      = eth_validate_addr,
2485         .ndo_set_mac_address    = eth_mac_addr,
2486         .ndo_fdb_add            = vxlan_fdb_add,
2487         .ndo_fdb_del            = vxlan_fdb_delete,
2488         .ndo_fdb_dump           = vxlan_fdb_dump,
2489         .ndo_fill_metadata_dst  = vxlan_fill_metadata_dst,
2490 };
2491 
2492 static const struct net_device_ops vxlan_netdev_raw_ops = {
2493         .ndo_init               = vxlan_init,
2494         .ndo_uninit             = vxlan_uninit,
2495         .ndo_open               = vxlan_open,
2496         .ndo_stop               = vxlan_stop,
2497         .ndo_start_xmit         = vxlan_xmit,
2498         .ndo_get_stats64        = ip_tunnel_get_stats64,
2499         .ndo_change_mtu         = vxlan_change_mtu,
2500         .ndo_fill_metadata_dst  = vxlan_fill_metadata_dst,
2501 };
2502 
2503 /* Info for udev, that this is a virtual tunnel endpoint */
2504 static struct device_type vxlan_type = {
2505         .name = "vxlan",
2506 };
2507 
2508 /* Calls the ndo_udp_tunnel_add of the caller in order to
2509  * supply the listening VXLAN udp ports. Callers are expected
2510  * to implement the ndo_udp_tunnel_add.
2511  */
2512 static void vxlan_push_rx_ports(struct net_device *dev)
2513 {
2514         struct vxlan_sock *vs;
2515         struct net *net = dev_net(dev);
2516         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2517         unsigned int i;
2518 
2519         spin_lock(&vn->sock_lock);
2520         for (i = 0; i < PORT_HASH_SIZE; ++i) {
2521                 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist)
2522                         udp_tunnel_push_rx_port(dev, vs->sock,
2523                                                 (vs->flags & VXLAN_F_GPE) ?
2524                                                 UDP_TUNNEL_TYPE_VXLAN_GPE :
2525                                                 UDP_TUNNEL_TYPE_VXLAN);
2526         }
2527         spin_unlock(&vn->sock_lock);
2528 }
2529 
2530 /* Initialize the device structure. */
2531 static void vxlan_setup(struct net_device *dev)
2532 {
2533         struct vxlan_dev *vxlan = netdev_priv(dev);
2534         unsigned int h;
2535 
2536         eth_hw_addr_random(dev);
2537         ether_setup(dev);
2538 
2539         dev->destructor = free_netdev;
2540         SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2541 
2542         dev->features   |= NETIF_F_LLTX;
2543         dev->features   |= NETIF_F_SG | NETIF_F_HW_CSUM;
2544         dev->features   |= NETIF_F_RXCSUM;
2545         dev->features   |= NETIF_F_GSO_SOFTWARE;
2546 
2547         dev->vlan_features = dev->features;
2548         dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2549         dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2550         dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2551         dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2552         netif_keep_dst(dev);
2553         dev->priv_flags |= IFF_NO_QUEUE;
2554 
2555         INIT_LIST_HEAD(&vxlan->next);
2556         spin_lock_init(&vxlan->hash_lock);
2557 
2558         init_timer_deferrable(&vxlan->age_timer);
2559         vxlan->age_timer.function = vxlan_cleanup;
2560         vxlan->age_timer.data = (unsigned long) vxlan;
2561 
2562         vxlan->cfg.dst_port = htons(vxlan_port);
2563 
2564         vxlan->dev = dev;
2565 
2566         gro_cells_init(&vxlan->gro_cells, dev);
2567 
2568         for (h = 0; h < FDB_HASH_SIZE; ++h)
2569                 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2570 }
2571 
2572 static void vxlan_ether_setup(struct net_device *dev)
2573 {
2574         dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2575         dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2576         dev->netdev_ops = &vxlan_netdev_ether_ops;
2577 }
2578 
2579 static void vxlan_raw_setup(struct net_device *dev)
2580 {
2581         dev->header_ops = NULL;
2582         dev->type = ARPHRD_NONE;
2583         dev->hard_header_len = 0;
2584         dev->addr_len = 0;
2585         dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2586         dev->netdev_ops = &vxlan_netdev_raw_ops;
2587 }
2588 
2589 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2590         [IFLA_VXLAN_ID]         = { .type = NLA_U32 },
2591         [IFLA_VXLAN_GROUP]      = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2592         [IFLA_VXLAN_GROUP6]     = { .len = sizeof(struct in6_addr) },
2593         [IFLA_VXLAN_LINK]       = { .type = NLA_U32 },
2594         [IFLA_VXLAN_LOCAL]      = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2595         [IFLA_VXLAN_LOCAL6]     = { .len = sizeof(struct in6_addr) },
2596         [IFLA_VXLAN_TOS]        = { .type = NLA_U8 },
2597         [IFLA_VXLAN_TTL]        = { .type = NLA_U8 },
2598         [IFLA_VXLAN_LABEL]      = { .type = NLA_U32 },
2599         [IFLA_VXLAN_LEARNING]   = { .type = NLA_U8 },
2600         [IFLA_VXLAN_AGEING]     = { .type = NLA_U32 },
2601         [IFLA_VXLAN_LIMIT]      = { .type = NLA_U32 },
2602         [IFLA_VXLAN_PORT_RANGE] = { .len  = sizeof(struct ifla_vxlan_port_range) },
2603         [IFLA_VXLAN_PROXY]      = { .type = NLA_U8 },
2604         [IFLA_VXLAN_RSC]        = { .type = NLA_U8 },
2605         [IFLA_VXLAN_L2MISS]     = { .type = NLA_U8 },
2606         [IFLA_VXLAN_L3MISS]     = { .type = NLA_U8 },
2607         [IFLA_VXLAN_COLLECT_METADATA]   = { .type = NLA_U8 },
2608         [IFLA_VXLAN_PORT]       = { .type = NLA_U16 },
2609         [IFLA_VXLAN_UDP_CSUM]   = { .type = NLA_U8 },
2610         [IFLA_VXLAN_UDP_ZERO_CSUM6_TX]  = { .type = NLA_U8 },
2611         [IFLA_VXLAN_UDP_ZERO_CSUM6_RX]  = { .type = NLA_U8 },
2612         [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2613         [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
2614         [IFLA_VXLAN_GBP]        = { .type = NLA_FLAG, },
2615         [IFLA_VXLAN_GPE]        = { .type = NLA_FLAG, },
2616         [IFLA_VXLAN_REMCSUM_NOPARTIAL]  = { .type = NLA_FLAG },
2617 };
2618 
2619 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2620 {
2621         if (tb[IFLA_ADDRESS]) {
2622                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2623                         pr_debug("invalid link address (not ethernet)\n");
2624                         return -EINVAL;
2625                 }
2626 
2627                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2628                         pr_debug("invalid all zero ethernet address\n");
2629                         return -EADDRNOTAVAIL;
2630                 }
2631         }
2632 
2633         if (!data)
2634                 return -EINVAL;
2635 
2636         if (data[IFLA_VXLAN_ID]) {
2637                 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2638                 if (id >= VXLAN_VID_MASK)
2639                         return -ERANGE;
2640         }
2641 
2642         if (data[IFLA_VXLAN_PORT_RANGE]) {
2643                 const struct ifla_vxlan_port_range *p
2644                         = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2645 
2646                 if (ntohs(p->high) < ntohs(p->low)) {
2647                         pr_debug("port range %u .. %u not valid\n",
2648                                  ntohs(p->low), ntohs(p->high));
2649                         return -EINVAL;
2650                 }
2651         }
2652 
2653         return 0;
2654 }
2655 
2656 static void vxlan_get_drvinfo(struct net_device *netdev,
2657                               struct ethtool_drvinfo *drvinfo)
2658 {
2659         strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2660         strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2661 }
2662 
2663 static const struct ethtool_ops vxlan_ethtool_ops = {
2664         .get_drvinfo    = vxlan_get_drvinfo,
2665         .get_link       = ethtool_op_get_link,
2666 };
2667 
2668 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2669                                         __be16 port, u32 flags)
2670 {
2671         struct socket *sock;
2672         struct udp_port_cfg udp_conf;
2673         int err;
2674 
2675         memset(&udp_conf, 0, sizeof(udp_conf));
2676 
2677         if (ipv6) {
2678                 udp_conf.family = AF_INET6;
2679                 udp_conf.use_udp6_rx_checksums =
2680                     !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
2681                 udp_conf.ipv6_v6only = 1;
2682         } else {
2683                 udp_conf.family = AF_INET;
2684         }
2685 
2686         udp_conf.local_udp_port = port;
2687 
2688         /* Open UDP socket */
2689         err = udp_sock_create(net, &udp_conf, &sock);
2690         if (err < 0)
2691                 return ERR_PTR(err);
2692 
2693         return sock;
2694 }
2695 
2696 /* Create new listen socket if needed */
2697 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2698                                               __be16 port, u32 flags)
2699 {
2700         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2701         struct vxlan_sock *vs;
2702         struct socket *sock;
2703         unsigned int h;
2704         struct udp_tunnel_sock_cfg tunnel_cfg;
2705 
2706         vs = kzalloc(sizeof(*vs), GFP_KERNEL);
2707         if (!vs)
2708                 return ERR_PTR(-ENOMEM);
2709 
2710         for (h = 0; h < VNI_HASH_SIZE; ++h)
2711                 INIT_HLIST_HEAD(&vs->vni_list[h]);
2712 
2713         sock = vxlan_create_sock(net, ipv6, port, flags);
2714         if (IS_ERR(sock)) {
2715                 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2716                         PTR_ERR(sock));
2717                 kfree(vs);
2718                 return ERR_CAST(sock);
2719         }
2720 
2721         vs->sock = sock;
2722         atomic_set(&vs->refcnt, 1);
2723         vs->flags = (flags & VXLAN_F_RCV_FLAGS);
2724 
2725         spin_lock(&vn->sock_lock);
2726         hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2727         udp_tunnel_notify_add_rx_port(sock,
2728                                       (vs->flags & VXLAN_F_GPE) ?
2729                                       UDP_TUNNEL_TYPE_VXLAN_GPE :
2730                                       UDP_TUNNEL_TYPE_VXLAN);
2731         spin_unlock(&vn->sock_lock);
2732 
2733         /* Mark socket as an encapsulation socket. */
2734         memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
2735         tunnel_cfg.sk_user_data = vs;
2736         tunnel_cfg.encap_type = 1;
2737         tunnel_cfg.encap_rcv = vxlan_rcv;
2738         tunnel_cfg.encap_destroy = NULL;
2739         tunnel_cfg.gro_receive = vxlan_gro_receive;
2740         tunnel_cfg.gro_complete = vxlan_gro_complete;
2741 
2742         setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
2743 
2744         return vs;
2745 }
2746 
2747 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
2748 {
2749         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2750         struct vxlan_sock *vs = NULL;
2751 
2752         if (!vxlan->cfg.no_share) {
2753                 spin_lock(&vn->sock_lock);
2754                 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2755                                      vxlan->cfg.dst_port, vxlan->flags);
2756                 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
2757                         spin_unlock(&vn->sock_lock);
2758                         return -EBUSY;
2759                 }
2760                 spin_unlock(&vn->sock_lock);
2761         }
2762         if (!vs)
2763                 vs = vxlan_socket_create(vxlan->net, ipv6,
2764                                          vxlan->cfg.dst_port, vxlan->flags);
2765         if (IS_ERR(vs))
2766                 return PTR_ERR(vs);
2767 #if IS_ENABLED(CONFIG_IPV6)
2768         if (ipv6)
2769                 rcu_assign_pointer(vxlan->vn6_sock, vs);
2770         else
2771 #endif
2772                 rcu_assign_pointer(vxlan->vn4_sock, vs);
2773         vxlan_vs_add_dev(vs, vxlan);
2774         return 0;
2775 }
2776 
2777 static int vxlan_sock_add(struct vxlan_dev *vxlan)
2778 {
2779         bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2780         bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2781         int ret = 0;
2782 
2783         RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
2784 #if IS_ENABLED(CONFIG_IPV6)
2785         RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
2786         if (ipv6 || metadata)
2787                 ret = __vxlan_sock_add(vxlan, true);
2788 #endif
2789         if (!ret && (!ipv6 || metadata))
2790                 ret = __vxlan_sock_add(vxlan, false);
2791         if (ret < 0)
2792                 vxlan_sock_release(vxlan);
2793         return ret;
2794 }
2795 
2796 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2797                                struct vxlan_config *conf)
2798 {
2799         struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
2800         struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
2801         struct vxlan_rdst *dst = &vxlan->default_dst;
2802         unsigned short needed_headroom = ETH_HLEN;
2803         int err;
2804         bool use_ipv6 = false;
2805         __be16 default_port = vxlan->cfg.dst_port;
2806         struct net_device *lowerdev = NULL;
2807 
2808         if (conf->flags & VXLAN_F_GPE) {
2809                 /* For now, allow GPE only together with COLLECT_METADATA.
2810                  * This can be relaxed later; in such case, the other side
2811                  * of the PtP link will have to be provided.
2812                  */
2813                 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
2814                     !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
2815                         pr_info("unsupported combination of extensions\n");
2816                         return -EINVAL;
2817                 }
2818 
2819                 vxlan_raw_setup(dev);
2820         } else {
2821                 vxlan_ether_setup(dev);
2822         }
2823 
2824         vxlan->net = src_net;
2825 
2826         dst->remote_vni = conf->vni;
2827 
2828         memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
2829 
2830         /* Unless IPv6 is explicitly requested, assume IPv4 */
2831         if (!dst->remote_ip.sa.sa_family)
2832                 dst->remote_ip.sa.sa_family = AF_INET;
2833 
2834         if (dst->remote_ip.sa.sa_family == AF_INET6 ||
2835             vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2836                 if (!IS_ENABLED(CONFIG_IPV6))
2837                         return -EPFNOSUPPORT;
2838                 use_ipv6 = true;
2839                 vxlan->flags |= VXLAN_F_IPV6;
2840         }
2841 
2842         if (conf->label && !use_ipv6) {
2843                 pr_info("label only supported in use with IPv6\n");
2844                 return -EINVAL;
2845         }
2846 
2847         if (conf->remote_ifindex) {
2848                 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
2849                 dst->remote_ifindex = conf->remote_ifindex;
2850 
2851                 if (!lowerdev) {
2852                         pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
2853                         return -ENODEV;
2854                 }
2855 
2856 #if IS_ENABLED(CONFIG_IPV6)
2857                 if (use_ipv6) {
2858                         struct inet6_dev *idev = __in6_dev_get(lowerdev);
2859                         if (idev && idev->cnf.disable_ipv6) {
2860                                 pr_info("IPv6 is disabled via sysctl\n");
2861                                 return -EPERM;
2862                         }
2863                 }
2864 #endif
2865 
2866                 if (!conf->mtu)
2867                         dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2868 
2869                 needed_headroom = lowerdev->hard_header_len;
2870         } else if (vxlan_addr_multicast(&dst->remote_ip)) {
2871                 pr_info("multicast destination requires interface to be specified\n");
2872                 return -EINVAL;
2873         }
2874 
2875         if (conf->mtu) {
2876                 err = __vxlan_change_mtu(dev, lowerdev, dst, conf->mtu, false);
2877                 if (err)
2878                         return err;
2879         }
2880 
2881         if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2882                 needed_headroom += VXLAN6_HEADROOM;
2883         else
2884                 needed_headroom += VXLAN_HEADROOM;
2885         dev->needed_headroom = needed_headroom;
2886 
2887         memcpy(&vxlan->cfg, conf, sizeof(*conf));
2888         if (!vxlan->cfg.dst_port) {
2889                 if (conf->flags & VXLAN_F_GPE)
2890                         vxlan->cfg.dst_port = 4790; /* IANA assigned VXLAN-GPE port */
2891                 else
2892                         vxlan->cfg.dst_port = default_port;
2893         }
2894         vxlan->flags |= conf->flags;
2895 
2896         if (!vxlan->cfg.age_interval)
2897                 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2898 
2899         list_for_each_entry(tmp, &vn->vxlan_list, next) {
2900                 if (tmp->cfg.vni == conf->vni &&
2901                     (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2902                      tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2903                     tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2904                     (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2905                     (vxlan->flags & VXLAN_F_RCV_FLAGS)) {
2906                         pr_info("duplicate VNI %u\n", be32_to_cpu(conf->vni));
2907                         return -EEXIST;
2908                 }
2909         }
2910 
2911         dev->ethtool_ops = &vxlan_ethtool_ops;
2912 
2913         /* create an fdb entry for a valid default destination */
2914         if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2915                 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2916                                        &vxlan->default_dst.remote_ip,
2917                                        NUD_REACHABLE|NUD_PERMANENT,
2918                                        NLM_F_EXCL|NLM_F_CREATE,
2919                                        vxlan->cfg.dst_port,
2920                                        vxlan->default_dst.remote_vni,
2921                                        vxlan->default_dst.remote_ifindex,
2922                                        NTF_SELF);
2923                 if (err)
2924                         return err;
2925         }
2926 
2927         err = register_netdevice(dev);
2928         if (err) {
2929                 vxlan_fdb_delete_default(vxlan);
2930                 return err;
2931         }
2932 
2933         list_add(&vxlan->next, &vn->vxlan_list);
2934 
2935         return 0;
2936 }
2937 
2938 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2939                          struct nlattr *tb[], struct nlattr *data[])
2940 {
2941         struct vxlan_config conf;
2942 
2943         memset(&conf, 0, sizeof(conf));
2944 
2945         if (data[IFLA_VXLAN_ID])
2946                 conf.vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
2947 
2948         if (data[IFLA_VXLAN_GROUP]) {
2949                 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2950         } else if (data[IFLA_VXLAN_GROUP6]) {
2951                 if (!IS_ENABLED(CONFIG_IPV6))
2952                         return -EPFNOSUPPORT;
2953 
2954                 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2955                 conf.remote_ip.sa.sa_family = AF_INET6;
2956         }
2957 
2958         if (data[IFLA_VXLAN_LOCAL]) {
2959                 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2960                 conf.saddr.sa.sa_family = AF_INET;
2961         } else if (data[IFLA_VXLAN_LOCAL6]) {
2962                 if (!IS_ENABLED(CONFIG_IPV6))
2963                         return -EPFNOSUPPORT;
2964 
2965                 /* TODO: respect scope id */
2966                 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2967                 conf.saddr.sa.sa_family = AF_INET6;
2968         }
2969 
2970         if (data[IFLA_VXLAN_LINK])
2971                 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2972 
2973         if (data[IFLA_VXLAN_TOS])
2974                 conf.tos  = nla_get_u8(data[IFLA_VXLAN_TOS]);
2975 
2976         if (data[IFLA_VXLAN_TTL])
2977                 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2978 
2979         if (data[IFLA_VXLAN_LABEL])
2980                 conf.label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
2981                              IPV6_FLOWLABEL_MASK;
2982 
2983         if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
2984                 conf.flags |= VXLAN_F_LEARN;
2985 
2986         if (data[IFLA_VXLAN_AGEING])
2987                 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2988 
2989         if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2990                 conf.flags |= VXLAN_F_PROXY;
2991 
2992         if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2993                 conf.flags |= VXLAN_F_RSC;
2994 
2995         if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2996                 conf.flags |= VXLAN_F_L2MISS;
2997 
2998         if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2999                 conf.flags |= VXLAN_F_L3MISS;
3000 
3001         if (data[IFLA_VXLAN_LIMIT])
3002                 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3003 
3004         if (data[IFLA_VXLAN_COLLECT_METADATA] &&
3005             nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3006                 conf.flags |= VXLAN_F_COLLECT_METADATA;
3007 
3008         if (data[IFLA_VXLAN_PORT_RANGE]) {
3009                 const struct ifla_vxlan_port_range *p
3010                         = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3011                 conf.port_min = ntohs(p->low);
3012                 conf.port_max = ntohs(p->high);
3013         }
3014 
3015         if (data[IFLA_VXLAN_PORT])
3016                 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3017 
3018         if (data[IFLA_VXLAN_UDP_CSUM] &&
3019             !nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3020                 conf.flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3021 
3022         if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
3023             nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
3024                 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
3025 
3026         if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
3027             nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
3028                 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
3029 
3030         if (data[IFLA_VXLAN_REMCSUM_TX] &&
3031             nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
3032                 conf.flags |= VXLAN_F_REMCSUM_TX;
3033 
3034         if (data[IFLA_VXLAN_REMCSUM_RX] &&
3035             nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3036                 conf.flags |= VXLAN_F_REMCSUM_RX;
3037 
3038         if (data[IFLA_VXLAN_GBP])
3039                 conf.flags |= VXLAN_F_GBP;
3040 
3041         if (data[IFLA_VXLAN_GPE])
3042                 conf.flags |= VXLAN_F_GPE;
3043 
3044         if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
3045                 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
3046 
3047         if (tb[IFLA_MTU])
3048                 conf.mtu = nla_get_u32(tb[IFLA_MTU]);
3049 
3050         return vxlan_dev_configure(src_net, dev, &conf);
3051 }
3052 
3053 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3054 {
3055         struct vxlan_dev *vxlan = netdev_priv(dev);
3056         struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3057 
3058         spin_lock(&vn->sock_lock);
3059         if (!hlist_unhashed(&vxlan->hlist))
3060                 hlist_del_rcu(&vxlan->hlist);
3061         spin_unlock(&vn->sock_lock);
3062 
3063         gro_cells_destroy(&vxlan->gro_cells);
3064         list_del(&vxlan->next);
3065         unregister_netdevice_queue(dev, head);
3066 }
3067 
3068 static size_t vxlan_get_size(const struct net_device *dev)
3069 {
3070 
3071         return nla_total_size(sizeof(__u32)) +  /* IFLA_VXLAN_ID */
3072                 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
3073                 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
3074                 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
3075                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_TTL */
3076                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_TOS */
3077                 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
3078                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_LEARNING */
3079                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_PROXY */
3080                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_RSC */
3081                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_L2MISS */
3082                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_L3MISS */
3083                 nla_total_size(sizeof(__u8)) +  /* IFLA_VXLAN_COLLECT_METADATA */
3084                 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3085                 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
3086                 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
3087                 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3088                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3089                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3090                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3091                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3092                 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
3093                 0;
3094 }
3095 
3096 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3097 {
3098         const struct vxlan_dev *vxlan = netdev_priv(dev);
3099         const struct vxlan_rdst *dst = &vxlan->default_dst;
3100         struct ifla_vxlan_port_range ports = {
3101                 .low =  htons(vxlan->cfg.port_min),
3102                 .high = htons(vxlan->cfg.port_max),
3103         };
3104 
3105         if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
3106                 goto nla_put_failure;
3107 
3108         if (!vxlan_addr_any(&dst->remote_ip)) {
3109                 if (dst->remote_ip.sa.sa_family == AF_INET) {
3110                         if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3111                                             dst->remote_ip.sin.sin_addr.s_addr))
3112                                 goto nla_put_failure;
3113 #if IS_ENABLED(CONFIG_IPV6)
3114                 } else {
3115                         if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3116                                              &dst->remote_ip.sin6.sin6_addr))
3117                                 goto nla_put_failure;
3118 #endif
3119                 }
3120         }
3121 
3122         if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
3123                 goto nla_put_failure;
3124 
3125         if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3126                 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
3127                         if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
3128                                             vxlan->cfg.saddr.sin.sin_addr.s_addr))
3129                                 goto nla_put_failure;
3130 #if IS_ENABLED(CONFIG_IPV6)
3131                 } else {
3132                         if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
3133                                              &vxlan->cfg.saddr.sin6.sin6_addr))
3134                                 goto nla_put_failure;
3135 #endif
3136                 }
3137         }
3138 
3139         if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3140             nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
3141             nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
3142             nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3143                         !!(vxlan->flags & VXLAN_F_LEARN)) ||
3144             nla_put_u8(skb, IFLA_VXLAN_PROXY,
3145                         !!(vxlan->flags & VXLAN_F_PROXY)) ||
3146             nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3147             nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3148                         !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3149             nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3150                         !!(vxlan->flags & VXLAN_F_L3MISS)) ||
3151             nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3152                        !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
3153             nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3154             nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3155             nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
3156             nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
3157                         !(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
3158             nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3159                         !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3160             nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
3161                         !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3162             nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3163                         !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3164             nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3165                         !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
3166                 goto nla_put_failure;
3167 
3168         if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3169                 goto nla_put_failure;
3170 
3171         if (vxlan->flags & VXLAN_F_GBP &&
3172             nla_put_flag(skb, IFLA_VXLAN_GBP))
3173                 goto nla_put_failure;
3174 
3175         if (vxlan->flags & VXLAN_F_GPE &&
3176             nla_put_flag(skb, IFLA_VXLAN_GPE))
3177                 goto nla_put_failure;
3178 
3179         if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3180             nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3181                 goto nla_put_failure;
3182 
3183         return 0;
3184 
3185 nla_put_failure:
3186         return -EMSGSIZE;
3187 }
3188 
3189 static struct net *vxlan_get_link_net(const struct net_device *dev)
3190 {
3191         struct vxlan_dev *vxlan = netdev_priv(dev);
3192 
3193         return vxlan->net;
3194 }
3195 
3196 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3197         .kind           = "vxlan",
3198         .maxtype        = IFLA_VXLAN_MAX,
3199         .policy         = vxlan_policy,
3200         .priv_size      = sizeof(struct vxlan_dev),
3201         .setup          = vxlan_setup,
3202         .validate       = vxlan_validate,
3203         .newlink        = vxlan_newlink,
3204         .dellink        = vxlan_dellink,
3205         .get_size       = vxlan_get_size,
3206         .fill_info      = vxlan_fill_info,
3207         .get_link_net   = vxlan_get_link_net,
3208 };
3209 
3210 struct net_device *vxlan_dev_create(struct net *net, const char *name,
3211                                     u8 name_assign_type,
3212                                     struct vxlan_config *conf)
3213 {
3214         struct nlattr *tb[IFLA_MAX + 1];
3215         struct net_device *dev;
3216         int err;
3217 
3218         memset(&tb, 0, sizeof(tb));
3219 
3220         dev = rtnl_create_link(net, name, name_assign_type,
3221                                &vxlan_link_ops, tb);
3222         if (IS_ERR(dev))
3223                 return dev;
3224 
3225         err = vxlan_dev_configure(net, dev, conf);
3226         if (err < 0) {
3227                 free_netdev(dev);
3228                 return ERR_PTR(err);
3229         }
3230 
3231         err = rtnl_configure_link(dev, NULL);
3232         if (err < 0) {
3233                 LIST_HEAD(list_kill);
3234 
3235                 vxlan_dellink(dev, &list_kill);
3236                 unregister_netdevice_many(&list_kill);
3237                 return ERR_PTR(err);
3238         }
3239 
3240         return dev;
3241 }
3242 EXPORT_SYMBOL_GPL(vxlan_dev_create);
3243 
3244 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3245                                              struct net_device *dev)
3246 {
3247         struct vxlan_dev *vxlan, *next;
3248         LIST_HEAD(list_kill);
3249 
3250         list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3251                 struct vxlan_rdst *dst = &vxlan->default_dst;
3252 
3253                 /* In case we created vxlan device with carrier
3254                  * and we loose the carrier due to module unload
3255                  * we also need to remove vxlan device. In other
3256                  * cases, it's not necessary and remote_ifindex
3257                  * is 0 here, so no matches.
3258                  */
3259                 if (dst->remote_ifindex == dev->ifindex)
3260                         vxlan_dellink(vxlan->dev, &list_kill);
3261         }
3262 
3263         unregister_netdevice_many(&list_kill);
3264 }
3265 
3266 static int vxlan_netdevice_event(struct notifier_block *unused,
3267                                  unsigned long event, void *ptr)
3268 {
3269         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3270         struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
3271 
3272         if (event == NETDEV_UNREGISTER)
3273                 vxlan_handle_lowerdev_unregister(vn, dev);
3274         else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
3275                 vxlan_push_rx_ports(dev);
3276 
3277         return NOTIFY_DONE;
3278 }
3279 
3280 static struct notifier_block vxlan_notifier_block __read_mostly = {
3281         .notifier_call = vxlan_netdevice_event,
3282 };
3283 
3284 static __net_init int vxlan_init_net(struct net *net)
3285 {
3286         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3287         unsigned int h;
3288 
3289         INIT_LIST_HEAD(&vn->vxlan_list);
3290         spin_lock_init(&vn->sock_lock);
3291 
3292         for (h = 0; h < PORT_HASH_SIZE; ++h)
3293                 INIT_HLIST_HEAD(&vn->sock_list[h]);
3294 
3295         return 0;
3296 }
3297 
3298 static void __net_exit vxlan_exit_net(struct net *net)
3299 {
3300         struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3301         struct vxlan_dev *vxlan, *next;
3302         struct net_device *dev, *aux;
3303         LIST_HEAD(list);
3304 
3305         rtnl_lock();
3306         for_each_netdev_safe(net, dev, aux)
3307                 if (dev->rtnl_link_ops == &vxlan_link_ops)
3308                         unregister_netdevice_queue(dev, &list);
3309 
3310         list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3311                 /* If vxlan->dev is in the same netns, it has already been added
3312                  * to the list by the previous loop.
3313                  */
3314                 if (!net_eq(dev_net(vxlan->dev), net)) {
3315                         gro_cells_destroy(&vxlan->gro_cells);
3316                         unregister_netdevice_queue(vxlan->dev, &list);
3317                 }
3318         }
3319 
3320         unregister_netdevice_many(&list);
3321         rtnl_unlock();
3322 }
3323 
3324 static struct pernet_operations vxlan_net_ops = {
3325         .init = vxlan_init_net,
3326         .exit = vxlan_exit_net,
3327         .id   = &vxlan_net_id,
3328         .size = sizeof(struct vxlan_net),
3329 };
3330 
3331 static int __init vxlan_init_module(void)
3332 {
3333         int rc;
3334 
3335         get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3336 
3337         rc = register_pernet_subsys(&vxlan_net_ops);
3338         if (rc)
3339                 goto out1;
3340 
3341         rc = register_netdevice_notifier(&vxlan_notifier_block);
3342         if (rc)
3343                 goto out2;
3344 
3345         rc = rtnl_link_register(&vxlan_link_ops);
3346         if (rc)
3347                 goto out3;
3348 
3349         return 0;
3350 out3:
3351         unregister_netdevice_notifier(&vxlan_notifier_block);
3352 out2:
3353         unregister_pernet_subsys(&vxlan_net_ops);
3354 out1:
3355         return rc;
3356 }
3357 late_initcall(vxlan_init_module);
3358 
3359 static void __exit vxlan_cleanup_module(void)
3360 {
3361         rtnl_link_unregister(&vxlan_link_ops);
3362         unregister_netdevice_notifier(&vxlan_notifier_block);
3363         unregister_pernet_subsys(&vxlan_net_ops);
3364         /* rcu_barrier() is called by netns */
3365 }
3366 module_exit(vxlan_cleanup_module);
3367 
3368 MODULE_LICENSE("GPL");
3369 MODULE_VERSION(VXLAN_VERSION);
3370 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3371 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3372 MODULE_ALIAS_RTNL_LINK("vxlan");
3373 

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