?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 3.12.70
[linux/fpc-iii.git] / drivers / net / vxlan.c
blob73790abf0c2ab33eca5c77b39d9b30db4bbcb8c9
1 /*
2 * VXLAN: Virtual eXtensible Local Area Network
3 *
4 * Copyright (c) 2012-2013 Vyatta Inc.
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
9 */
10
11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13 #include <linux/kernel.h>
14 #include <linux/types.h>
15 #include <linux/module.h>
16 #include <linux/errno.h>
17 #include <linux/slab.h>
18 #include <linux/skbuff.h>
19 #include <linux/rculist.h>
20 #include <linux/netdevice.h>
21 #include <linux/in.h>
22 #include <linux/ip.h>
23 #include <linux/udp.h>
24 #include <linux/igmp.h>
25 #include <linux/etherdevice.h>
26 #include <linux/if_ether.h>
27 #include <linux/if_vlan.h>
28 #include <linux/hash.h>
29 #include <linux/ethtool.h>
30 #include <net/arp.h>
31 #include <net/ndisc.h>
32 #include <net/ip.h>
33 #include <net/ip_tunnels.h>
34 #include <net/icmp.h>
35 #include <net/udp.h>
36 #include <net/rtnetlink.h>
37 #include <net/route.h>
38 #include <net/dsfield.h>
39 #include <net/inet_ecn.h>
40 #include <net/net_namespace.h>
41 #include <net/netns/generic.h>
42 #include <net/vxlan.h>
43 #if IS_ENABLED(CONFIG_IPV6)
44 #include <net/ipv6.h>
45 #include <net/addrconf.h>
46 #include <net/ip6_tunnel.h>
47 #include <net/ip6_checksum.h>
48 #endif
49
50 #define VXLAN_VERSION "0.1"
51
52 #define PORT_HASH_BITS 8
53 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
54 #define VNI_HASH_BITS 10
55 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
56 #define FDB_HASH_BITS 8
57 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
58 #define FDB_AGE_DEFAULT 300 /* 5 min */
59 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
60
61 #define VXLAN_N_VID (1u << 24)
62 #define VXLAN_VID_MASK (VXLAN_N_VID - 1)
63 /* IP header + UDP + VXLAN + Ethernet header */
64 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
65 /* IPv6 header + UDP + VXLAN + Ethernet header */
66 #define VXLAN6_HEADROOM (40 + 8 + 8 + 14)
67 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
68
69 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
70
71 /* VXLAN protocol header */
72 struct vxlanhdr {
73 __be32 vx_flags;
74 __be32 vx_vni;
75 };
76
77 /* UDP port for VXLAN traffic.
78 * The IANA assigned port is 4789, but the Linux default is 8472
79 * for compatibility with early adopters.
80 */
81 static unsigned short vxlan_port __read_mostly = 8472;
82 module_param_named(udp_port, vxlan_port, ushort, 0444);
83 MODULE_PARM_DESC(udp_port, "Destination UDP port");
84
85 static bool log_ecn_error = true;
86 module_param(log_ecn_error, bool, 0644);
87 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
88
89 static int vxlan_net_id;
90
91 static const u8 all_zeros_mac[ETH_ALEN];
92
93 /* per-network namespace private data for this module */
94 struct vxlan_net {
95 struct list_head vxlan_list;
96 struct hlist_head sock_list[PORT_HASH_SIZE];
97 spinlock_t sock_lock;
98 };
99
100 union vxlan_addr {
101 struct sockaddr_in sin;
102 struct sockaddr_in6 sin6;
103 struct sockaddr sa;
104 };
105
106 struct vxlan_rdst {
107 union vxlan_addr remote_ip;
108 __be16 remote_port;
109 u32 remote_vni;
110 u32 remote_ifindex;
111 struct list_head list;
112 struct rcu_head rcu;
113 };
114
115 /* Forwarding table entry */
116 struct vxlan_fdb {
117 struct hlist_node hlist; /* linked list of entries */
118 struct rcu_head rcu;
119 unsigned long updated; /* jiffies */
120 unsigned long used;
121 struct list_head remotes;
122 u16 state; /* see ndm_state */
123 u8 flags; /* see ndm_flags */
124 u8 eth_addr[ETH_ALEN];
125 };
126
127 /* Pseudo network device */
128 struct vxlan_dev {
129 struct hlist_node hlist; /* vni hash table */
130 struct list_head next; /* vxlan's per namespace list */
131 struct vxlan_sock *vn_sock; /* listening socket */
132 struct net_device *dev;
133 struct vxlan_rdst default_dst; /* default destination */
134 union vxlan_addr saddr; /* source address */
135 __be16 dst_port;
136 __u16 port_min; /* source port range */
137 __u16 port_max;
138 __u8 tos; /* TOS override */
139 __u8 ttl;
140 u32 flags; /* VXLAN_F_* below */
141
142 struct work_struct sock_work;
143 struct work_struct igmp_join;
144 struct work_struct igmp_leave;
145
146 unsigned long age_interval;
147 struct timer_list age_timer;
148 spinlock_t hash_lock;
149 unsigned int addrcnt;
150 unsigned int addrmax;
151
152 struct hlist_head fdb_head[FDB_HASH_SIZE];
153 };
154
155 #define VXLAN_F_LEARN 0x01
156 #define VXLAN_F_PROXY 0x02
157 #define VXLAN_F_RSC 0x04
158 #define VXLAN_F_L2MISS 0x08
159 #define VXLAN_F_L3MISS 0x10
160 #define VXLAN_F_IPV6 0x20 /* internal flag */
161
162 /* salt for hash table */
163 static u32 vxlan_salt __read_mostly;
164 static struct workqueue_struct *vxlan_wq;
165
166 static void vxlan_sock_work(struct work_struct *work);
167
168 #if IS_ENABLED(CONFIG_IPV6)
169 static inline
170 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
171 {
172 if (a->sa.sa_family != b->sa.sa_family)
173 return false;
174 if (a->sa.sa_family == AF_INET6)
175 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
176 else
177 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
178 }
179
180 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
181 {
182 if (ipa->sa.sa_family == AF_INET6)
183 return ipv6_addr_any(&ipa->sin6.sin6_addr);
184 else
185 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
186 }
187
188 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
189 {
190 if (ipa->sa.sa_family == AF_INET6)
191 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
192 else
193 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
194 }
195
196 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
197 {
198 if (nla_len(nla) >= sizeof(struct in6_addr)) {
199 nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr));
200 ip->sa.sa_family = AF_INET6;
201 return 0;
202 } else if (nla_len(nla) >= sizeof(__be32)) {
203 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
204 ip->sa.sa_family = AF_INET;
205 return 0;
206 } else {
207 return -EAFNOSUPPORT;
208 }
209 }
210
211 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
212 const union vxlan_addr *ip)
213 {
214 if (ip->sa.sa_family == AF_INET6)
215 return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr);
216 else
217 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
218 }
219
220 #else /* !CONFIG_IPV6 */
221
222 static inline
223 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
224 {
225 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
226 }
227
228 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
229 {
230 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
231 }
232
233 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
234 {
235 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
236 }
237
238 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
239 {
240 if (nla_len(nla) >= sizeof(struct in6_addr)) {
241 return -EAFNOSUPPORT;
242 } else if (nla_len(nla) >= sizeof(__be32)) {
243 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
244 ip->sa.sa_family = AF_INET;
245 return 0;
246 } else {
247 return -EAFNOSUPPORT;
248 }
249 }
250
251 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
252 const union vxlan_addr *ip)
253 {
254 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
255 }
256 #endif
257
258 /* Virtual Network hash table head */
259 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
260 {
261 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
262 }
263
264 /* Socket hash table head */
265 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
266 {
267 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
268
269 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
270 }
271
272 /* First remote destination for a forwarding entry.
273 * Guaranteed to be non-NULL because remotes are never deleted.
274 */
275 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
276 {
277 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
278 }
279
280 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
281 {
282 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
283 }
284
285 /* Find VXLAN socket based on network namespace, address family and UDP port */
286 static struct vxlan_sock *vxlan_find_sock(struct net *net,
287 sa_family_t family, __be16 port)
288 {
289 struct vxlan_sock *vs;
290
291 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
292 if (inet_sk(vs->sock->sk)->inet_sport == port &&
293 inet_sk(vs->sock->sk)->sk.sk_family == family)
294 return vs;
295 }
296 return NULL;
297 }
298
299 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
300 {
301 struct vxlan_dev *vxlan;
302
303 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
304 if (vxlan->default_dst.remote_vni == id)
305 return vxlan;
306 }
307
308 return NULL;
309 }
310
311 /* Look up VNI in a per net namespace table */
312 static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id,
313 sa_family_t family, __be16 port)
314 {
315 struct vxlan_sock *vs;
316
317 vs = vxlan_find_sock(net, family, port);
318 if (!vs)
319 return NULL;
320
321 return vxlan_vs_find_vni(vs, id);
322 }
323
324 /* Fill in neighbour message in skbuff. */
325 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
326 const struct vxlan_fdb *fdb,
327 u32 portid, u32 seq, int type, unsigned int flags,
328 const struct vxlan_rdst *rdst)
329 {
330 unsigned long now = jiffies;
331 struct nda_cacheinfo ci;
332 struct nlmsghdr *nlh;
333 struct ndmsg *ndm;
334 bool send_ip, send_eth;
335
336 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
337 if (nlh == NULL)
338 return -EMSGSIZE;
339
340 ndm = nlmsg_data(nlh);
341 memset(ndm, 0, sizeof(*ndm));
342
343 send_eth = send_ip = true;
344
345 if (type == RTM_GETNEIGH) {
346 ndm->ndm_family = AF_INET;
347 send_ip = !vxlan_addr_any(&rdst->remote_ip);
348 send_eth = !is_zero_ether_addr(fdb->eth_addr);
349 } else
350 ndm->ndm_family = AF_BRIDGE;
351 ndm->ndm_state = fdb->state;
352 ndm->ndm_ifindex = vxlan->dev->ifindex;
353 ndm->ndm_flags = fdb->flags;
354 ndm->ndm_type = NDA_DST;
355
356 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
357 goto nla_put_failure;
358
359 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
360 goto nla_put_failure;
361
362 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
363 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
364 goto nla_put_failure;
365 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
366 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
367 goto nla_put_failure;
368 if (rdst->remote_ifindex &&
369 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
370 goto nla_put_failure;
371
372 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
373 ci.ndm_confirmed = 0;
374 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
375 ci.ndm_refcnt = 0;
376
377 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
378 goto nla_put_failure;
379
380 return nlmsg_end(skb, nlh);
381
382 nla_put_failure:
383 nlmsg_cancel(skb, nlh);
384 return -EMSGSIZE;
385 }
386
387 static inline size_t vxlan_nlmsg_size(void)
388 {
389 return NLMSG_ALIGN(sizeof(struct ndmsg))
390 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
391 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
392 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
393 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
394 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
395 + nla_total_size(sizeof(struct nda_cacheinfo));
396 }
397
398 static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
399 struct vxlan_fdb *fdb, int type)
400 {
401 struct net *net = dev_net(vxlan->dev);
402 struct sk_buff *skb;
403 int err = -ENOBUFS;
404
405 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
406 if (skb == NULL)
407 goto errout;
408
409 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0,
410 first_remote_rtnl(fdb));
411 if (err < 0) {
412 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
413 WARN_ON(err == -EMSGSIZE);
414 kfree_skb(skb);
415 goto errout;
416 }
417
418 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
419 return;
420 errout:
421 if (err < 0)
422 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
423 }
424
425 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
426 {
427 struct vxlan_dev *vxlan = netdev_priv(dev);
428 struct vxlan_fdb f = {
429 .state = NUD_STALE,
430 };
431 struct vxlan_rdst remote = {
432 .remote_ip = *ipa, /* goes to NDA_DST */
433 .remote_vni = VXLAN_N_VID,
434 };
435
436 INIT_LIST_HEAD(&f.remotes);
437 list_add_rcu(&remote.list, &f.remotes);
438
439 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
440 }
441
442 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
443 {
444 struct vxlan_fdb f = {
445 .state = NUD_STALE,
446 };
447
448 INIT_LIST_HEAD(&f.remotes);
449 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
450
451 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
452 }
453
454 /* Hash Ethernet address */
455 static u32 eth_hash(const unsigned char *addr)
456 {
457 u64 value = get_unaligned((u64 *)addr);
458
459 /* only want 6 bytes */
460 #ifdef __BIG_ENDIAN
461 value >>= 16;
462 #else
463 value <<= 16;
464 #endif
465 return hash_64(value, FDB_HASH_BITS);
466 }
467
468 /* Hash chain to use given mac address */
469 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
470 const u8 *mac)
471 {
472 return &vxlan->fdb_head[eth_hash(mac)];
473 }
474
475 /* Look up Ethernet address in forwarding table */
476 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
477 const u8 *mac)
478
479 {
480 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
481 struct vxlan_fdb *f;
482
483 hlist_for_each_entry_rcu(f, head, hlist) {
484 if (ether_addr_equal(mac, f->eth_addr))
485 return f;
486 }
487
488 return NULL;
489 }
490
491 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
492 const u8 *mac)
493 {
494 struct vxlan_fdb *f;
495
496 f = __vxlan_find_mac(vxlan, mac);
497 if (f)
498 f->used = jiffies;
499
500 return f;
501 }
502
503 /* caller should hold vxlan->hash_lock */
504 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
505 union vxlan_addr *ip, __be16 port,
506 __u32 vni, __u32 ifindex)
507 {
508 struct vxlan_rdst *rd;
509
510 list_for_each_entry(rd, &f->remotes, list) {
511 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
512 rd->remote_port == port &&
513 rd->remote_vni == vni &&
514 rd->remote_ifindex == ifindex)
515 return rd;
516 }
517
518 return NULL;
519 }
520
521 /* Replace destination of unicast mac */
522 static int vxlan_fdb_replace(struct vxlan_fdb *f,
523 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
524 {
525 struct vxlan_rdst *rd;
526
527 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
528 if (rd)
529 return 0;
530
531 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
532 if (!rd)
533 return 0;
534 rd->remote_ip = *ip;
535 rd->remote_port = port;
536 rd->remote_vni = vni;
537 rd->remote_ifindex = ifindex;
538 return 1;
539 }
540
541 /* Add/update destinations for multicast */
542 static int vxlan_fdb_append(struct vxlan_fdb *f,
543 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
544 {
545 struct vxlan_rdst *rd;
546
547 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
548 if (rd)
549 return 0;
550
551 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
552 if (rd == NULL)
553 return -ENOBUFS;
554 rd->remote_ip = *ip;
555 rd->remote_port = port;
556 rd->remote_vni = vni;
557 rd->remote_ifindex = ifindex;
558
559 list_add_tail_rcu(&rd->list, &f->remotes);
560
561 return 1;
562 }
563
564 /* Notify netdevs that UDP port started listening */
565 static void vxlan_notify_add_rx_port(struct sock *sk)
566 {
567 struct net_device *dev;
568 struct net *net = sock_net(sk);
569 sa_family_t sa_family = sk->sk_family;
570 __be16 port = inet_sk(sk)->inet_sport;
571
572 rcu_read_lock();
573 for_each_netdev_rcu(net, dev) {
574 if (dev->netdev_ops->ndo_add_vxlan_port)
575 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
576 port);
577 }
578 rcu_read_unlock();
579 }
580
581 /* Notify netdevs that UDP port is no more listening */
582 static void vxlan_notify_del_rx_port(struct sock *sk)
583 {
584 struct net_device *dev;
585 struct net *net = sock_net(sk);
586 sa_family_t sa_family = sk->sk_family;
587 __be16 port = inet_sk(sk)->inet_sport;
588
589 rcu_read_lock();
590 for_each_netdev_rcu(net, dev) {
591 if (dev->netdev_ops->ndo_del_vxlan_port)
592 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
593 port);
594 }
595 rcu_read_unlock();
596 }
597
598 /* Add new entry to forwarding table -- assumes lock held */
599 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
600 const u8 *mac, union vxlan_addr *ip,
601 __u16 state, __u16 flags,
602 __be16 port, __u32 vni, __u32 ifindex,
603 __u8 ndm_flags)
604 {
605 struct vxlan_fdb *f;
606 int notify = 0;
607
608 f = __vxlan_find_mac(vxlan, mac);
609 if (f) {
610 if (flags & NLM_F_EXCL) {
611 netdev_dbg(vxlan->dev,
612 "lost race to create %pM\n", mac);
613 return -EEXIST;
614 }
615 if (f->state != state) {
616 f->state = state;
617 f->updated = jiffies;
618 notify = 1;
619 }
620 if (f->flags != ndm_flags) {
621 f->flags = ndm_flags;
622 f->updated = jiffies;
623 notify = 1;
624 }
625 if ((flags & NLM_F_REPLACE)) {
626 /* Only change unicasts */
627 if (!(is_multicast_ether_addr(f->eth_addr) ||
628 is_zero_ether_addr(f->eth_addr))) {
629 int rc = vxlan_fdb_replace(f, ip, port, vni,
630 ifindex);
631
632 if (rc < 0)
633 return rc;
634 notify |= rc;
635 } else
636 return -EOPNOTSUPP;
637 }
638 if ((flags & NLM_F_APPEND) &&
639 (is_multicast_ether_addr(f->eth_addr) ||
640 is_zero_ether_addr(f->eth_addr))) {
641 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
642
643 if (rc < 0)
644 return rc;
645 notify |= rc;
646 }
647 } else {
648 if (!(flags & NLM_F_CREATE))
649 return -ENOENT;
650
651 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
652 return -ENOSPC;
653
654 /* Disallow replace to add a multicast entry */
655 if ((flags & NLM_F_REPLACE) &&
656 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
657 return -EOPNOTSUPP;
658
659 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
660 f = kmalloc(sizeof(*f), GFP_ATOMIC);
661 if (!f)
662 return -ENOMEM;
663
664 notify = 1;
665 f->state = state;
666 f->flags = ndm_flags;
667 f->updated = f->used = jiffies;
668 INIT_LIST_HEAD(&f->remotes);
669 memcpy(f->eth_addr, mac, ETH_ALEN);
670
671 vxlan_fdb_append(f, ip, port, vni, ifindex);
672
673 ++vxlan->addrcnt;
674 hlist_add_head_rcu(&f->hlist,
675 vxlan_fdb_head(vxlan, mac));
676 }
677
678 if (notify)
679 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
680
681 return 0;
682 }
683
684 static void vxlan_fdb_free(struct rcu_head *head)
685 {
686 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
687 struct vxlan_rdst *rd, *nd;
688
689 list_for_each_entry_safe(rd, nd, &f->remotes, list)
690 kfree(rd);
691 kfree(f);
692 }
693
694 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
695 {
696 netdev_dbg(vxlan->dev,
697 "delete %pM\n", f->eth_addr);
698
699 --vxlan->addrcnt;
700 vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
701
702 hlist_del_rcu(&f->hlist);
703 call_rcu(&f->rcu, vxlan_fdb_free);
704 }
705
706 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
707 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
708 {
709 struct net *net = dev_net(vxlan->dev);
710 int err;
711
712 if (tb[NDA_DST]) {
713 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
714 if (err)
715 return err;
716 } else {
717 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
718 if (remote->sa.sa_family == AF_INET) {
719 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
720 ip->sa.sa_family = AF_INET;
721 #if IS_ENABLED(CONFIG_IPV6)
722 } else {
723 ip->sin6.sin6_addr = in6addr_any;
724 ip->sa.sa_family = AF_INET6;
725 #endif
726 }
727 }
728
729 if (tb[NDA_PORT]) {
730 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
731 return -EINVAL;
732 *port = nla_get_be16(tb[NDA_PORT]);
733 } else {
734 *port = vxlan->dst_port;
735 }
736
737 if (tb[NDA_VNI]) {
738 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
739 return -EINVAL;
740 *vni = nla_get_u32(tb[NDA_VNI]);
741 } else {
742 *vni = vxlan->default_dst.remote_vni;
743 }
744
745 if (tb[NDA_IFINDEX]) {
746 struct net_device *tdev;
747
748 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
749 return -EINVAL;
750 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
751 tdev = dev_get_by_index(net, *ifindex);
752 if (!tdev)
753 return -EADDRNOTAVAIL;
754 dev_put(tdev);
755 } else {
756 *ifindex = 0;
757 }
758
759 return 0;
760 }
761
762 /* Add static entry (via netlink) */
763 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
764 struct net_device *dev,
765 const unsigned char *addr, u16 flags)
766 {
767 struct vxlan_dev *vxlan = netdev_priv(dev);
768 /* struct net *net = dev_net(vxlan->dev); */
769 union vxlan_addr ip;
770 __be16 port;
771 u32 vni, ifindex;
772 int err;
773
774 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
775 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
776 ndm->ndm_state);
777 return -EINVAL;
778 }
779
780 if (tb[NDA_DST] == NULL)
781 return -EINVAL;
782
783 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
784 if (err)
785 return err;
786
787 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
788 return -EAFNOSUPPORT;
789
790 spin_lock_bh(&vxlan->hash_lock);
791 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
792 port, vni, ifindex, ndm->ndm_flags);
793 spin_unlock_bh(&vxlan->hash_lock);
794
795 return err;
796 }
797
798 /* Delete entry (via netlink) */
799 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
800 struct net_device *dev,
801 const unsigned char *addr)
802 {
803 struct vxlan_dev *vxlan = netdev_priv(dev);
804 struct vxlan_fdb *f;
805 struct vxlan_rdst *rd = NULL;
806 union vxlan_addr ip;
807 __be16 port;
808 u32 vni, ifindex;
809 int err;
810
811 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
812 if (err)
813 return err;
814
815 err = -ENOENT;
816
817 spin_lock_bh(&vxlan->hash_lock);
818 f = vxlan_find_mac(vxlan, addr);
819 if (!f)
820 goto out;
821
822 if (!vxlan_addr_any(&ip)) {
823 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
824 if (!rd)
825 goto out;
826 }
827
828 err = 0;
829
830 /* remove a destination if it's not the only one on the list,
831 * otherwise destroy the fdb entry
832 */
833 if (rd && !list_is_singular(&f->remotes)) {
834 list_del_rcu(&rd->list);
835 kfree_rcu(rd, rcu);
836 goto out;
837 }
838
839 vxlan_fdb_destroy(vxlan, f);
840
841 out:
842 spin_unlock_bh(&vxlan->hash_lock);
843
844 return err;
845 }
846
847 /* Dump forwarding table */
848 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
849 struct net_device *dev, int idx)
850 {
851 struct vxlan_dev *vxlan = netdev_priv(dev);
852 unsigned int h;
853
854 for (h = 0; h < FDB_HASH_SIZE; ++h) {
855 struct vxlan_fdb *f;
856 int err;
857
858 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
859 struct vxlan_rdst *rd;
860
861 if (idx < cb->args[0])
862 goto skip;
863
864 list_for_each_entry_rcu(rd, &f->remotes, list) {
865 err = vxlan_fdb_info(skb, vxlan, f,
866 NETLINK_CB(cb->skb).portid,
867 cb->nlh->nlmsg_seq,
868 RTM_NEWNEIGH,
869 NLM_F_MULTI, rd);
870 if (err < 0)
871 goto out;
872 }
873 skip:
874 ++idx;
875 }
876 }
877 out:
878 return idx;
879 }
880
881 /* Watch incoming packets to learn mapping between Ethernet address
882 * and Tunnel endpoint.
883 * Return true if packet is bogus and should be droppped.
884 */
885 static bool vxlan_snoop(struct net_device *dev,
886 union vxlan_addr *src_ip, const u8 *src_mac)
887 {
888 struct vxlan_dev *vxlan = netdev_priv(dev);
889 struct vxlan_fdb *f;
890
891 f = vxlan_find_mac(vxlan, src_mac);
892 if (likely(f)) {
893 struct vxlan_rdst *rdst = first_remote_rcu(f);
894
895 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
896 return false;
897
898 /* Don't migrate static entries, drop packets */
899 if (f->state & NUD_NOARP)
900 return true;
901
902 if (net_ratelimit())
903 netdev_info(dev,
904 "%pM migrated from %pIS to %pIS\n",
905 src_mac, &rdst->remote_ip, &src_ip);
906
907 rdst->remote_ip = *src_ip;
908 f->updated = jiffies;
909 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
910 } else {
911 /* learned new entry */
912 spin_lock(&vxlan->hash_lock);
913
914 /* close off race between vxlan_flush and incoming packets */
915 if (netif_running(dev))
916 vxlan_fdb_create(vxlan, src_mac, src_ip,
917 NUD_REACHABLE,
918 NLM_F_EXCL|NLM_F_CREATE,
919 vxlan->dst_port,
920 vxlan->default_dst.remote_vni,
921 0, NTF_SELF);
922 spin_unlock(&vxlan->hash_lock);
923 }
924
925 return false;
926 }
927
928 /* See if multicast group is already in use by other ID */
929 static bool vxlan_group_used(struct vxlan_net *vn, union vxlan_addr *remote_ip)
930 {
931 struct vxlan_dev *vxlan;
932
933 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
934 if (!netif_running(vxlan->dev))
935 continue;
936
937 if (vxlan_addr_equal(&vxlan->default_dst.remote_ip,
938 remote_ip))
939 return true;
940 }
941
942 return false;
943 }
944
945 static void vxlan_sock_hold(struct vxlan_sock *vs)
946 {
947 atomic_inc(&vs->refcnt);
948 }
949
950 void vxlan_sock_release(struct vxlan_sock *vs)
951 {
952 struct sock *sk = vs->sock->sk;
953 struct net *net = sock_net(sk);
954 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
955
956 if (!atomic_dec_and_test(&vs->refcnt))
957 return;
958
959 spin_lock(&vn->sock_lock);
960 hlist_del_rcu(&vs->hlist);
961 rcu_assign_sk_user_data(vs->sock->sk, NULL);
962 vxlan_notify_del_rx_port(sk);
963 spin_unlock(&vn->sock_lock);
964
965 queue_work(vxlan_wq, &vs->del_work);
966 }
967 EXPORT_SYMBOL_GPL(vxlan_sock_release);
968
969 /* Callback to update multicast group membership when first VNI on
970 * multicast asddress is brought up
971 * Done as workqueue because ip_mc_join_group acquires RTNL.
972 */
973 static void vxlan_igmp_join(struct work_struct *work)
974 {
975 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
976 struct vxlan_sock *vs = vxlan->vn_sock;
977 struct sock *sk = vs->sock->sk;
978 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
979 int ifindex = vxlan->default_dst.remote_ifindex;
980
981 lock_sock(sk);
982 if (ip->sa.sa_family == AF_INET) {
983 struct ip_mreqn mreq = {
984 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
985 .imr_ifindex = ifindex,
986 };
987
988 ip_mc_join_group(sk, &mreq);
989 #if IS_ENABLED(CONFIG_IPV6)
990 } else {
991 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
992 &ip->sin6.sin6_addr);
993 #endif
994 }
995 release_sock(sk);
996
997 vxlan_sock_release(vs);
998 dev_put(vxlan->dev);
999 }
1000
1001 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1002 static void vxlan_igmp_leave(struct work_struct *work)
1003 {
1004 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
1005 struct vxlan_sock *vs = vxlan->vn_sock;
1006 struct sock *sk = vs->sock->sk;
1007 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1008 int ifindex = vxlan->default_dst.remote_ifindex;
1009
1010 lock_sock(sk);
1011 if (ip->sa.sa_family == AF_INET) {
1012 struct ip_mreqn mreq = {
1013 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1014 .imr_ifindex = ifindex,
1015 };
1016
1017 ip_mc_leave_group(sk, &mreq);
1018 #if IS_ENABLED(CONFIG_IPV6)
1019 } else {
1020 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1021 &ip->sin6.sin6_addr);
1022 #endif
1023 }
1024
1025 release_sock(sk);
1026
1027 vxlan_sock_release(vs);
1028 dev_put(vxlan->dev);
1029 }
1030
1031 /* Callback from net/ipv4/udp.c to receive packets */
1032 static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1033 {
1034 struct vxlan_sock *vs;
1035 struct vxlanhdr *vxh;
1036 __be16 port;
1037
1038 /* Need Vxlan and inner Ethernet header to be present */
1039 if (!pskb_may_pull(skb, VXLAN_HLEN))
1040 goto drop;
1041
1042 /* Return packets with reserved bits set */
1043 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1044 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
1045 (vxh->vx_vni & htonl(0xff))) {
1046 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1047 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1048 goto drop;
1049 }
1050
1051 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1052 goto drop;
1053
1054 port = inet_sk(sk)->inet_sport;
1055
1056 vs = rcu_dereference_sk_user_data(sk);
1057 if (!vs)
1058 goto drop;
1059
1060 /* If the NIC driver gave us an encapsulated packet
1061 * with the encapsulation mark, the device checksummed it
1062 * for us. Otherwise force the upper layers to verify it.
1063 */
1064 if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation)
1065 skb->ip_summed = CHECKSUM_NONE;
1066
1067 skb->encapsulation = 0;
1068
1069 vs->rcv(vs, skb, vxh->vx_vni);
1070 return 0;
1071
1072 drop:
1073 /* Consume bad packet */
1074 kfree_skb(skb);
1075 return 0;
1076 }
1077
1078 static void vxlan_rcv(struct vxlan_sock *vs,
1079 struct sk_buff *skb, __be32 vx_vni)
1080 {
1081 struct iphdr *oip = NULL;
1082 struct ipv6hdr *oip6 = NULL;
1083 struct vxlan_dev *vxlan;
1084 struct pcpu_tstats *stats;
1085 union vxlan_addr saddr;
1086 __u32 vni;
1087 int err = 0;
1088 union vxlan_addr *remote_ip;
1089
1090 vni = ntohl(vx_vni) >> 8;
1091 /* Is this VNI defined? */
1092 vxlan = vxlan_vs_find_vni(vs, vni);
1093 if (!vxlan)
1094 goto drop;
1095
1096 remote_ip = &vxlan->default_dst.remote_ip;
1097 skb_reset_mac_header(skb);
1098 skb->protocol = eth_type_trans(skb, vxlan->dev);
1099
1100 /* Ignore packet loops (and multicast echo) */
1101 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1102 goto drop;
1103
1104 /* Re-examine inner Ethernet packet */
1105 if (remote_ip->sa.sa_family == AF_INET) {
1106 oip = ip_hdr(skb);
1107 saddr.sin.sin_addr.s_addr = oip->saddr;
1108 saddr.sa.sa_family = AF_INET;
1109 #if IS_ENABLED(CONFIG_IPV6)
1110 } else {
1111 oip6 = ipv6_hdr(skb);
1112 saddr.sin6.sin6_addr = oip6->saddr;
1113 saddr.sa.sa_family = AF_INET6;
1114 #endif
1115 }
1116
1117 if ((vxlan->flags & VXLAN_F_LEARN) &&
1118 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1119 goto drop;
1120
1121 skb_reset_network_header(skb);
1122
1123 if (oip6)
1124 err = IP6_ECN_decapsulate(oip6, skb);
1125 if (oip)
1126 err = IP_ECN_decapsulate(oip, skb);
1127
1128 if (unlikely(err)) {
1129 if (log_ecn_error) {
1130 if (oip6)
1131 net_info_ratelimited("non-ECT from %pI6\n",
1132 &oip6->saddr);
1133 if (oip)
1134 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1135 &oip->saddr, oip->tos);
1136 }
1137 if (err > 1) {
1138 ++vxlan->dev->stats.rx_frame_errors;
1139 ++vxlan->dev->stats.rx_errors;
1140 goto drop;
1141 }
1142 }
1143
1144 stats = this_cpu_ptr(vxlan->dev->tstats);
1145 u64_stats_update_begin(&stats->syncp);
1146 stats->rx_packets++;
1147 stats->rx_bytes += skb->len;
1148 u64_stats_update_end(&stats->syncp);
1149
1150 netif_rx(skb);
1151
1152 return;
1153 drop:
1154 /* Consume bad packet */
1155 kfree_skb(skb);
1156 }
1157
1158 static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1159 {
1160 struct vxlan_dev *vxlan = netdev_priv(dev);
1161 struct arphdr *parp;
1162 u8 *arpptr, *sha;
1163 __be32 sip, tip;
1164 struct neighbour *n;
1165
1166 if (dev->flags & IFF_NOARP)
1167 goto out;
1168
1169 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1170 dev->stats.tx_dropped++;
1171 goto out;
1172 }
1173 parp = arp_hdr(skb);
1174
1175 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1176 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1177 parp->ar_pro != htons(ETH_P_IP) ||
1178 parp->ar_op != htons(ARPOP_REQUEST) ||
1179 parp->ar_hln != dev->addr_len ||
1180 parp->ar_pln != 4)
1181 goto out;
1182 arpptr = (u8 *)parp + sizeof(struct arphdr);
1183 sha = arpptr;
1184 arpptr += dev->addr_len; /* sha */
1185 memcpy(&sip, arpptr, sizeof(sip));
1186 arpptr += sizeof(sip);
1187 arpptr += dev->addr_len; /* tha */
1188 memcpy(&tip, arpptr, sizeof(tip));
1189
1190 if (ipv4_is_loopback(tip) ||
1191 ipv4_is_multicast(tip))
1192 goto out;
1193
1194 n = neigh_lookup(&arp_tbl, &tip, dev);
1195
1196 if (n) {
1197 struct vxlan_fdb *f;
1198 struct sk_buff *reply;
1199
1200 if (!(n->nud_state & NUD_CONNECTED)) {
1201 neigh_release(n);
1202 goto out;
1203 }
1204
1205 f = vxlan_find_mac(vxlan, n->ha);
1206 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1207 /* bridge-local neighbor */
1208 neigh_release(n);
1209 goto out;
1210 }
1211
1212 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1213 n->ha, sha);
1214
1215 neigh_release(n);
1216
1217 if (reply == NULL)
1218 goto out;
1219
1220 skb_reset_mac_header(reply);
1221 __skb_pull(reply, skb_network_offset(reply));
1222 reply->ip_summed = CHECKSUM_UNNECESSARY;
1223 reply->pkt_type = PACKET_HOST;
1224
1225 if (netif_rx_ni(reply) == NET_RX_DROP)
1226 dev->stats.rx_dropped++;
1227 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1228 union vxlan_addr ipa = {
1229 .sin.sin_addr.s_addr = tip,
1230 .sin.sin_family = AF_INET,
1231 };
1232
1233 vxlan_ip_miss(dev, &ipa);
1234 }
1235 out:
1236 consume_skb(skb);
1237 return NETDEV_TX_OK;
1238 }
1239
1240 #if IS_ENABLED(CONFIG_IPV6)
1241
1242 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1243 struct neighbour *n, bool isrouter)
1244 {
1245 struct net_device *dev = request->dev;
1246 struct sk_buff *reply;
1247 struct nd_msg *ns, *na;
1248 struct ipv6hdr *pip6;
1249 u8 *daddr;
1250 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1251 int ns_olen;
1252 int i, len;
1253
1254 if (dev == NULL)
1255 return NULL;
1256
1257 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1258 sizeof(*na) + na_olen + dev->needed_tailroom;
1259 reply = alloc_skb(len, GFP_ATOMIC);
1260 if (reply == NULL)
1261 return NULL;
1262
1263 reply->protocol = htons(ETH_P_IPV6);
1264 reply->dev = dev;
1265 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1266 skb_push(reply, sizeof(struct ethhdr));
1267 skb_set_mac_header(reply, 0);
1268
1269 ns = (struct nd_msg *)skb_transport_header(request);
1270
1271 daddr = eth_hdr(request)->h_source;
1272 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1273 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1274 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1275 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1276 break;
1277 }
1278 }
1279
1280 /* Ethernet header */
1281 memcpy(eth_hdr(reply)->h_dest, daddr, ETH_ALEN);
1282 memcpy(eth_hdr(reply)->h_source, n->ha, ETH_ALEN);
1283 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1284 reply->protocol = htons(ETH_P_IPV6);
1285
1286 skb_pull(reply, sizeof(struct ethhdr));
1287 skb_set_network_header(reply, 0);
1288 skb_put(reply, sizeof(struct ipv6hdr));
1289
1290 /* IPv6 header */
1291
1292 pip6 = ipv6_hdr(reply);
1293 memset(pip6, 0, sizeof(struct ipv6hdr));
1294 pip6->version = 6;
1295 pip6->priority = ipv6_hdr(request)->priority;
1296 pip6->nexthdr = IPPROTO_ICMPV6;
1297 pip6->hop_limit = 255;
1298 pip6->daddr = ipv6_hdr(request)->saddr;
1299 pip6->saddr = *(struct in6_addr *)n->primary_key;
1300
1301 skb_pull(reply, sizeof(struct ipv6hdr));
1302 skb_set_transport_header(reply, 0);
1303
1304 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1305
1306 /* Neighbor Advertisement */
1307 memset(na, 0, sizeof(*na)+na_olen);
1308 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1309 na->icmph.icmp6_router = isrouter;
1310 na->icmph.icmp6_override = 1;
1311 na->icmph.icmp6_solicited = 1;
1312 na->target = ns->target;
1313 memcpy(&na->opt[2], n->ha, ETH_ALEN);
1314 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1315 na->opt[1] = na_olen >> 3;
1316
1317 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1318 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1319 csum_partial(na, sizeof(*na)+na_olen, 0));
1320
1321 pip6->payload_len = htons(sizeof(*na)+na_olen);
1322
1323 skb_push(reply, sizeof(struct ipv6hdr));
1324
1325 reply->ip_summed = CHECKSUM_UNNECESSARY;
1326
1327 return reply;
1328 }
1329
1330 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1331 {
1332 struct vxlan_dev *vxlan = netdev_priv(dev);
1333 struct nd_msg *msg;
1334 const struct ipv6hdr *iphdr;
1335 const struct in6_addr *saddr, *daddr;
1336 struct neighbour *n;
1337 struct inet6_dev *in6_dev;
1338
1339 in6_dev = __in6_dev_get(dev);
1340 if (!in6_dev)
1341 goto out;
1342
1343 iphdr = ipv6_hdr(skb);
1344 saddr = &iphdr->saddr;
1345 daddr = &iphdr->daddr;
1346
1347 msg = (struct nd_msg *)skb_transport_header(skb);
1348 if (msg->icmph.icmp6_code != 0 ||
1349 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1350 goto out;
1351
1352 if (ipv6_addr_loopback(daddr) ||
1353 ipv6_addr_is_multicast(&msg->target))
1354 goto out;
1355
1356 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
1357
1358 if (n) {
1359 struct vxlan_fdb *f;
1360 struct sk_buff *reply;
1361
1362 if (!(n->nud_state & NUD_CONNECTED)) {
1363 neigh_release(n);
1364 goto out;
1365 }
1366
1367 f = vxlan_find_mac(vxlan, n->ha);
1368 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1369 /* bridge-local neighbor */
1370 neigh_release(n);
1371 goto out;
1372 }
1373
1374 reply = vxlan_na_create(skb, n,
1375 !!(f ? f->flags & NTF_ROUTER : 0));
1376
1377 neigh_release(n);
1378
1379 if (reply == NULL)
1380 goto out;
1381
1382 if (netif_rx_ni(reply) == NET_RX_DROP)
1383 dev->stats.rx_dropped++;
1384
1385 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1386 union vxlan_addr ipa = {
1387 .sin6.sin6_addr = msg->target,
1388 .sin6.sin6_family = AF_INET6,
1389 };
1390
1391 vxlan_ip_miss(dev, &ipa);
1392 }
1393
1394 out:
1395 consume_skb(skb);
1396 return NETDEV_TX_OK;
1397 }
1398 #endif
1399
1400 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1401 {
1402 struct vxlan_dev *vxlan = netdev_priv(dev);
1403 struct neighbour *n;
1404
1405 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1406 return false;
1407
1408 n = NULL;
1409 switch (ntohs(eth_hdr(skb)->h_proto)) {
1410 case ETH_P_IP:
1411 {
1412 struct iphdr *pip;
1413
1414 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1415 return false;
1416 pip = ip_hdr(skb);
1417 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1418 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1419 union vxlan_addr ipa = {
1420 .sin.sin_addr.s_addr = pip->daddr,
1421 .sin.sin_family = AF_INET,
1422 };
1423
1424 vxlan_ip_miss(dev, &ipa);
1425 return false;
1426 }
1427
1428 break;
1429 }
1430 #if IS_ENABLED(CONFIG_IPV6)
1431 case ETH_P_IPV6:
1432 {
1433 struct ipv6hdr *pip6;
1434
1435 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1436 return false;
1437 pip6 = ipv6_hdr(skb);
1438 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1439 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1440 union vxlan_addr ipa = {
1441 .sin6.sin6_addr = pip6->daddr,
1442 .sin6.sin6_family = AF_INET6,
1443 };
1444
1445 vxlan_ip_miss(dev, &ipa);
1446 return false;
1447 }
1448
1449 break;
1450 }
1451 #endif
1452 default:
1453 return false;
1454 }
1455
1456 if (n) {
1457 bool diff;
1458
1459 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1460 if (diff) {
1461 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1462 dev->addr_len);
1463 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1464 }
1465 neigh_release(n);
1466 return diff;
1467 }
1468
1469 return false;
1470 }
1471
1472 static void vxlan_sock_put(struct sk_buff *skb)
1473 {
1474 sock_put(skb->sk);
1475 }
1476
1477 /* On transmit, associate with the tunnel socket */
1478 static void vxlan_set_owner(struct sock *sk, struct sk_buff *skb)
1479 {
1480 skb_orphan(skb);
1481 sock_hold(sk);
1482 skb->sk = sk;
1483 skb->destructor = vxlan_sock_put;
1484 }
1485
1486 /* Compute source port for outgoing packet
1487 * first choice to use L4 flow hash since it will spread
1488 * better and maybe available from hardware
1489 * secondary choice is to use jhash on the Ethernet header
1490 */
1491 __be16 vxlan_src_port(__u16 port_min, __u16 port_max, struct sk_buff *skb)
1492 {
1493 unsigned int range = (port_max - port_min) + 1;
1494 u32 hash;
1495
1496 hash = skb_get_rxhash(skb);
1497 if (!hash)
1498 hash = jhash(skb->data, 2 * ETH_ALEN,
1499 (__force u32) skb->protocol);
1500
1501 return htons((((u64) hash * range) >> 32) + port_min);
1502 }
1503 EXPORT_SYMBOL_GPL(vxlan_src_port);
1504
1505 static int handle_offloads(struct sk_buff *skb)
1506 {
1507 if (skb_is_gso(skb)) {
1508 int err = skb_unclone(skb, GFP_ATOMIC);
1509 if (unlikely(err))
1510 return err;
1511
1512 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
1513 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
1514 skb->ip_summed = CHECKSUM_NONE;
1515
1516 return 0;
1517 }
1518
1519 #if IS_ENABLED(CONFIG_IPV6)
1520 static int vxlan6_xmit_skb(struct vxlan_sock *vs,
1521 struct dst_entry *dst, struct sk_buff *skb,
1522 struct net_device *dev, struct in6_addr *saddr,
1523 struct in6_addr *daddr, __u8 prio, __u8 ttl,
1524 __be16 src_port, __be16 dst_port, __be32 vni)
1525 {
1526 struct ipv6hdr *ip6h;
1527 struct vxlanhdr *vxh;
1528 struct udphdr *uh;
1529 int min_headroom;
1530 int err;
1531
1532 if (!skb->encapsulation) {
1533 skb_reset_inner_headers(skb);
1534 skb->encapsulation = 1;
1535 }
1536
1537 skb_scrub_packet(skb, false);
1538
1539 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1540 + VXLAN_HLEN + sizeof(struct ipv6hdr)
1541 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1542
1543 /* Need space for new headers (invalidates iph ptr) */
1544 err = skb_cow_head(skb, min_headroom);
1545 if (unlikely(err))
1546 return err;
1547
1548 if (vlan_tx_tag_present(skb)) {
1549 if (WARN_ON(!__vlan_put_tag(skb,
1550 skb->vlan_proto,
1551 vlan_tx_tag_get(skb))))
1552 return -ENOMEM;
1553
1554 skb->vlan_tci = 0;
1555 }
1556
1557 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1558 vxh->vx_flags = htonl(VXLAN_FLAGS);
1559 vxh->vx_vni = vni;
1560
1561 __skb_push(skb, sizeof(*uh));
1562 skb_reset_transport_header(skb);
1563 uh = udp_hdr(skb);
1564
1565 uh->dest = dst_port;
1566 uh->source = src_port;
1567
1568 uh->len = htons(skb->len);
1569 uh->check = 0;
1570
1571 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1572 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1573 IPSKB_REROUTED);
1574 skb_dst_set(skb, dst);
1575
1576 if (!skb_is_gso(skb) && !(dst->dev->features & NETIF_F_IPV6_CSUM)) {
1577 __wsum csum = skb_checksum(skb, 0, skb->len, 0);
1578 skb->ip_summed = CHECKSUM_UNNECESSARY;
1579 uh->check = csum_ipv6_magic(saddr, daddr, skb->len,
1580 IPPROTO_UDP, csum);
1581 if (uh->check == 0)
1582 uh->check = CSUM_MANGLED_0;
1583 } else {
1584 skb->ip_summed = CHECKSUM_PARTIAL;
1585 skb->csum_start = skb_transport_header(skb) - skb->head;
1586 skb->csum_offset = offsetof(struct udphdr, check);
1587 uh->check = ~csum_ipv6_magic(saddr, daddr,
1588 skb->len, IPPROTO_UDP, 0);
1589 }
1590
1591 __skb_push(skb, sizeof(*ip6h));
1592 skb_reset_network_header(skb);
1593 ip6h = ipv6_hdr(skb);
1594 ip6h->version = 6;
1595 ip6h->priority = prio;
1596 ip6h->flow_lbl[0] = 0;
1597 ip6h->flow_lbl[1] = 0;
1598 ip6h->flow_lbl[2] = 0;
1599 ip6h->payload_len = htons(skb->len);
1600 ip6h->nexthdr = IPPROTO_UDP;
1601 ip6h->hop_limit = ttl;
1602 ip6h->daddr = *daddr;
1603 ip6h->saddr = *saddr;
1604
1605 vxlan_set_owner(vs->sock->sk, skb);
1606
1607 err = handle_offloads(skb);
1608 if (err)
1609 return err;
1610
1611 ip6tunnel_xmit(skb, dev);
1612 return 0;
1613 }
1614 #endif
1615
1616 int vxlan_xmit_skb(struct vxlan_sock *vs,
1617 struct rtable *rt, struct sk_buff *skb,
1618 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
1619 __be16 src_port, __be16 dst_port, __be32 vni)
1620 {
1621 struct vxlanhdr *vxh;
1622 struct udphdr *uh;
1623 int min_headroom;
1624 int err;
1625
1626 if (!skb->encapsulation) {
1627 skb_reset_inner_headers(skb);
1628 skb->encapsulation = 1;
1629 }
1630
1631 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1632 + VXLAN_HLEN + sizeof(struct iphdr)
1633 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1634
1635 /* Need space for new headers (invalidates iph ptr) */
1636 err = skb_cow_head(skb, min_headroom);
1637 if (unlikely(err))
1638 return err;
1639
1640 if (vlan_tx_tag_present(skb)) {
1641 if (WARN_ON(!__vlan_put_tag(skb,
1642 skb->vlan_proto,
1643 vlan_tx_tag_get(skb))))
1644 return -ENOMEM;
1645
1646 skb->vlan_tci = 0;
1647 }
1648
1649 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1650 vxh->vx_flags = htonl(VXLAN_FLAGS);
1651 vxh->vx_vni = vni;
1652
1653 __skb_push(skb, sizeof(*uh));
1654 skb_reset_transport_header(skb);
1655 uh = udp_hdr(skb);
1656
1657 uh->dest = dst_port;
1658 uh->source = src_port;
1659
1660 uh->len = htons(skb->len);
1661 uh->check = 0;
1662
1663 vxlan_set_owner(vs->sock->sk, skb);
1664
1665 err = handle_offloads(skb);
1666 if (err)
1667 return err;
1668
1669 return iptunnel_xmit(rt, skb, src, dst, IPPROTO_UDP, tos, ttl, df,
1670 false);
1671 }
1672 EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1673
1674 /* Bypass encapsulation if the destination is local */
1675 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1676 struct vxlan_dev *dst_vxlan)
1677 {
1678 struct pcpu_tstats *tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1679 struct pcpu_tstats *rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1680 union vxlan_addr loopback;
1681 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
1682 struct net_device *dev = skb->dev;
1683 int len = skb->len;
1684
1685 skb->pkt_type = PACKET_HOST;
1686 skb->encapsulation = 0;
1687 skb->dev = dst_vxlan->dev;
1688 __skb_pull(skb, skb_network_offset(skb));
1689
1690 if (remote_ip->sa.sa_family == AF_INET) {
1691 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1692 loopback.sa.sa_family = AF_INET;
1693 #if IS_ENABLED(CONFIG_IPV6)
1694 } else {
1695 loopback.sin6.sin6_addr = in6addr_loopback;
1696 loopback.sa.sa_family = AF_INET6;
1697 #endif
1698 }
1699
1700 if (dst_vxlan->flags & VXLAN_F_LEARN)
1701 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
1702
1703 u64_stats_update_begin(&tx_stats->syncp);
1704 tx_stats->tx_packets++;
1705 tx_stats->tx_bytes += len;
1706 u64_stats_update_end(&tx_stats->syncp);
1707
1708 if (netif_rx(skb) == NET_RX_SUCCESS) {
1709 u64_stats_update_begin(&rx_stats->syncp);
1710 rx_stats->rx_packets++;
1711 rx_stats->rx_bytes += len;
1712 u64_stats_update_end(&rx_stats->syncp);
1713 } else {
1714 dev->stats.rx_dropped++;
1715 }
1716 }
1717
1718 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1719 struct vxlan_rdst *rdst, bool did_rsc)
1720 {
1721 struct vxlan_dev *vxlan = netdev_priv(dev);
1722 struct rtable *rt = NULL;
1723 const struct iphdr *old_iph;
1724 struct flowi4 fl4;
1725 union vxlan_addr *dst;
1726 __be16 src_port = 0, dst_port;
1727 u32 vni;
1728 __be16 df = 0;
1729 __u8 tos, ttl;
1730 int err;
1731
1732 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
1733 vni = rdst->remote_vni;
1734 dst = &rdst->remote_ip;
1735
1736 if (vxlan_addr_any(dst)) {
1737 if (did_rsc) {
1738 /* short-circuited back to local bridge */
1739 vxlan_encap_bypass(skb, vxlan, vxlan);
1740 return;
1741 }
1742 goto drop;
1743 }
1744
1745 old_iph = ip_hdr(skb);
1746
1747 ttl = vxlan->ttl;
1748 if (!ttl && vxlan_addr_multicast(dst))
1749 ttl = 1;
1750
1751 tos = vxlan->tos;
1752 if (tos == 1)
1753 tos = ip_tunnel_get_dsfield(old_iph, skb);
1754
1755 src_port = vxlan_src_port(vxlan->port_min, vxlan->port_max, skb);
1756
1757 if (dst->sa.sa_family == AF_INET) {
1758 memset(&fl4, 0, sizeof(fl4));
1759 fl4.flowi4_oif = rdst->remote_ifindex;
1760 fl4.flowi4_tos = RT_TOS(tos);
1761 fl4.daddr = dst->sin.sin_addr.s_addr;
1762 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1763
1764 rt = ip_route_output_key(dev_net(dev), &fl4);
1765 if (IS_ERR(rt)) {
1766 netdev_dbg(dev, "no route to %pI4\n",
1767 &dst->sin.sin_addr.s_addr);
1768 dev->stats.tx_carrier_errors++;
1769 goto tx_error;
1770 }
1771
1772 if (rt->dst.dev == dev) {
1773 netdev_dbg(dev, "circular route to %pI4\n",
1774 &dst->sin.sin_addr.s_addr);
1775 dev->stats.collisions++;
1776 goto rt_tx_error;
1777 }
1778
1779 /* Bypass encapsulation if the destination is local */
1780 if (rt->rt_flags & RTCF_LOCAL &&
1781 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1782 struct vxlan_dev *dst_vxlan;
1783
1784 ip_rt_put(rt);
1785 dst_vxlan = vxlan_find_vni(dev_net(dev), vni,
1786 dst->sa.sa_family, dst_port);
1787 if (!dst_vxlan)
1788 goto tx_error;
1789 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1790 return;
1791 }
1792
1793 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1794 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
1795
1796 err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
1797 fl4.saddr, dst->sin.sin_addr.s_addr,
1798 tos, ttl, df, src_port, dst_port,
1799 htonl(vni << 8));
1800
1801 if (err < 0)
1802 goto rt_tx_error;
1803 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1804 #if IS_ENABLED(CONFIG_IPV6)
1805 } else {
1806 struct sock *sk = vxlan->vn_sock->sock->sk;
1807 struct dst_entry *ndst;
1808 struct flowi6 fl6;
1809 u32 flags;
1810
1811 memset(&fl6, 0, sizeof(fl6));
1812 fl6.flowi6_oif = rdst->remote_ifindex;
1813 fl6.daddr = dst->sin6.sin6_addr;
1814 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
1815 fl6.flowi6_proto = IPPROTO_UDP;
1816
1817 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1818 netdev_dbg(dev, "no route to %pI6\n",
1819 &dst->sin6.sin6_addr);
1820 dev->stats.tx_carrier_errors++;
1821 goto tx_error;
1822 }
1823
1824 if (ndst->dev == dev) {
1825 netdev_dbg(dev, "circular route to %pI6\n",
1826 &dst->sin6.sin6_addr);
1827 dst_release(ndst);
1828 dev->stats.collisions++;
1829 goto tx_error;
1830 }
1831
1832 /* Bypass encapsulation if the destination is local */
1833 flags = ((struct rt6_info *)ndst)->rt6i_flags;
1834 if (flags & RTF_LOCAL &&
1835 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1836 struct vxlan_dev *dst_vxlan;
1837
1838 dst_release(ndst);
1839 dst_vxlan = vxlan_find_vni(dev_net(dev), vni,
1840 dst->sa.sa_family, dst_port);
1841 if (!dst_vxlan)
1842 goto tx_error;
1843 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1844 return;
1845 }
1846
1847 ttl = ttl ? : ip6_dst_hoplimit(ndst);
1848
1849 err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
1850 dev, &fl6.saddr, &fl6.daddr, 0, ttl,
1851 src_port, dst_port, htonl(vni << 8));
1852 #endif
1853 }
1854
1855 return;
1856
1857 drop:
1858 dev->stats.tx_dropped++;
1859 goto tx_free;
1860
1861 rt_tx_error:
1862 ip_rt_put(rt);
1863 tx_error:
1864 dev->stats.tx_errors++;
1865 tx_free:
1866 dev_kfree_skb(skb);
1867 }
1868
1869 /* Transmit local packets over Vxlan
1870 *
1871 * Outer IP header inherits ECN and DF from inner header.
1872 * Outer UDP destination is the VXLAN assigned port.
1873 * source port is based on hash of flow
1874 */
1875 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1876 {
1877 struct vxlan_dev *vxlan = netdev_priv(dev);
1878 struct ethhdr *eth;
1879 bool did_rsc = false;
1880 struct vxlan_rdst *rdst;
1881 struct vxlan_fdb *f;
1882
1883 skb_reset_mac_header(skb);
1884 eth = eth_hdr(skb);
1885
1886 if ((vxlan->flags & VXLAN_F_PROXY)) {
1887 if (ntohs(eth->h_proto) == ETH_P_ARP)
1888 return arp_reduce(dev, skb);
1889 #if IS_ENABLED(CONFIG_IPV6)
1890 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
1891 pskb_may_pull(skb, sizeof(struct ipv6hdr)
1892 + sizeof(struct nd_msg)) &&
1893 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
1894 struct nd_msg *msg;
1895
1896 msg = (struct nd_msg *)skb_transport_header(skb);
1897 if (msg->icmph.icmp6_code == 0 &&
1898 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
1899 return neigh_reduce(dev, skb);
1900 }
1901 eth = eth_hdr(skb);
1902 #endif
1903 }
1904
1905 f = vxlan_find_mac(vxlan, eth->h_dest);
1906 did_rsc = false;
1907
1908 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
1909 (ntohs(eth->h_proto) == ETH_P_IP ||
1910 ntohs(eth->h_proto) == ETH_P_IPV6)) {
1911 did_rsc = route_shortcircuit(dev, skb);
1912 if (did_rsc)
1913 f = vxlan_find_mac(vxlan, eth->h_dest);
1914 }
1915
1916 if (f == NULL) {
1917 f = vxlan_find_mac(vxlan, all_zeros_mac);
1918 if (f == NULL) {
1919 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1920 !is_multicast_ether_addr(eth->h_dest))
1921 vxlan_fdb_miss(vxlan, eth->h_dest);
1922
1923 dev->stats.tx_dropped++;
1924 dev_kfree_skb(skb);
1925 return NETDEV_TX_OK;
1926 }
1927 }
1928
1929 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1930 struct sk_buff *skb1;
1931
1932 skb1 = skb_clone(skb, GFP_ATOMIC);
1933 if (skb1)
1934 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
1935 }
1936
1937 dev_kfree_skb(skb);
1938 return NETDEV_TX_OK;
1939 }
1940
1941 /* Walk the forwarding table and purge stale entries */
1942 static void vxlan_cleanup(unsigned long arg)
1943 {
1944 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1945 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1946 unsigned int h;
1947
1948 if (!netif_running(vxlan->dev))
1949 return;
1950
1951 spin_lock_bh(&vxlan->hash_lock);
1952 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1953 struct hlist_node *p, *n;
1954 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1955 struct vxlan_fdb *f
1956 = container_of(p, struct vxlan_fdb, hlist);
1957 unsigned long timeout;
1958
1959 if (f->state & NUD_PERMANENT)
1960 continue;
1961
1962 timeout = f->used + vxlan->age_interval * HZ;
1963 if (time_before_eq(timeout, jiffies)) {
1964 netdev_dbg(vxlan->dev,
1965 "garbage collect %pM\n",
1966 f->eth_addr);
1967 f->state = NUD_STALE;
1968 vxlan_fdb_destroy(vxlan, f);
1969 } else if (time_before(timeout, next_timer))
1970 next_timer = timeout;
1971 }
1972 }
1973 spin_unlock_bh(&vxlan->hash_lock);
1974
1975 mod_timer(&vxlan->age_timer, next_timer);
1976 }
1977
1978 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
1979 {
1980 __u32 vni = vxlan->default_dst.remote_vni;
1981
1982 vxlan->vn_sock = vs;
1983 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
1984 }
1985
1986 /* Setup stats when device is created */
1987 static int vxlan_init(struct net_device *dev)
1988 {
1989 struct vxlan_dev *vxlan = netdev_priv(dev);
1990 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1991 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
1992 struct vxlan_sock *vs;
1993
1994 dev->tstats = alloc_percpu(struct pcpu_tstats);
1995 if (!dev->tstats)
1996 return -ENOMEM;
1997
1998 spin_lock(&vn->sock_lock);
1999 vs = vxlan_find_sock(dev_net(dev), ipv6 ? AF_INET6 : AF_INET,
2000 vxlan->dst_port);
2001 if (vs && atomic_add_unless(&vs->refcnt, 1, 0)) {
2002 /* If we have a socket with same port already, reuse it */
2003 vxlan_vs_add_dev(vs, vxlan);
2004 } else {
2005 /* otherwise make new socket outside of RTNL */
2006 dev_hold(dev);
2007 queue_work(vxlan_wq, &vxlan->sock_work);
2008 }
2009 spin_unlock(&vn->sock_lock);
2010
2011 return 0;
2012 }
2013
2014 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
2015 {
2016 struct vxlan_fdb *f;
2017
2018 spin_lock_bh(&vxlan->hash_lock);
2019 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2020 if (f)
2021 vxlan_fdb_destroy(vxlan, f);
2022 spin_unlock_bh(&vxlan->hash_lock);
2023 }
2024
2025 static void vxlan_uninit(struct net_device *dev)
2026 {
2027 struct vxlan_dev *vxlan = netdev_priv(dev);
2028 struct vxlan_sock *vs = vxlan->vn_sock;
2029
2030 vxlan_fdb_delete_default(vxlan);
2031
2032 if (vs)
2033 vxlan_sock_release(vs);
2034 free_percpu(dev->tstats);
2035 }
2036
2037 /* Start ageing timer and join group when device is brought up */
2038 static int vxlan_open(struct net_device *dev)
2039 {
2040 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
2041 struct vxlan_dev *vxlan = netdev_priv(dev);
2042 struct vxlan_sock *vs = vxlan->vn_sock;
2043
2044 /* socket hasn't been created */
2045 if (!vs)
2046 return -ENOTCONN;
2047
2048 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2049 vxlan_group_used(vn, &vxlan->default_dst.remote_ip)) {
2050 vxlan_sock_hold(vs);
2051 dev_hold(dev);
2052 queue_work(vxlan_wq, &vxlan->igmp_join);
2053 }
2054
2055 if (vxlan->age_interval)
2056 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2057
2058 return 0;
2059 }
2060
2061 /* Purge the forwarding table */
2062 static void vxlan_flush(struct vxlan_dev *vxlan)
2063 {
2064 unsigned int h;
2065
2066 spin_lock_bh(&vxlan->hash_lock);
2067 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2068 struct hlist_node *p, *n;
2069 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2070 struct vxlan_fdb *f
2071 = container_of(p, struct vxlan_fdb, hlist);
2072 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2073 if (!is_zero_ether_addr(f->eth_addr))
2074 vxlan_fdb_destroy(vxlan, f);
2075 }
2076 }
2077 spin_unlock_bh(&vxlan->hash_lock);
2078 }
2079
2080 /* Cleanup timer and forwarding table on shutdown */
2081 static int vxlan_stop(struct net_device *dev)
2082 {
2083 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
2084 struct vxlan_dev *vxlan = netdev_priv(dev);
2085 struct vxlan_sock *vs = vxlan->vn_sock;
2086
2087 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2088 ! vxlan_group_used(vn, &vxlan->default_dst.remote_ip)) {
2089 vxlan_sock_hold(vs);
2090 dev_hold(dev);
2091 queue_work(vxlan_wq, &vxlan->igmp_leave);
2092 }
2093
2094 del_timer_sync(&vxlan->age_timer);
2095
2096 vxlan_flush(vxlan);
2097
2098 return 0;
2099 }
2100
2101 /* Stub, nothing needs to be done. */
2102 static void vxlan_set_multicast_list(struct net_device *dev)
2103 {
2104 }
2105
2106 static const struct net_device_ops vxlan_netdev_ops = {
2107 .ndo_init = vxlan_init,
2108 .ndo_uninit = vxlan_uninit,
2109 .ndo_open = vxlan_open,
2110 .ndo_stop = vxlan_stop,
2111 .ndo_start_xmit = vxlan_xmit,
2112 .ndo_get_stats64 = ip_tunnel_get_stats64,
2113 .ndo_set_rx_mode = vxlan_set_multicast_list,
2114 .ndo_change_mtu = eth_change_mtu,
2115 .ndo_validate_addr = eth_validate_addr,
2116 .ndo_set_mac_address = eth_mac_addr,
2117 .ndo_fdb_add = vxlan_fdb_add,
2118 .ndo_fdb_del = vxlan_fdb_delete,
2119 .ndo_fdb_dump = vxlan_fdb_dump,
2120 };
2121
2122 /* Info for udev, that this is a virtual tunnel endpoint */
2123 static struct device_type vxlan_type = {
2124 .name = "vxlan",
2125 };
2126
2127 /* Calls the ndo_add_vxlan_port of the caller in order to
2128 * supply the listening VXLAN udp ports. Callers are expected
2129 * to implement the ndo_add_vxlan_port.
2130 */
2131 void vxlan_get_rx_port(struct net_device *dev)
2132 {
2133 struct vxlan_sock *vs;
2134 struct net *net = dev_net(dev);
2135 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2136 sa_family_t sa_family;
2137 __be16 port;
2138 unsigned int i;
2139
2140 spin_lock(&vn->sock_lock);
2141 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2142 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2143 port = inet_sk(vs->sock->sk)->inet_sport;
2144 sa_family = vs->sock->sk->sk_family;
2145 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2146 port);
2147 }
2148 }
2149 spin_unlock(&vn->sock_lock);
2150 }
2151 EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2152
2153 /* Initialize the device structure. */
2154 static void vxlan_setup(struct net_device *dev)
2155 {
2156 struct vxlan_dev *vxlan = netdev_priv(dev);
2157 unsigned int h;
2158 int low, high;
2159
2160 eth_hw_addr_random(dev);
2161 ether_setup(dev);
2162 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2163 dev->needed_headroom = ETH_HLEN + VXLAN6_HEADROOM;
2164 else
2165 dev->needed_headroom = ETH_HLEN + VXLAN_HEADROOM;
2166
2167 dev->netdev_ops = &vxlan_netdev_ops;
2168 dev->destructor = free_netdev;
2169 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2170
2171 dev->tx_queue_len = 0;
2172 dev->features |= NETIF_F_LLTX;
2173 dev->features |= NETIF_F_NETNS_LOCAL;
2174 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
2175 dev->features |= NETIF_F_RXCSUM;
2176 dev->features |= NETIF_F_GSO_SOFTWARE;
2177
2178 dev->vlan_features = dev->features;
2179 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2180 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2181 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2182 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2183 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
2184 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2185
2186 INIT_LIST_HEAD(&vxlan->next);
2187 spin_lock_init(&vxlan->hash_lock);
2188 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2189 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
2190 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
2191
2192 init_timer_deferrable(&vxlan->age_timer);
2193 vxlan->age_timer.function = vxlan_cleanup;
2194 vxlan->age_timer.data = (unsigned long) vxlan;
2195
2196 inet_get_local_port_range(&low, &high);
2197 vxlan->port_min = low;
2198 vxlan->port_max = high;
2199 vxlan->dst_port = htons(vxlan_port);
2200
2201 vxlan->dev = dev;
2202
2203 for (h = 0; h < FDB_HASH_SIZE; ++h)
2204 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2205 }
2206
2207 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2208 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
2209 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2210 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
2211 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2212 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2213 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
2214 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2215 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2216 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2217 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2218 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
2219 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
2220 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2221 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2222 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2223 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
2224 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
2225 };
2226
2227 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2228 {
2229 if (tb[IFLA_ADDRESS]) {
2230 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2231 pr_debug("invalid link address (not ethernet)\n");
2232 return -EINVAL;
2233 }
2234
2235 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2236 pr_debug("invalid all zero ethernet address\n");
2237 return -EADDRNOTAVAIL;
2238 }
2239 }
2240
2241 if (!data)
2242 return -EINVAL;
2243
2244 if (data[IFLA_VXLAN_ID]) {
2245 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2246 if (id >= VXLAN_VID_MASK)
2247 return -ERANGE;
2248 }
2249
2250 if (data[IFLA_VXLAN_PORT_RANGE]) {
2251 const struct ifla_vxlan_port_range *p
2252 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2253
2254 if (ntohs(p->high) < ntohs(p->low)) {
2255 pr_debug("port range %u .. %u not valid\n",
2256 ntohs(p->low), ntohs(p->high));
2257 return -EINVAL;
2258 }
2259 }
2260
2261 return 0;
2262 }
2263
2264 static void vxlan_get_drvinfo(struct net_device *netdev,
2265 struct ethtool_drvinfo *drvinfo)
2266 {
2267 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2268 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2269 }
2270
2271 static const struct ethtool_ops vxlan_ethtool_ops = {
2272 .get_drvinfo = vxlan_get_drvinfo,
2273 .get_link = ethtool_op_get_link,
2274 };
2275
2276 static void vxlan_del_work(struct work_struct *work)
2277 {
2278 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2279
2280 sk_release_kernel(vs->sock->sk);
2281 kfree_rcu(vs, rcu);
2282 }
2283
2284 #if IS_ENABLED(CONFIG_IPV6)
2285 /* Create UDP socket for encapsulation receive. AF_INET6 socket
2286 * could be used for both IPv4 and IPv6 communications, but
2287 * users may set bindv6only=1.
2288 */
2289 static int create_v6_sock(struct net *net, __be16 port, struct socket **psock)
2290 {
2291 struct sock *sk;
2292 struct socket *sock;
2293 struct sockaddr_in6 vxlan_addr = {
2294 .sin6_family = AF_INET6,
2295 .sin6_port = port,
2296 };
2297 int rc, val = 1;
2298
2299 rc = sock_create_kern(AF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
2300 if (rc < 0) {
2301 pr_debug("UDPv6 socket create failed\n");
2302 return rc;
2303 }
2304
2305 /* Put in proper namespace */
2306 sk = sock->sk;
2307 sk_change_net(sk, net);
2308
2309 kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
2310 (char *)&val, sizeof(val));
2311 rc = kernel_bind(sock, (struct sockaddr *)&vxlan_addr,
2312 sizeof(struct sockaddr_in6));
2313 if (rc < 0) {
2314 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2315 &vxlan_addr.sin6_addr, ntohs(vxlan_addr.sin6_port), rc);
2316 sk_release_kernel(sk);
2317 return rc;
2318 }
2319 /* At this point, IPv6 module should have been loaded in
2320 * sock_create_kern().
2321 */
2322 BUG_ON(!ipv6_stub);
2323
2324 *psock = sock;
2325 /* Disable multicast loopback */
2326 inet_sk(sk)->mc_loop = 0;
2327 return 0;
2328 }
2329
2330 #else
2331
2332 static int create_v6_sock(struct net *net, __be16 port, struct socket **psock)
2333 {
2334 return -EPFNOSUPPORT;
2335 }
2336 #endif
2337
2338 static int create_v4_sock(struct net *net, __be16 port, struct socket **psock)
2339 {
2340 struct sock *sk;
2341 struct socket *sock;
2342 struct sockaddr_in vxlan_addr = {
2343 .sin_family = AF_INET,
2344 .sin_addr.s_addr = htonl(INADDR_ANY),
2345 .sin_port = port,
2346 };
2347 int rc;
2348
2349 /* Create UDP socket for encapsulation receive. */
2350 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
2351 if (rc < 0) {
2352 pr_debug("UDP socket create failed\n");
2353 return rc;
2354 }
2355
2356 /* Put in proper namespace */
2357 sk = sock->sk;
2358 sk_change_net(sk, net);
2359
2360 rc = kernel_bind(sock, (struct sockaddr *) &vxlan_addr,
2361 sizeof(vxlan_addr));
2362 if (rc < 0) {
2363 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2364 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
2365 sk_release_kernel(sk);
2366 return rc;
2367 }
2368
2369 *psock = sock;
2370 /* Disable multicast loopback */
2371 inet_sk(sk)->mc_loop = 0;
2372 return 0;
2373 }
2374
2375 /* Create new listen socket if needed */
2376 static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
2377 vxlan_rcv_t *rcv, void *data, bool ipv6)
2378 {
2379 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2380 struct vxlan_sock *vs;
2381 struct socket *sock;
2382 struct sock *sk;
2383 int rc = 0;
2384 unsigned int h;
2385
2386 vs = kmalloc(sizeof(*vs), GFP_KERNEL);
2387 if (!vs)
2388 return ERR_PTR(-ENOMEM);
2389
2390 for (h = 0; h < VNI_HASH_SIZE; ++h)
2391 INIT_HLIST_HEAD(&vs->vni_list[h]);
2392
2393 INIT_WORK(&vs->del_work, vxlan_del_work);
2394
2395 if (ipv6)
2396 rc = create_v6_sock(net, port, &sock);
2397 else
2398 rc = create_v4_sock(net, port, &sock);
2399 if (rc < 0) {
2400 kfree(vs);
2401 return ERR_PTR(rc);
2402 }
2403
2404 vs->sock = sock;
2405 sk = sock->sk;
2406 atomic_set(&vs->refcnt, 1);
2407 vs->rcv = rcv;
2408 vs->data = data;
2409 rcu_assign_sk_user_data(vs->sock->sk, vs);
2410
2411 spin_lock(&vn->sock_lock);
2412 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2413 vxlan_notify_add_rx_port(sk);
2414 spin_unlock(&vn->sock_lock);
2415
2416 /* Mark socket as an encapsulation socket. */
2417 udp_sk(sk)->encap_type = 1;
2418 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
2419 #if IS_ENABLED(CONFIG_IPV6)
2420 if (ipv6)
2421 ipv6_stub->udpv6_encap_enable();
2422 else
2423 #endif
2424 udp_encap_enable();
2425
2426 return vs;
2427 }
2428
2429 struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2430 vxlan_rcv_t *rcv, void *data,
2431 bool no_share, bool ipv6)
2432 {
2433 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2434 struct vxlan_sock *vs;
2435
2436 vs = vxlan_socket_create(net, port, rcv, data, ipv6);
2437 if (!IS_ERR(vs))
2438 return vs;
2439
2440 if (no_share) /* Return error if sharing is not allowed. */
2441 return vs;
2442
2443 spin_lock(&vn->sock_lock);
2444 vs = vxlan_find_sock(net, ipv6 ? AF_INET6 : AF_INET, port);
2445 if (vs && ((vs->rcv != rcv) ||
2446 !atomic_add_unless(&vs->refcnt, 1, 0)))
2447 vs = ERR_PTR(-EBUSY);
2448 spin_unlock(&vn->sock_lock);
2449
2450 if (!vs)
2451 vs = ERR_PTR(-EINVAL);
2452
2453 return vs;
2454 }
2455 EXPORT_SYMBOL_GPL(vxlan_sock_add);
2456
2457 /* Scheduled at device creation to bind to a socket */
2458 static void vxlan_sock_work(struct work_struct *work)
2459 {
2460 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
2461 struct net *net = dev_net(vxlan->dev);
2462 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2463 __be16 port = vxlan->dst_port;
2464 struct vxlan_sock *nvs;
2465
2466 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags & VXLAN_F_IPV6);
2467 spin_lock(&vn->sock_lock);
2468 if (!IS_ERR(nvs))
2469 vxlan_vs_add_dev(nvs, vxlan);
2470 spin_unlock(&vn->sock_lock);
2471
2472 dev_put(vxlan->dev);
2473 }
2474
2475 static int vxlan_newlink(struct net *net, struct net_device *dev,
2476 struct nlattr *tb[], struct nlattr *data[])
2477 {
2478 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2479 struct vxlan_dev *vxlan = netdev_priv(dev);
2480 struct vxlan_rdst *dst = &vxlan->default_dst;
2481 __u32 vni;
2482 int err;
2483 bool use_ipv6 = false;
2484
2485 if (!data[IFLA_VXLAN_ID])
2486 return -EINVAL;
2487
2488 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
2489 dst->remote_vni = vni;
2490
2491 /* Unless IPv6 is explicitly requested, assume IPv4 */
2492 dst->remote_ip.sa.sa_family = AF_INET;
2493 if (data[IFLA_VXLAN_GROUP]) {
2494 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
2495 } else if (data[IFLA_VXLAN_GROUP6]) {
2496 if (!IS_ENABLED(CONFIG_IPV6))
2497 return -EPFNOSUPPORT;
2498
2499 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2500 sizeof(struct in6_addr));
2501 dst->remote_ip.sa.sa_family = AF_INET6;
2502 use_ipv6 = true;
2503 }
2504
2505 if (data[IFLA_VXLAN_LOCAL]) {
2506 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2507 vxlan->saddr.sa.sa_family = AF_INET;
2508 } else if (data[IFLA_VXLAN_LOCAL6]) {
2509 if (!IS_ENABLED(CONFIG_IPV6))
2510 return -EPFNOSUPPORT;
2511
2512 /* TODO: respect scope id */
2513 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2514 sizeof(struct in6_addr));
2515 vxlan->saddr.sa.sa_family = AF_INET6;
2516 use_ipv6 = true;
2517 }
2518
2519 if (data[IFLA_VXLAN_LINK] &&
2520 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
2521 struct net_device *lowerdev
2522 = __dev_get_by_index(net, dst->remote_ifindex);
2523
2524 if (!lowerdev) {
2525 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
2526 return -ENODEV;
2527 }
2528
2529 #if IS_ENABLED(CONFIG_IPV6)
2530 if (use_ipv6) {
2531 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2532 if (idev && idev->cnf.disable_ipv6) {
2533 pr_info("IPv6 is disabled via sysctl\n");
2534 return -EPERM;
2535 }
2536 vxlan->flags |= VXLAN_F_IPV6;
2537 }
2538 #endif
2539
2540 if (!tb[IFLA_MTU])
2541 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2542
2543 dev->needed_headroom = lowerdev->hard_header_len +
2544 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2545 }
2546
2547 if (data[IFLA_VXLAN_TOS])
2548 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2549
2550 if (data[IFLA_VXLAN_TTL])
2551 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2552
2553 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
2554 vxlan->flags |= VXLAN_F_LEARN;
2555
2556 if (data[IFLA_VXLAN_AGEING])
2557 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2558 else
2559 vxlan->age_interval = FDB_AGE_DEFAULT;
2560
2561 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2562 vxlan->flags |= VXLAN_F_PROXY;
2563
2564 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2565 vxlan->flags |= VXLAN_F_RSC;
2566
2567 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2568 vxlan->flags |= VXLAN_F_L2MISS;
2569
2570 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2571 vxlan->flags |= VXLAN_F_L3MISS;
2572
2573 if (data[IFLA_VXLAN_LIMIT])
2574 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2575
2576 if (data[IFLA_VXLAN_PORT_RANGE]) {
2577 const struct ifla_vxlan_port_range *p
2578 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2579 vxlan->port_min = ntohs(p->low);
2580 vxlan->port_max = ntohs(p->high);
2581 }
2582
2583 if (data[IFLA_VXLAN_PORT])
2584 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2585
2586 if (vxlan_find_vni(net, vni, use_ipv6 ? AF_INET6 : AF_INET,
2587 vxlan->dst_port)) {
2588 pr_info("duplicate VNI %u\n", vni);
2589 return -EEXIST;
2590 }
2591
2592 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
2593
2594 /* create an fdb entry for a valid default destination */
2595 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2596 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2597 &vxlan->default_dst.remote_ip,
2598 NUD_REACHABLE|NUD_PERMANENT,
2599 NLM_F_EXCL|NLM_F_CREATE,
2600 vxlan->dst_port,
2601 vxlan->default_dst.remote_vni,
2602 vxlan->default_dst.remote_ifindex,
2603 NTF_SELF);
2604 if (err)
2605 return err;
2606 }
2607
2608 err = register_netdevice(dev);
2609 if (err) {
2610 vxlan_fdb_delete_default(vxlan);
2611 return err;
2612 }
2613
2614 list_add(&vxlan->next, &vn->vxlan_list);
2615
2616 return 0;
2617 }
2618
2619 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2620 {
2621 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
2622 struct vxlan_dev *vxlan = netdev_priv(dev);
2623
2624 spin_lock(&vn->sock_lock);
2625 if (!hlist_unhashed(&vxlan->hlist))
2626 hlist_del_rcu(&vxlan->hlist);
2627 spin_unlock(&vn->sock_lock);
2628
2629 list_del(&vxlan->next);
2630 unregister_netdevice_queue(dev, head);
2631 }
2632
2633 static size_t vxlan_get_size(const struct net_device *dev)
2634 {
2635
2636 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
2637 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
2638 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
2639 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
2640 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2641 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2642 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
2643 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2644 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2645 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2646 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
2647 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2648 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
2649 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
2650 nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */
2651 0;
2652 }
2653
2654 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2655 {
2656 const struct vxlan_dev *vxlan = netdev_priv(dev);
2657 const struct vxlan_rdst *dst = &vxlan->default_dst;
2658 struct ifla_vxlan_port_range ports = {
2659 .low = htons(vxlan->port_min),
2660 .high = htons(vxlan->port_max),
2661 };
2662
2663 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
2664 goto nla_put_failure;
2665
2666 if (!vxlan_addr_any(&dst->remote_ip)) {
2667 if (dst->remote_ip.sa.sa_family == AF_INET) {
2668 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2669 dst->remote_ip.sin.sin_addr.s_addr))
2670 goto nla_put_failure;
2671 #if IS_ENABLED(CONFIG_IPV6)
2672 } else {
2673 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2674 &dst->remote_ip.sin6.sin6_addr))
2675 goto nla_put_failure;
2676 #endif
2677 }
2678 }
2679
2680 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
2681 goto nla_put_failure;
2682
2683 if (!vxlan_addr_any(&vxlan->saddr)) {
2684 if (vxlan->saddr.sa.sa_family == AF_INET) {
2685 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2686 vxlan->saddr.sin.sin_addr.s_addr))
2687 goto nla_put_failure;
2688 #if IS_ENABLED(CONFIG_IPV6)
2689 } else {
2690 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2691 &vxlan->saddr.sin6.sin6_addr))
2692 goto nla_put_failure;
2693 #endif
2694 }
2695 }
2696
2697 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2698 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
2699 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2700 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2701 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2702 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2703 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2704 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2705 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2706 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2707 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
2708 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
2709 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
2710 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port))
2711 goto nla_put_failure;
2712
2713 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2714 goto nla_put_failure;
2715
2716 return 0;
2717
2718 nla_put_failure:
2719 return -EMSGSIZE;
2720 }
2721
2722 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2723 .kind = "vxlan",
2724 .maxtype = IFLA_VXLAN_MAX,
2725 .policy = vxlan_policy,
2726 .priv_size = sizeof(struct vxlan_dev),
2727 .setup = vxlan_setup,
2728 .validate = vxlan_validate,
2729 .newlink = vxlan_newlink,
2730 .dellink = vxlan_dellink,
2731 .get_size = vxlan_get_size,
2732 .fill_info = vxlan_fill_info,
2733 };
2734
2735 static __net_init int vxlan_init_net(struct net *net)
2736 {
2737 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2738 unsigned int h;
2739
2740 INIT_LIST_HEAD(&vn->vxlan_list);
2741 spin_lock_init(&vn->sock_lock);
2742
2743 for (h = 0; h < PORT_HASH_SIZE; ++h)
2744 INIT_HLIST_HEAD(&vn->sock_list[h]);
2745
2746 return 0;
2747 }
2748
2749 static __net_exit void vxlan_exit_net(struct net *net)
2750 {
2751 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2752 struct vxlan_dev *vxlan;
2753 LIST_HEAD(list);
2754
2755 rtnl_lock();
2756 list_for_each_entry(vxlan, &vn->vxlan_list, next)
2757 unregister_netdevice_queue(vxlan->dev, &list);
2758 unregister_netdevice_many(&list);
2759 rtnl_unlock();
2760 }
2761
2762 static struct pernet_operations vxlan_net_ops = {
2763 .init = vxlan_init_net,
2764 .exit = vxlan_exit_net,
2765 .id = &vxlan_net_id,
2766 .size = sizeof(struct vxlan_net),
2767 };
2768
2769 static int __init vxlan_init_module(void)
2770 {
2771 int rc;
2772
2773 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2774 if (!vxlan_wq)
2775 return -ENOMEM;
2776
2777 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2778
2779 rc = register_pernet_device(&vxlan_net_ops);
2780 if (rc)
2781 goto out1;
2782
2783 rc = rtnl_link_register(&vxlan_link_ops);
2784 if (rc)
2785 goto out2;
2786
2787 return 0;
2788
2789 out2:
2790 unregister_pernet_device(&vxlan_net_ops);
2791 out1:
2792 destroy_workqueue(vxlan_wq);
2793 return rc;
2794 }
2795 late_initcall(vxlan_init_module);
2796
2797 static void __exit vxlan_cleanup_module(void)
2798 {
2799 rtnl_link_unregister(&vxlan_link_ops);
2800 destroy_workqueue(vxlan_wq);
2801 unregister_pernet_device(&vxlan_net_ops);
2802 rcu_barrier();
2803 }
2804 module_exit(vxlan_cleanup_module);
2805
2806 MODULE_LICENSE("GPL");
2807 MODULE_VERSION(VXLAN_VERSION);
2808 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
2809 MODULE_ALIAS_RTNL_LINK("vxlan");
Linux with added FPC-III support