projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
i40e: use kernel specific defines
[deliverable/linux.git] / drivers / net / vxlan.c
1 /*
2 * VXLAN: Virtual eXtensible Local Area Network
3 *
4 * Copyright (c) 2012-2013 Vyatta Inc.
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
9 */
10
11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13 #include <linux/kernel.h>
14 #include <linux/types.h>
15 #include <linux/module.h>
16 #include <linux/errno.h>
17 #include <linux/slab.h>
18 #include <linux/skbuff.h>
19 #include <linux/rculist.h>
20 #include <linux/netdevice.h>
21 #include <linux/in.h>
22 #include <linux/ip.h>
23 #include <linux/udp.h>
24 #include <linux/igmp.h>
25 #include <linux/etherdevice.h>
26 #include <linux/if_ether.h>
27 #include <linux/if_vlan.h>
28 #include <linux/hash.h>
29 #include <linux/ethtool.h>
30 #include <net/arp.h>
31 #include <net/ndisc.h>
32 #include <net/ip.h>
33 #include <net/ip_tunnels.h>
34 #include <net/icmp.h>
35 #include <net/udp.h>
36 #include <net/rtnetlink.h>
37 #include <net/route.h>
38 #include <net/dsfield.h>
39 #include <net/inet_ecn.h>
40 #include <net/net_namespace.h>
41 #include <net/netns/generic.h>
42 #include <net/vxlan.h>
43 #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 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
64
65 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
66
67 /* VXLAN protocol header */
68 struct vxlanhdr {
69 __be32 vx_flags;
70 __be32 vx_vni;
71 };
72
73 /* UDP port for VXLAN traffic.
74 * The IANA assigned port is 4789, but the Linux default is 8472
75 * for compatibility with early adopters.
76 */
77 static unsigned short vxlan_port __read_mostly = 8472;
78 module_param_named(udp_port, vxlan_port, ushort, 0444);
79 MODULE_PARM_DESC(udp_port, "Destination UDP port");
80
81 static bool log_ecn_error = true;
82 module_param(log_ecn_error, bool, 0644);
83 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
84
85 static int vxlan_net_id;
86
87 static const u8 all_zeros_mac[ETH_ALEN];
88
89 /* per-network namespace private data for this module */
90 struct vxlan_net {
91 struct list_head vxlan_list;
92 struct hlist_head sock_list[PORT_HASH_SIZE];
93 spinlock_t sock_lock;
94 };
95
96 union vxlan_addr {
97 struct sockaddr_in sin;
98 struct sockaddr_in6 sin6;
99 struct sockaddr sa;
100 };
101
102 struct vxlan_rdst {
103 union vxlan_addr remote_ip;
104 __be16 remote_port;
105 u32 remote_vni;
106 u32 remote_ifindex;
107 struct list_head list;
108 struct rcu_head rcu;
109 };
110
111 /* Forwarding table entry */
112 struct vxlan_fdb {
113 struct hlist_node hlist; /* linked list of entries */
114 struct rcu_head rcu;
115 unsigned long updated; /* jiffies */
116 unsigned long used;
117 struct list_head remotes;
118 u16 state; /* see ndm_state */
119 u8 flags; /* see ndm_flags */
120 u8 eth_addr[ETH_ALEN];
121 };
122
123 /* Pseudo network device */
124 struct vxlan_dev {
125 struct hlist_node hlist; /* vni hash table */
126 struct list_head next; /* vxlan's per namespace list */
127 struct vxlan_sock *vn_sock; /* listening socket */
128 struct net_device *dev;
129 struct vxlan_rdst default_dst; /* default destination */
130 union vxlan_addr saddr; /* source address */
131 __be16 dst_port;
132 __u16 port_min; /* source port range */
133 __u16 port_max;
134 __u8 tos; /* TOS override */
135 __u8 ttl;
136 u32 flags; /* VXLAN_F_* below */
137
138 struct work_struct sock_work;
139 struct work_struct igmp_join;
140 struct work_struct igmp_leave;
141
142 unsigned long age_interval;
143 struct timer_list age_timer;
144 spinlock_t hash_lock;
145 unsigned int addrcnt;
146 unsigned int addrmax;
147
148 struct hlist_head fdb_head[FDB_HASH_SIZE];
149 };
150
151 #define VXLAN_F_LEARN 0x01
152 #define VXLAN_F_PROXY 0x02
153 #define VXLAN_F_RSC 0x04
154 #define VXLAN_F_L2MISS 0x08
155 #define VXLAN_F_L3MISS 0x10
156 #define VXLAN_F_IPV6 0x20 /* internal flag */
157
158 /* salt for hash table */
159 static u32 vxlan_salt __read_mostly;
160 static struct workqueue_struct *vxlan_wq;
161
162 static void vxlan_sock_work(struct work_struct *work);
163
164 #if IS_ENABLED(CONFIG_IPV6)
165 static inline
166 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
167 {
168 if (a->sa.sa_family != b->sa.sa_family)
169 return false;
170 if (a->sa.sa_family == AF_INET6)
171 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
172 else
173 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
174 }
175
176 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
177 {
178 if (ipa->sa.sa_family == AF_INET6)
179 return ipv6_addr_any(&ipa->sin6.sin6_addr);
180 else
181 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
182 }
183
184 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
185 {
186 if (ipa->sa.sa_family == AF_INET6)
187 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
188 else
189 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
190 }
191
192 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
193 {
194 if (nla_len(nla) >= sizeof(struct in6_addr)) {
195 nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr));
196 ip->sa.sa_family = AF_INET6;
197 return 0;
198 } else if (nla_len(nla) >= sizeof(__be32)) {
199 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
200 ip->sa.sa_family = AF_INET;
201 return 0;
202 } else {
203 return -EAFNOSUPPORT;
204 }
205 }
206
207 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
208 const union vxlan_addr *ip)
209 {
210 if (ip->sa.sa_family == AF_INET6)
211 return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr);
212 else
213 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
214 }
215
216 #else /* !CONFIG_IPV6 */
217
218 static inline
219 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
220 {
221 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
222 }
223
224 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
225 {
226 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
227 }
228
229 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
230 {
231 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
232 }
233
234 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
235 {
236 if (nla_len(nla) >= sizeof(struct in6_addr)) {
237 return -EAFNOSUPPORT;
238 } else if (nla_len(nla) >= sizeof(__be32)) {
239 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
240 ip->sa.sa_family = AF_INET;
241 return 0;
242 } else {
243 return -EAFNOSUPPORT;
244 }
245 }
246
247 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
248 const union vxlan_addr *ip)
249 {
250 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
251 }
252 #endif
253
254 /* Virtual Network hash table head */
255 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
256 {
257 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
258 }
259
260 /* Socket hash table head */
261 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
262 {
263 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
264
265 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
266 }
267
268 /* First remote destination for a forwarding entry.
269 * Guaranteed to be non-NULL because remotes are never deleted.
270 */
271 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
272 {
273 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
274 }
275
276 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
277 {
278 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
279 }
280
281 /* Find VXLAN socket based on network namespace and UDP port */
282 static struct vxlan_sock *vxlan_find_sock(struct net *net, __be16 port)
283 {
284 struct vxlan_sock *vs;
285
286 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
287 if (inet_sk(vs->sock->sk)->inet_sport == port)
288 return vs;
289 }
290 return NULL;
291 }
292
293 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
294 {
295 struct vxlan_dev *vxlan;
296
297 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
298 if (vxlan->default_dst.remote_vni == id)
299 return vxlan;
300 }
301
302 return NULL;
303 }
304
305 /* Look up VNI in a per net namespace table */
306 static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port)
307 {
308 struct vxlan_sock *vs;
309
310 vs = vxlan_find_sock(net, port);
311 if (!vs)
312 return NULL;
313
314 return vxlan_vs_find_vni(vs, id);
315 }
316
317 /* Fill in neighbour message in skbuff. */
318 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
319 const struct vxlan_fdb *fdb,
320 u32 portid, u32 seq, int type, unsigned int flags,
321 const struct vxlan_rdst *rdst)
322 {
323 unsigned long now = jiffies;
324 struct nda_cacheinfo ci;
325 struct nlmsghdr *nlh;
326 struct ndmsg *ndm;
327 bool send_ip, send_eth;
328
329 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
330 if (nlh == NULL)
331 return -EMSGSIZE;
332
333 ndm = nlmsg_data(nlh);
334 memset(ndm, 0, sizeof(*ndm));
335
336 send_eth = send_ip = true;
337
338 if (type == RTM_GETNEIGH) {
339 ndm->ndm_family = AF_INET;
340 send_ip = !vxlan_addr_any(&rdst->remote_ip);
341 send_eth = !is_zero_ether_addr(fdb->eth_addr);
342 } else
343 ndm->ndm_family = AF_BRIDGE;
344 ndm->ndm_state = fdb->state;
345 ndm->ndm_ifindex = vxlan->dev->ifindex;
346 ndm->ndm_flags = fdb->flags;
347 ndm->ndm_type = NDA_DST;
348
349 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
350 goto nla_put_failure;
351
352 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
353 goto nla_put_failure;
354
355 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
356 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
357 goto nla_put_failure;
358 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
359 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
360 goto nla_put_failure;
361 if (rdst->remote_ifindex &&
362 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
363 goto nla_put_failure;
364
365 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
366 ci.ndm_confirmed = 0;
367 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
368 ci.ndm_refcnt = 0;
369
370 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
371 goto nla_put_failure;
372
373 return nlmsg_end(skb, nlh);
374
375 nla_put_failure:
376 nlmsg_cancel(skb, nlh);
377 return -EMSGSIZE;
378 }
379
380 static inline size_t vxlan_nlmsg_size(void)
381 {
382 return NLMSG_ALIGN(sizeof(struct ndmsg))
383 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
384 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
385 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
386 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
387 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
388 + nla_total_size(sizeof(struct nda_cacheinfo));
389 }
390
391 static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
392 struct vxlan_fdb *fdb, int type)
393 {
394 struct net *net = dev_net(vxlan->dev);
395 struct sk_buff *skb;
396 int err = -ENOBUFS;
397
398 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
399 if (skb == NULL)
400 goto errout;
401
402 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0,
403 first_remote_rtnl(fdb));
404 if (err < 0) {
405 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
406 WARN_ON(err == -EMSGSIZE);
407 kfree_skb(skb);
408 goto errout;
409 }
410
411 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
412 return;
413 errout:
414 if (err < 0)
415 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
416 }
417
418 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
419 {
420 struct vxlan_dev *vxlan = netdev_priv(dev);
421 struct vxlan_fdb f = {
422 .state = NUD_STALE,
423 };
424 struct vxlan_rdst remote = {
425 .remote_ip = *ipa, /* goes to NDA_DST */
426 .remote_vni = VXLAN_N_VID,
427 };
428
429 INIT_LIST_HEAD(&f.remotes);
430 list_add_rcu(&remote.list, &f.remotes);
431
432 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
433 }
434
435 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
436 {
437 struct vxlan_fdb f = {
438 .state = NUD_STALE,
439 };
440
441 INIT_LIST_HEAD(&f.remotes);
442 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
443
444 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
445 }
446
447 /* Hash Ethernet address */
448 static u32 eth_hash(const unsigned char *addr)
449 {
450 u64 value = get_unaligned((u64 *)addr);
451
452 /* only want 6 bytes */
453 #ifdef __BIG_ENDIAN
454 value >>= 16;
455 #else
456 value <<= 16;
457 #endif
458 return hash_64(value, FDB_HASH_BITS);
459 }
460
461 /* Hash chain to use given mac address */
462 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
463 const u8 *mac)
464 {
465 return &vxlan->fdb_head[eth_hash(mac)];
466 }
467
468 /* Look up Ethernet address in forwarding table */
469 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
470 const u8 *mac)
471
472 {
473 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
474 struct vxlan_fdb *f;
475
476 hlist_for_each_entry_rcu(f, head, hlist) {
477 if (ether_addr_equal(mac, f->eth_addr))
478 return f;
479 }
480
481 return NULL;
482 }
483
484 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
485 const u8 *mac)
486 {
487 struct vxlan_fdb *f;
488
489 f = __vxlan_find_mac(vxlan, mac);
490 if (f)
491 f->used = jiffies;
492
493 return f;
494 }
495
496 /* caller should hold vxlan->hash_lock */
497 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
498 union vxlan_addr *ip, __be16 port,
499 __u32 vni, __u32 ifindex)
500 {
501 struct vxlan_rdst *rd;
502
503 list_for_each_entry(rd, &f->remotes, list) {
504 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
505 rd->remote_port == port &&
506 rd->remote_vni == vni &&
507 rd->remote_ifindex == ifindex)
508 return rd;
509 }
510
511 return NULL;
512 }
513
514 /* Replace destination of unicast mac */
515 static int vxlan_fdb_replace(struct vxlan_fdb *f,
516 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
517 {
518 struct vxlan_rdst *rd;
519
520 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
521 if (rd)
522 return 0;
523
524 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
525 if (!rd)
526 return 0;
527 rd->remote_ip = *ip;
528 rd->remote_port = port;
529 rd->remote_vni = vni;
530 rd->remote_ifindex = ifindex;
531 return 1;
532 }
533
534 /* Add/update destinations for multicast */
535 static int vxlan_fdb_append(struct vxlan_fdb *f,
536 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
537 {
538 struct vxlan_rdst *rd;
539
540 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
541 if (rd)
542 return 0;
543
544 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
545 if (rd == NULL)
546 return -ENOBUFS;
547 rd->remote_ip = *ip;
548 rd->remote_port = port;
549 rd->remote_vni = vni;
550 rd->remote_ifindex = ifindex;
551
552 list_add_tail_rcu(&rd->list, &f->remotes);
553
554 return 1;
555 }
556
557 /* Notify netdevs that UDP port started listening */
558 static void vxlan_notify_add_rx_port(struct sock *sk)
559 {
560 struct net_device *dev;
561 struct net *net = sock_net(sk);
562 sa_family_t sa_family = sk->sk_family;
563 __be16 port = inet_sk(sk)->inet_sport;
564
565 rcu_read_lock();
566 for_each_netdev_rcu(net, dev) {
567 if (dev->netdev_ops->ndo_add_vxlan_port)
568 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
569 port);
570 }
571 rcu_read_unlock();
572 }
573
574 /* Notify netdevs that UDP port is no more listening */
575 static void vxlan_notify_del_rx_port(struct sock *sk)
576 {
577 struct net_device *dev;
578 struct net *net = sock_net(sk);
579 sa_family_t sa_family = sk->sk_family;
580 __be16 port = inet_sk(sk)->inet_sport;
581
582 rcu_read_lock();
583 for_each_netdev_rcu(net, dev) {
584 if (dev->netdev_ops->ndo_del_vxlan_port)
585 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
586 port);
587 }
588 rcu_read_unlock();
589 }
590
591 /* Add new entry to forwarding table -- assumes lock held */
592 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
593 const u8 *mac, union vxlan_addr *ip,
594 __u16 state, __u16 flags,
595 __be16 port, __u32 vni, __u32 ifindex,
596 __u8 ndm_flags)
597 {
598 struct vxlan_fdb *f;
599 int notify = 0;
600
601 f = __vxlan_find_mac(vxlan, mac);
602 if (f) {
603 if (flags & NLM_F_EXCL) {
604 netdev_dbg(vxlan->dev,
605 "lost race to create %pM\n", mac);
606 return -EEXIST;
607 }
608 if (f->state != state) {
609 f->state = state;
610 f->updated = jiffies;
611 notify = 1;
612 }
613 if (f->flags != ndm_flags) {
614 f->flags = ndm_flags;
615 f->updated = jiffies;
616 notify = 1;
617 }
618 if ((flags & NLM_F_REPLACE)) {
619 /* Only change unicasts */
620 if (!(is_multicast_ether_addr(f->eth_addr) ||
621 is_zero_ether_addr(f->eth_addr))) {
622 int rc = vxlan_fdb_replace(f, ip, port, vni,
623 ifindex);
624
625 if (rc < 0)
626 return rc;
627 notify |= rc;
628 } else
629 return -EOPNOTSUPP;
630 }
631 if ((flags & NLM_F_APPEND) &&
632 (is_multicast_ether_addr(f->eth_addr) ||
633 is_zero_ether_addr(f->eth_addr))) {
634 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
635
636 if (rc < 0)
637 return rc;
638 notify |= rc;
639 }
640 } else {
641 if (!(flags & NLM_F_CREATE))
642 return -ENOENT;
643
644 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
645 return -ENOSPC;
646
647 /* Disallow replace to add a multicast entry */
648 if ((flags & NLM_F_REPLACE) &&
649 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
650 return -EOPNOTSUPP;
651
652 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
653 f = kmalloc(sizeof(*f), GFP_ATOMIC);
654 if (!f)
655 return -ENOMEM;
656
657 notify = 1;
658 f->state = state;
659 f->flags = ndm_flags;
660 f->updated = f->used = jiffies;
661 INIT_LIST_HEAD(&f->remotes);
662 memcpy(f->eth_addr, mac, ETH_ALEN);
663
664 vxlan_fdb_append(f, ip, port, vni, ifindex);
665
666 ++vxlan->addrcnt;
667 hlist_add_head_rcu(&f->hlist,
668 vxlan_fdb_head(vxlan, mac));
669 }
670
671 if (notify)
672 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
673
674 return 0;
675 }
676
677 static void vxlan_fdb_free(struct rcu_head *head)
678 {
679 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
680 struct vxlan_rdst *rd, *nd;
681
682 list_for_each_entry_safe(rd, nd, &f->remotes, list)
683 kfree(rd);
684 kfree(f);
685 }
686
687 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
688 {
689 netdev_dbg(vxlan->dev,
690 "delete %pM\n", f->eth_addr);
691
692 --vxlan->addrcnt;
693 vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
694
695 hlist_del_rcu(&f->hlist);
696 call_rcu(&f->rcu, vxlan_fdb_free);
697 }
698
699 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
700 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
701 {
702 struct net *net = dev_net(vxlan->dev);
703 int err;
704
705 if (tb[NDA_DST]) {
706 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
707 if (err)
708 return err;
709 } else {
710 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
711 if (remote->sa.sa_family == AF_INET) {
712 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
713 ip->sa.sa_family = AF_INET;
714 #if IS_ENABLED(CONFIG_IPV6)
715 } else {
716 ip->sin6.sin6_addr = in6addr_any;
717 ip->sa.sa_family = AF_INET6;
718 #endif
719 }
720 }
721
722 if (tb[NDA_PORT]) {
723 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
724 return -EINVAL;
725 *port = nla_get_be16(tb[NDA_PORT]);
726 } else {
727 *port = vxlan->dst_port;
728 }
729
730 if (tb[NDA_VNI]) {
731 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
732 return -EINVAL;
733 *vni = nla_get_u32(tb[NDA_VNI]);
734 } else {
735 *vni = vxlan->default_dst.remote_vni;
736 }
737
738 if (tb[NDA_IFINDEX]) {
739 struct net_device *tdev;
740
741 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
742 return -EINVAL;
743 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
744 tdev = dev_get_by_index(net, *ifindex);
745 if (!tdev)
746 return -EADDRNOTAVAIL;
747 dev_put(tdev);
748 } else {
749 *ifindex = 0;
750 }
751
752 return 0;
753 }
754
755 /* Add static entry (via netlink) */
756 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
757 struct net_device *dev,
758 const unsigned char *addr, u16 flags)
759 {
760 struct vxlan_dev *vxlan = netdev_priv(dev);
761 /* struct net *net = dev_net(vxlan->dev); */
762 union vxlan_addr ip;
763 __be16 port;
764 u32 vni, ifindex;
765 int err;
766
767 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
768 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
769 ndm->ndm_state);
770 return -EINVAL;
771 }
772
773 if (tb[NDA_DST] == NULL)
774 return -EINVAL;
775
776 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
777 if (err)
778 return err;
779
780 spin_lock_bh(&vxlan->hash_lock);
781 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
782 port, vni, ifindex, ndm->ndm_flags);
783 spin_unlock_bh(&vxlan->hash_lock);
784
785 return err;
786 }
787
788 /* Delete entry (via netlink) */
789 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
790 struct net_device *dev,
791 const unsigned char *addr)
792 {
793 struct vxlan_dev *vxlan = netdev_priv(dev);
794 struct vxlan_fdb *f;
795 struct vxlan_rdst *rd = NULL;
796 union vxlan_addr ip;
797 __be16 port;
798 u32 vni, ifindex;
799 int err;
800
801 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
802 if (err)
803 return err;
804
805 err = -ENOENT;
806
807 spin_lock_bh(&vxlan->hash_lock);
808 f = vxlan_find_mac(vxlan, addr);
809 if (!f)
810 goto out;
811
812 if (!vxlan_addr_any(&ip)) {
813 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
814 if (!rd)
815 goto out;
816 }
817
818 err = 0;
819
820 /* remove a destination if it's not the only one on the list,
821 * otherwise destroy the fdb entry
822 */
823 if (rd && !list_is_singular(&f->remotes)) {
824 list_del_rcu(&rd->list);
825 kfree_rcu(rd, rcu);
826 goto out;
827 }
828
829 vxlan_fdb_destroy(vxlan, f);
830
831 out:
832 spin_unlock_bh(&vxlan->hash_lock);
833
834 return err;
835 }
836
837 /* Dump forwarding table */
838 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
839 struct net_device *dev, int idx)
840 {
841 struct vxlan_dev *vxlan = netdev_priv(dev);
842 unsigned int h;
843
844 for (h = 0; h < FDB_HASH_SIZE; ++h) {
845 struct vxlan_fdb *f;
846 int err;
847
848 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
849 struct vxlan_rdst *rd;
850
851 if (idx < cb->args[0])
852 goto skip;
853
854 list_for_each_entry_rcu(rd, &f->remotes, list) {
855 err = vxlan_fdb_info(skb, vxlan, f,
856 NETLINK_CB(cb->skb).portid,
857 cb->nlh->nlmsg_seq,
858 RTM_NEWNEIGH,
859 NLM_F_MULTI, rd);
860 if (err < 0)
861 goto out;
862 }
863 skip:
864 ++idx;
865 }
866 }
867 out:
868 return idx;
869 }
870
871 /* Watch incoming packets to learn mapping between Ethernet address
872 * and Tunnel endpoint.
873 * Return true if packet is bogus and should be droppped.
874 */
875 static bool vxlan_snoop(struct net_device *dev,
876 union vxlan_addr *src_ip, const u8 *src_mac)
877 {
878 struct vxlan_dev *vxlan = netdev_priv(dev);
879 struct vxlan_fdb *f;
880
881 f = vxlan_find_mac(vxlan, src_mac);
882 if (likely(f)) {
883 struct vxlan_rdst *rdst = first_remote_rcu(f);
884
885 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
886 return false;
887
888 /* Don't migrate static entries, drop packets */
889 if (f->state & NUD_NOARP)
890 return true;
891
892 if (net_ratelimit())
893 netdev_info(dev,
894 "%pM migrated from %pIS to %pIS\n",
895 src_mac, &rdst->remote_ip, &src_ip);
896
897 rdst->remote_ip = *src_ip;
898 f->updated = jiffies;
899 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
900 } else {
901 /* learned new entry */
902 spin_lock(&vxlan->hash_lock);
903
904 /* close off race between vxlan_flush and incoming packets */
905 if (netif_running(dev))
906 vxlan_fdb_create(vxlan, src_mac, src_ip,
907 NUD_REACHABLE,
908 NLM_F_EXCL|NLM_F_CREATE,
909 vxlan->dst_port,
910 vxlan->default_dst.remote_vni,
911 0, NTF_SELF);
912 spin_unlock(&vxlan->hash_lock);
913 }
914
915 return false;
916 }
917
918 /* See if multicast group is already in use by other ID */
919 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
920 {
921 struct vxlan_dev *vxlan;
922
923 /* The vxlan_sock is only used by dev, leaving group has
924 * no effect on other vxlan devices.
925 */
926 if (atomic_read(&dev->vn_sock->refcnt) == 1)
927 return false;
928
929 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
930 if (!netif_running(vxlan->dev) || vxlan == dev)
931 continue;
932
933 if (vxlan->vn_sock != dev->vn_sock)
934 continue;
935
936 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
937 &dev->default_dst.remote_ip))
938 continue;
939
940 if (vxlan->default_dst.remote_ifindex !=
941 dev->default_dst.remote_ifindex)
942 continue;
943
944 return true;
945 }
946
947 return false;
948 }
949
950 static void vxlan_sock_hold(struct vxlan_sock *vs)
951 {
952 atomic_inc(&vs->refcnt);
953 }
954
955 void vxlan_sock_release(struct vxlan_sock *vs)
956 {
957 struct sock *sk = vs->sock->sk;
958 struct net *net = sock_net(sk);
959 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
960
961 if (!atomic_dec_and_test(&vs->refcnt))
962 return;
963
964 spin_lock(&vn->sock_lock);
965 hlist_del_rcu(&vs->hlist);
966 rcu_assign_sk_user_data(vs->sock->sk, NULL);
967 vxlan_notify_del_rx_port(sk);
968 spin_unlock(&vn->sock_lock);
969
970 queue_work(vxlan_wq, &vs->del_work);
971 }
972 EXPORT_SYMBOL_GPL(vxlan_sock_release);
973
974 /* Callback to update multicast group membership when first VNI on
975 * multicast asddress is brought up
976 * Done as workqueue because ip_mc_join_group acquires RTNL.
977 */
978 static void vxlan_igmp_join(struct work_struct *work)
979 {
980 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
981 struct vxlan_sock *vs = vxlan->vn_sock;
982 struct sock *sk = vs->sock->sk;
983 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
984 int ifindex = vxlan->default_dst.remote_ifindex;
985
986 lock_sock(sk);
987 if (ip->sa.sa_family == AF_INET) {
988 struct ip_mreqn mreq = {
989 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
990 .imr_ifindex = ifindex,
991 };
992
993 ip_mc_join_group(sk, &mreq);
994 #if IS_ENABLED(CONFIG_IPV6)
995 } else {
996 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
997 &ip->sin6.sin6_addr);
998 #endif
999 }
1000 release_sock(sk);
1001
1002 vxlan_sock_release(vs);
1003 dev_put(vxlan->dev);
1004 }
1005
1006 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1007 static void vxlan_igmp_leave(struct work_struct *work)
1008 {
1009 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
1010 struct vxlan_sock *vs = vxlan->vn_sock;
1011 struct sock *sk = vs->sock->sk;
1012 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1013 int ifindex = vxlan->default_dst.remote_ifindex;
1014
1015 lock_sock(sk);
1016 if (ip->sa.sa_family == AF_INET) {
1017 struct ip_mreqn mreq = {
1018 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1019 .imr_ifindex = ifindex,
1020 };
1021
1022 ip_mc_leave_group(sk, &mreq);
1023 #if IS_ENABLED(CONFIG_IPV6)
1024 } else {
1025 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1026 &ip->sin6.sin6_addr);
1027 #endif
1028 }
1029
1030 release_sock(sk);
1031
1032 vxlan_sock_release(vs);
1033 dev_put(vxlan->dev);
1034 }
1035
1036 /* Callback from net/ipv4/udp.c to receive packets */
1037 static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1038 {
1039 struct vxlan_sock *vs;
1040 struct vxlanhdr *vxh;
1041 __be16 port;
1042
1043 /* Need Vxlan and inner Ethernet header to be present */
1044 if (!pskb_may_pull(skb, VXLAN_HLEN))
1045 goto error;
1046
1047 /* Return packets with reserved bits set */
1048 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1049 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
1050 (vxh->vx_vni & htonl(0xff))) {
1051 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1052 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1053 goto error;
1054 }
1055
1056 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1057 goto drop;
1058
1059 port = inet_sk(sk)->inet_sport;
1060
1061 vs = rcu_dereference_sk_user_data(sk);
1062 if (!vs)
1063 goto drop;
1064
1065 vs->rcv(vs, skb, vxh->vx_vni);
1066 return 0;
1067
1068 drop:
1069 /* Consume bad packet */
1070 kfree_skb(skb);
1071 return 0;
1072
1073 error:
1074 /* Return non vxlan pkt */
1075 return 1;
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_sw_netstats *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 the NIC driver gave us an encapsulated packet with
1124 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
1125 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
1126 * for us. Otherwise force the upper layers to verify it.
1127 */
1128 if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation ||
1129 !(vxlan->dev->features & NETIF_F_RXCSUM))
1130 skb->ip_summed = CHECKSUM_NONE;
1131
1132 skb->encapsulation = 0;
1133
1134 if (oip6)
1135 err = IP6_ECN_decapsulate(oip6, skb);
1136 if (oip)
1137 err = IP_ECN_decapsulate(oip, skb);
1138
1139 if (unlikely(err)) {
1140 if (log_ecn_error) {
1141 if (oip6)
1142 net_info_ratelimited("non-ECT from %pI6\n",
1143 &oip6->saddr);
1144 if (oip)
1145 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1146 &oip->saddr, oip->tos);
1147 }
1148 if (err > 1) {
1149 ++vxlan->dev->stats.rx_frame_errors;
1150 ++vxlan->dev->stats.rx_errors;
1151 goto drop;
1152 }
1153 }
1154
1155 stats = this_cpu_ptr(vxlan->dev->tstats);
1156 u64_stats_update_begin(&stats->syncp);
1157 stats->rx_packets++;
1158 stats->rx_bytes += skb->len;
1159 u64_stats_update_end(&stats->syncp);
1160
1161 netif_rx(skb);
1162
1163 return;
1164 drop:
1165 /* Consume bad packet */
1166 kfree_skb(skb);
1167 }
1168
1169 static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1170 {
1171 struct vxlan_dev *vxlan = netdev_priv(dev);
1172 struct arphdr *parp;
1173 u8 *arpptr, *sha;
1174 __be32 sip, tip;
1175 struct neighbour *n;
1176
1177 if (dev->flags & IFF_NOARP)
1178 goto out;
1179
1180 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1181 dev->stats.tx_dropped++;
1182 goto out;
1183 }
1184 parp = arp_hdr(skb);
1185
1186 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1187 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1188 parp->ar_pro != htons(ETH_P_IP) ||
1189 parp->ar_op != htons(ARPOP_REQUEST) ||
1190 parp->ar_hln != dev->addr_len ||
1191 parp->ar_pln != 4)
1192 goto out;
1193 arpptr = (u8 *)parp + sizeof(struct arphdr);
1194 sha = arpptr;
1195 arpptr += dev->addr_len; /* sha */
1196 memcpy(&sip, arpptr, sizeof(sip));
1197 arpptr += sizeof(sip);
1198 arpptr += dev->addr_len; /* tha */
1199 memcpy(&tip, arpptr, sizeof(tip));
1200
1201 if (ipv4_is_loopback(tip) ||
1202 ipv4_is_multicast(tip))
1203 goto out;
1204
1205 n = neigh_lookup(&arp_tbl, &tip, dev);
1206
1207 if (n) {
1208 struct vxlan_fdb *f;
1209 struct sk_buff *reply;
1210
1211 if (!(n->nud_state & NUD_CONNECTED)) {
1212 neigh_release(n);
1213 goto out;
1214 }
1215
1216 f = vxlan_find_mac(vxlan, n->ha);
1217 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1218 /* bridge-local neighbor */
1219 neigh_release(n);
1220 goto out;
1221 }
1222
1223 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1224 n->ha, sha);
1225
1226 neigh_release(n);
1227
1228 skb_reset_mac_header(reply);
1229 __skb_pull(reply, skb_network_offset(reply));
1230 reply->ip_summed = CHECKSUM_UNNECESSARY;
1231 reply->pkt_type = PACKET_HOST;
1232
1233 if (netif_rx_ni(reply) == NET_RX_DROP)
1234 dev->stats.rx_dropped++;
1235 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1236 union vxlan_addr ipa = {
1237 .sin.sin_addr.s_addr = tip,
1238 .sa.sa_family = AF_INET,
1239 };
1240
1241 vxlan_ip_miss(dev, &ipa);
1242 }
1243 out:
1244 consume_skb(skb);
1245 return NETDEV_TX_OK;
1246 }
1247
1248 #if IS_ENABLED(CONFIG_IPV6)
1249 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1250 {
1251 struct vxlan_dev *vxlan = netdev_priv(dev);
1252 struct neighbour *n;
1253 union vxlan_addr ipa;
1254 const struct ipv6hdr *iphdr;
1255 const struct in6_addr *saddr, *daddr;
1256 struct nd_msg *msg;
1257 struct inet6_dev *in6_dev = NULL;
1258
1259 in6_dev = __in6_dev_get(dev);
1260 if (!in6_dev)
1261 goto out;
1262
1263 if (!pskb_may_pull(skb, skb->len))
1264 goto out;
1265
1266 iphdr = ipv6_hdr(skb);
1267 saddr = &iphdr->saddr;
1268 daddr = &iphdr->daddr;
1269
1270 if (ipv6_addr_loopback(daddr) ||
1271 ipv6_addr_is_multicast(daddr))
1272 goto out;
1273
1274 msg = (struct nd_msg *)skb_transport_header(skb);
1275 if (msg->icmph.icmp6_code != 0 ||
1276 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1277 goto out;
1278
1279 n = neigh_lookup(ipv6_stub->nd_tbl, daddr, dev);
1280
1281 if (n) {
1282 struct vxlan_fdb *f;
1283
1284 if (!(n->nud_state & NUD_CONNECTED)) {
1285 neigh_release(n);
1286 goto out;
1287 }
1288
1289 f = vxlan_find_mac(vxlan, n->ha);
1290 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1291 /* bridge-local neighbor */
1292 neigh_release(n);
1293 goto out;
1294 }
1295
1296 ipv6_stub->ndisc_send_na(dev, n, saddr, &msg->target,
1297 !!in6_dev->cnf.forwarding,
1298 true, false, false);
1299 neigh_release(n);
1300 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1301 ipa.sin6.sin6_addr = *daddr;
1302 ipa.sa.sa_family = AF_INET6;
1303 vxlan_ip_miss(dev, &ipa);
1304 }
1305
1306 out:
1307 consume_skb(skb);
1308 return NETDEV_TX_OK;
1309 }
1310 #endif
1311
1312 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1313 {
1314 struct vxlan_dev *vxlan = netdev_priv(dev);
1315 struct neighbour *n;
1316
1317 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1318 return false;
1319
1320 n = NULL;
1321 switch (ntohs(eth_hdr(skb)->h_proto)) {
1322 case ETH_P_IP:
1323 {
1324 struct iphdr *pip;
1325
1326 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1327 return false;
1328 pip = ip_hdr(skb);
1329 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1330 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1331 union vxlan_addr ipa = {
1332 .sin.sin_addr.s_addr = pip->daddr,
1333 .sa.sa_family = AF_INET,
1334 };
1335
1336 vxlan_ip_miss(dev, &ipa);
1337 return false;
1338 }
1339
1340 break;
1341 }
1342 #if IS_ENABLED(CONFIG_IPV6)
1343 case ETH_P_IPV6:
1344 {
1345 struct ipv6hdr *pip6;
1346
1347 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1348 return false;
1349 pip6 = ipv6_hdr(skb);
1350 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1351 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1352 union vxlan_addr ipa = {
1353 .sin6.sin6_addr = pip6->daddr,
1354 .sa.sa_family = AF_INET6,
1355 };
1356
1357 vxlan_ip_miss(dev, &ipa);
1358 return false;
1359 }
1360
1361 break;
1362 }
1363 #endif
1364 default:
1365 return false;
1366 }
1367
1368 if (n) {
1369 bool diff;
1370
1371 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1372 if (diff) {
1373 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1374 dev->addr_len);
1375 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1376 }
1377 neigh_release(n);
1378 return diff;
1379 }
1380
1381 return false;
1382 }
1383
1384 /* Compute source port for outgoing packet
1385 * first choice to use L4 flow hash since it will spread
1386 * better and maybe available from hardware
1387 * secondary choice is to use jhash on the Ethernet header
1388 */
1389 __be16 vxlan_src_port(__u16 port_min, __u16 port_max, struct sk_buff *skb)
1390 {
1391 unsigned int range = (port_max - port_min) + 1;
1392 u32 hash;
1393
1394 hash = skb_get_hash(skb);
1395 if (!hash)
1396 hash = jhash(skb->data, 2 * ETH_ALEN,
1397 (__force u32) skb->protocol);
1398
1399 return htons((((u64) hash * range) >> 32) + port_min);
1400 }
1401 EXPORT_SYMBOL_GPL(vxlan_src_port);
1402
1403 static int handle_offloads(struct sk_buff *skb)
1404 {
1405 if (skb_is_gso(skb)) {
1406 int err = skb_unclone(skb, GFP_ATOMIC);
1407 if (unlikely(err))
1408 return err;
1409
1410 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
1411 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
1412 skb->ip_summed = CHECKSUM_NONE;
1413
1414 return 0;
1415 }
1416
1417 #if IS_ENABLED(CONFIG_IPV6)
1418 static int vxlan6_xmit_skb(struct vxlan_sock *vs,
1419 struct dst_entry *dst, struct sk_buff *skb,
1420 struct net_device *dev, struct in6_addr *saddr,
1421 struct in6_addr *daddr, __u8 prio, __u8 ttl,
1422 __be16 src_port, __be16 dst_port, __be32 vni)
1423 {
1424 struct ipv6hdr *ip6h;
1425 struct vxlanhdr *vxh;
1426 struct udphdr *uh;
1427 int min_headroom;
1428 int err;
1429
1430 if (!skb->encapsulation) {
1431 skb_reset_inner_headers(skb);
1432 skb->encapsulation = 1;
1433 }
1434
1435 skb_scrub_packet(skb, false);
1436
1437 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1438 + VXLAN_HLEN + sizeof(struct ipv6hdr)
1439 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1440
1441 /* Need space for new headers (invalidates iph ptr) */
1442 err = skb_cow_head(skb, min_headroom);
1443 if (unlikely(err))
1444 return err;
1445
1446 if (vlan_tx_tag_present(skb)) {
1447 if (WARN_ON(!__vlan_put_tag(skb,
1448 skb->vlan_proto,
1449 vlan_tx_tag_get(skb))))
1450 return -ENOMEM;
1451
1452 skb->vlan_tci = 0;
1453 }
1454
1455 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1456 vxh->vx_flags = htonl(VXLAN_FLAGS);
1457 vxh->vx_vni = vni;
1458
1459 __skb_push(skb, sizeof(*uh));
1460 skb_reset_transport_header(skb);
1461 uh = udp_hdr(skb);
1462
1463 uh->dest = dst_port;
1464 uh->source = src_port;
1465
1466 uh->len = htons(skb->len);
1467 uh->check = 0;
1468
1469 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1470 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1471 IPSKB_REROUTED);
1472 skb_dst_set(skb, dst);
1473
1474 if (!skb_is_gso(skb) && !(dst->dev->features & NETIF_F_IPV6_CSUM)) {
1475 __wsum csum = skb_checksum(skb, 0, skb->len, 0);
1476 skb->ip_summed = CHECKSUM_UNNECESSARY;
1477 uh->check = csum_ipv6_magic(saddr, daddr, skb->len,
1478 IPPROTO_UDP, csum);
1479 if (uh->check == 0)
1480 uh->check = CSUM_MANGLED_0;
1481 } else {
1482 skb->ip_summed = CHECKSUM_PARTIAL;
1483 skb->csum_start = skb_transport_header(skb) - skb->head;
1484 skb->csum_offset = offsetof(struct udphdr, check);
1485 uh->check = ~csum_ipv6_magic(saddr, daddr,
1486 skb->len, IPPROTO_UDP, 0);
1487 }
1488
1489 __skb_push(skb, sizeof(*ip6h));
1490 skb_reset_network_header(skb);
1491 ip6h = ipv6_hdr(skb);
1492 ip6h->version = 6;
1493 ip6h->priority = prio;
1494 ip6h->flow_lbl[0] = 0;
1495 ip6h->flow_lbl[1] = 0;
1496 ip6h->flow_lbl[2] = 0;
1497 ip6h->payload_len = htons(skb->len);
1498 ip6h->nexthdr = IPPROTO_UDP;
1499 ip6h->hop_limit = ttl;
1500 ip6h->daddr = *daddr;
1501 ip6h->saddr = *saddr;
1502
1503 err = handle_offloads(skb);
1504 if (err)
1505 return err;
1506
1507 ip6tunnel_xmit(skb, dev);
1508 return 0;
1509 }
1510 #endif
1511
1512 int vxlan_xmit_skb(struct vxlan_sock *vs,
1513 struct rtable *rt, struct sk_buff *skb,
1514 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
1515 __be16 src_port, __be16 dst_port, __be32 vni)
1516 {
1517 struct vxlanhdr *vxh;
1518 struct udphdr *uh;
1519 int min_headroom;
1520 int err;
1521
1522 if (!skb->encapsulation) {
1523 skb_reset_inner_headers(skb);
1524 skb->encapsulation = 1;
1525 }
1526
1527 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1528 + VXLAN_HLEN + sizeof(struct iphdr)
1529 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1530
1531 /* Need space for new headers (invalidates iph ptr) */
1532 err = skb_cow_head(skb, min_headroom);
1533 if (unlikely(err))
1534 return err;
1535
1536 if (vlan_tx_tag_present(skb)) {
1537 if (WARN_ON(!__vlan_put_tag(skb,
1538 skb->vlan_proto,
1539 vlan_tx_tag_get(skb))))
1540 return -ENOMEM;
1541
1542 skb->vlan_tci = 0;
1543 }
1544
1545 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1546 vxh->vx_flags = htonl(VXLAN_FLAGS);
1547 vxh->vx_vni = vni;
1548
1549 __skb_push(skb, sizeof(*uh));
1550 skb_reset_transport_header(skb);
1551 uh = udp_hdr(skb);
1552
1553 uh->dest = dst_port;
1554 uh->source = src_port;
1555
1556 uh->len = htons(skb->len);
1557 uh->check = 0;
1558
1559 err = handle_offloads(skb);
1560 if (err)
1561 return err;
1562
1563 return iptunnel_xmit(rt, skb, src, dst, IPPROTO_UDP, tos, ttl, df,
1564 false);
1565 }
1566 EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1567
1568 /* Bypass encapsulation if the destination is local */
1569 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1570 struct vxlan_dev *dst_vxlan)
1571 {
1572 struct pcpu_sw_netstats *tx_stats, *rx_stats;
1573 union vxlan_addr loopback;
1574 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
1575
1576 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1577 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1578 skb->pkt_type = PACKET_HOST;
1579 skb->encapsulation = 0;
1580 skb->dev = dst_vxlan->dev;
1581 __skb_pull(skb, skb_network_offset(skb));
1582
1583 if (remote_ip->sa.sa_family == AF_INET) {
1584 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1585 loopback.sa.sa_family = AF_INET;
1586 #if IS_ENABLED(CONFIG_IPV6)
1587 } else {
1588 loopback.sin6.sin6_addr = in6addr_loopback;
1589 loopback.sa.sa_family = AF_INET6;
1590 #endif
1591 }
1592
1593 if (dst_vxlan->flags & VXLAN_F_LEARN)
1594 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
1595
1596 u64_stats_update_begin(&tx_stats->syncp);
1597 tx_stats->tx_packets++;
1598 tx_stats->tx_bytes += skb->len;
1599 u64_stats_update_end(&tx_stats->syncp);
1600
1601 if (netif_rx(skb) == NET_RX_SUCCESS) {
1602 u64_stats_update_begin(&rx_stats->syncp);
1603 rx_stats->rx_packets++;
1604 rx_stats->rx_bytes += skb->len;
1605 u64_stats_update_end(&rx_stats->syncp);
1606 } else {
1607 skb->dev->stats.rx_dropped++;
1608 }
1609 }
1610
1611 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1612 struct vxlan_rdst *rdst, bool did_rsc)
1613 {
1614 struct vxlan_dev *vxlan = netdev_priv(dev);
1615 struct rtable *rt = NULL;
1616 const struct iphdr *old_iph;
1617 struct flowi4 fl4;
1618 union vxlan_addr *dst;
1619 __be16 src_port = 0, dst_port;
1620 u32 vni;
1621 __be16 df = 0;
1622 __u8 tos, ttl;
1623 int err;
1624
1625 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
1626 vni = rdst->remote_vni;
1627 dst = &rdst->remote_ip;
1628
1629 if (vxlan_addr_any(dst)) {
1630 if (did_rsc) {
1631 /* short-circuited back to local bridge */
1632 vxlan_encap_bypass(skb, vxlan, vxlan);
1633 return;
1634 }
1635 goto drop;
1636 }
1637
1638 old_iph = ip_hdr(skb);
1639
1640 ttl = vxlan->ttl;
1641 if (!ttl && vxlan_addr_multicast(dst))
1642 ttl = 1;
1643
1644 tos = vxlan->tos;
1645 if (tos == 1)
1646 tos = ip_tunnel_get_dsfield(old_iph, skb);
1647
1648 src_port = vxlan_src_port(vxlan->port_min, vxlan->port_max, skb);
1649
1650 if (dst->sa.sa_family == AF_INET) {
1651 memset(&fl4, 0, sizeof(fl4));
1652 fl4.flowi4_oif = rdst->remote_ifindex;
1653 fl4.flowi4_tos = RT_TOS(tos);
1654 fl4.daddr = dst->sin.sin_addr.s_addr;
1655 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1656
1657 rt = ip_route_output_key(dev_net(dev), &fl4);
1658 if (IS_ERR(rt)) {
1659 netdev_dbg(dev, "no route to %pI4\n",
1660 &dst->sin.sin_addr.s_addr);
1661 dev->stats.tx_carrier_errors++;
1662 goto tx_error;
1663 }
1664
1665 if (rt->dst.dev == dev) {
1666 netdev_dbg(dev, "circular route to %pI4\n",
1667 &dst->sin.sin_addr.s_addr);
1668 dev->stats.collisions++;
1669 goto rt_tx_error;
1670 }
1671
1672 /* Bypass encapsulation if the destination is local */
1673 if (rt->rt_flags & RTCF_LOCAL &&
1674 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1675 struct vxlan_dev *dst_vxlan;
1676
1677 ip_rt_put(rt);
1678 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1679 if (!dst_vxlan)
1680 goto tx_error;
1681 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1682 return;
1683 }
1684
1685 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1686 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
1687
1688 err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
1689 fl4.saddr, dst->sin.sin_addr.s_addr,
1690 tos, ttl, df, src_port, dst_port,
1691 htonl(vni << 8));
1692
1693 if (err < 0)
1694 goto rt_tx_error;
1695 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1696 #if IS_ENABLED(CONFIG_IPV6)
1697 } else {
1698 struct sock *sk = vxlan->vn_sock->sock->sk;
1699 struct dst_entry *ndst;
1700 struct flowi6 fl6;
1701 u32 flags;
1702
1703 memset(&fl6, 0, sizeof(fl6));
1704 fl6.flowi6_oif = rdst->remote_ifindex;
1705 fl6.daddr = dst->sin6.sin6_addr;
1706 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
1707 fl6.flowi6_proto = IPPROTO_UDP;
1708
1709 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1710 netdev_dbg(dev, "no route to %pI6\n",
1711 &dst->sin6.sin6_addr);
1712 dev->stats.tx_carrier_errors++;
1713 goto tx_error;
1714 }
1715
1716 if (ndst->dev == dev) {
1717 netdev_dbg(dev, "circular route to %pI6\n",
1718 &dst->sin6.sin6_addr);
1719 dst_release(ndst);
1720 dev->stats.collisions++;
1721 goto tx_error;
1722 }
1723
1724 /* Bypass encapsulation if the destination is local */
1725 flags = ((struct rt6_info *)ndst)->rt6i_flags;
1726 if (flags & RTF_LOCAL &&
1727 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1728 struct vxlan_dev *dst_vxlan;
1729
1730 dst_release(ndst);
1731 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1732 if (!dst_vxlan)
1733 goto tx_error;
1734 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1735 return;
1736 }
1737
1738 ttl = ttl ? : ip6_dst_hoplimit(ndst);
1739
1740 err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
1741 dev, &fl6.saddr, &fl6.daddr, 0, ttl,
1742 src_port, dst_port, htonl(vni << 8));
1743 #endif
1744 }
1745
1746 return;
1747
1748 drop:
1749 dev->stats.tx_dropped++;
1750 goto tx_free;
1751
1752 rt_tx_error:
1753 ip_rt_put(rt);
1754 tx_error:
1755 dev->stats.tx_errors++;
1756 tx_free:
1757 dev_kfree_skb(skb);
1758 }
1759
1760 /* Transmit local packets over Vxlan
1761 *
1762 * Outer IP header inherits ECN and DF from inner header.
1763 * Outer UDP destination is the VXLAN assigned port.
1764 * source port is based on hash of flow
1765 */
1766 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1767 {
1768 struct vxlan_dev *vxlan = netdev_priv(dev);
1769 struct ethhdr *eth;
1770 bool did_rsc = false;
1771 struct vxlan_rdst *rdst, *fdst = NULL;
1772 struct vxlan_fdb *f;
1773
1774 skb_reset_mac_header(skb);
1775 eth = eth_hdr(skb);
1776
1777 if ((vxlan->flags & VXLAN_F_PROXY)) {
1778 if (ntohs(eth->h_proto) == ETH_P_ARP)
1779 return arp_reduce(dev, skb);
1780 #if IS_ENABLED(CONFIG_IPV6)
1781 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
1782 skb->len >= sizeof(struct ipv6hdr) + sizeof(struct nd_msg) &&
1783 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
1784 struct nd_msg *msg;
1785
1786 msg = (struct nd_msg *)skb_transport_header(skb);
1787 if (msg->icmph.icmp6_code == 0 &&
1788 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
1789 return neigh_reduce(dev, skb);
1790 }
1791 #endif
1792 }
1793
1794 f = vxlan_find_mac(vxlan, eth->h_dest);
1795 did_rsc = false;
1796
1797 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
1798 (ntohs(eth->h_proto) == ETH_P_IP ||
1799 ntohs(eth->h_proto) == ETH_P_IPV6)) {
1800 did_rsc = route_shortcircuit(dev, skb);
1801 if (did_rsc)
1802 f = vxlan_find_mac(vxlan, eth->h_dest);
1803 }
1804
1805 if (f == NULL) {
1806 f = vxlan_find_mac(vxlan, all_zeros_mac);
1807 if (f == NULL) {
1808 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1809 !is_multicast_ether_addr(eth->h_dest))
1810 vxlan_fdb_miss(vxlan, eth->h_dest);
1811
1812 dev->stats.tx_dropped++;
1813 kfree_skb(skb);
1814 return NETDEV_TX_OK;
1815 }
1816 }
1817
1818 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1819 struct sk_buff *skb1;
1820
1821 if (!fdst) {
1822 fdst = rdst;
1823 continue;
1824 }
1825 skb1 = skb_clone(skb, GFP_ATOMIC);
1826 if (skb1)
1827 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
1828 }
1829
1830 if (fdst)
1831 vxlan_xmit_one(skb, dev, fdst, did_rsc);
1832 else
1833 kfree_skb(skb);
1834 return NETDEV_TX_OK;
1835 }
1836
1837 /* Walk the forwarding table and purge stale entries */
1838 static void vxlan_cleanup(unsigned long arg)
1839 {
1840 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1841 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1842 unsigned int h;
1843
1844 if (!netif_running(vxlan->dev))
1845 return;
1846
1847 spin_lock_bh(&vxlan->hash_lock);
1848 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1849 struct hlist_node *p, *n;
1850 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1851 struct vxlan_fdb *f
1852 = container_of(p, struct vxlan_fdb, hlist);
1853 unsigned long timeout;
1854
1855 if (f->state & NUD_PERMANENT)
1856 continue;
1857
1858 timeout = f->used + vxlan->age_interval * HZ;
1859 if (time_before_eq(timeout, jiffies)) {
1860 netdev_dbg(vxlan->dev,
1861 "garbage collect %pM\n",
1862 f->eth_addr);
1863 f->state = NUD_STALE;
1864 vxlan_fdb_destroy(vxlan, f);
1865 } else if (time_before(timeout, next_timer))
1866 next_timer = timeout;
1867 }
1868 }
1869 spin_unlock_bh(&vxlan->hash_lock);
1870
1871 mod_timer(&vxlan->age_timer, next_timer);
1872 }
1873
1874 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
1875 {
1876 __u32 vni = vxlan->default_dst.remote_vni;
1877
1878 vxlan->vn_sock = vs;
1879 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
1880 }
1881
1882 /* Setup stats when device is created */
1883 static int vxlan_init(struct net_device *dev)
1884 {
1885 struct vxlan_dev *vxlan = netdev_priv(dev);
1886 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1887 struct vxlan_sock *vs;
1888 int i;
1889
1890 dev->tstats = alloc_percpu(struct pcpu_sw_netstats);
1891 if (!dev->tstats)
1892 return -ENOMEM;
1893
1894 for_each_possible_cpu(i) {
1895 struct pcpu_sw_netstats *vxlan_stats;
1896 vxlan_stats = per_cpu_ptr(dev->tstats, i);
1897 u64_stats_init(&vxlan_stats->syncp);
1898 }
1899
1900
1901 spin_lock(&vn->sock_lock);
1902 vs = vxlan_find_sock(dev_net(dev), vxlan->dst_port);
1903 if (vs) {
1904 /* If we have a socket with same port already, reuse it */
1905 atomic_inc(&vs->refcnt);
1906 vxlan_vs_add_dev(vs, vxlan);
1907 } else {
1908 /* otherwise make new socket outside of RTNL */
1909 dev_hold(dev);
1910 queue_work(vxlan_wq, &vxlan->sock_work);
1911 }
1912 spin_unlock(&vn->sock_lock);
1913
1914 return 0;
1915 }
1916
1917 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
1918 {
1919 struct vxlan_fdb *f;
1920
1921 spin_lock_bh(&vxlan->hash_lock);
1922 f = __vxlan_find_mac(vxlan, all_zeros_mac);
1923 if (f)
1924 vxlan_fdb_destroy(vxlan, f);
1925 spin_unlock_bh(&vxlan->hash_lock);
1926 }
1927
1928 static void vxlan_uninit(struct net_device *dev)
1929 {
1930 struct vxlan_dev *vxlan = netdev_priv(dev);
1931 struct vxlan_sock *vs = vxlan->vn_sock;
1932
1933 vxlan_fdb_delete_default(vxlan);
1934
1935 if (vs)
1936 vxlan_sock_release(vs);
1937 free_percpu(dev->tstats);
1938 }
1939
1940 /* Start ageing timer and join group when device is brought up */
1941 static int vxlan_open(struct net_device *dev)
1942 {
1943 struct vxlan_dev *vxlan = netdev_priv(dev);
1944 struct vxlan_sock *vs = vxlan->vn_sock;
1945
1946 /* socket hasn't been created */
1947 if (!vs)
1948 return -ENOTCONN;
1949
1950 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
1951 vxlan_sock_hold(vs);
1952 dev_hold(dev);
1953 queue_work(vxlan_wq, &vxlan->igmp_join);
1954 }
1955
1956 if (vxlan->age_interval)
1957 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1958
1959 return 0;
1960 }
1961
1962 /* Purge the forwarding table */
1963 static void vxlan_flush(struct vxlan_dev *vxlan)
1964 {
1965 unsigned int h;
1966
1967 spin_lock_bh(&vxlan->hash_lock);
1968 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1969 struct hlist_node *p, *n;
1970 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1971 struct vxlan_fdb *f
1972 = container_of(p, struct vxlan_fdb, hlist);
1973 /* the all_zeros_mac entry is deleted at vxlan_uninit */
1974 if (!is_zero_ether_addr(f->eth_addr))
1975 vxlan_fdb_destroy(vxlan, f);
1976 }
1977 }
1978 spin_unlock_bh(&vxlan->hash_lock);
1979 }
1980
1981 /* Cleanup timer and forwarding table on shutdown */
1982 static int vxlan_stop(struct net_device *dev)
1983 {
1984 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1985 struct vxlan_dev *vxlan = netdev_priv(dev);
1986 struct vxlan_sock *vs = vxlan->vn_sock;
1987
1988 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
1989 !vxlan_group_used(vn, vxlan)) {
1990 vxlan_sock_hold(vs);
1991 dev_hold(dev);
1992 queue_work(vxlan_wq, &vxlan->igmp_leave);
1993 }
1994
1995 del_timer_sync(&vxlan->age_timer);
1996
1997 vxlan_flush(vxlan);
1998
1999 return 0;
2000 }
2001
2002 /* Stub, nothing needs to be done. */
2003 static void vxlan_set_multicast_list(struct net_device *dev)
2004 {
2005 }
2006
2007 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2008 {
2009 struct vxlan_dev *vxlan = netdev_priv(dev);
2010 struct vxlan_rdst *dst = &vxlan->default_dst;
2011 struct net_device *lowerdev;
2012 int max_mtu;
2013
2014 lowerdev = __dev_get_by_index(dev_net(dev), dst->remote_ifindex);
2015 if (lowerdev == NULL)
2016 return eth_change_mtu(dev, new_mtu);
2017
2018 if (dst->remote_ip.sa.sa_family == AF_INET6)
2019 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2020 else
2021 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2022
2023 if (new_mtu < 68 || new_mtu > max_mtu)
2024 return -EINVAL;
2025
2026 dev->mtu = new_mtu;
2027 return 0;
2028 }
2029
2030 static const struct net_device_ops vxlan_netdev_ops = {
2031 .ndo_init = vxlan_init,
2032 .ndo_uninit = vxlan_uninit,
2033 .ndo_open = vxlan_open,
2034 .ndo_stop = vxlan_stop,
2035 .ndo_start_xmit = vxlan_xmit,
2036 .ndo_get_stats64 = ip_tunnel_get_stats64,
2037 .ndo_set_rx_mode = vxlan_set_multicast_list,
2038 .ndo_change_mtu = vxlan_change_mtu,
2039 .ndo_validate_addr = eth_validate_addr,
2040 .ndo_set_mac_address = eth_mac_addr,
2041 .ndo_fdb_add = vxlan_fdb_add,
2042 .ndo_fdb_del = vxlan_fdb_delete,
2043 .ndo_fdb_dump = vxlan_fdb_dump,
2044 };
2045
2046 /* Info for udev, that this is a virtual tunnel endpoint */
2047 static struct device_type vxlan_type = {
2048 .name = "vxlan",
2049 };
2050
2051 /* Calls the ndo_add_vxlan_port of the caller in order to
2052 * supply the listening VXLAN udp ports. Callers are expected
2053 * to implement the ndo_add_vxlan_port.
2054 */
2055 void vxlan_get_rx_port(struct net_device *dev)
2056 {
2057 struct vxlan_sock *vs;
2058 struct net *net = dev_net(dev);
2059 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2060 sa_family_t sa_family;
2061 __be16 port;
2062 unsigned int i;
2063
2064 spin_lock(&vn->sock_lock);
2065 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2066 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2067 port = inet_sk(vs->sock->sk)->inet_sport;
2068 sa_family = vs->sock->sk->sk_family;
2069 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2070 port);
2071 }
2072 }
2073 spin_unlock(&vn->sock_lock);
2074 }
2075 EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2076
2077 /* Initialize the device structure. */
2078 static void vxlan_setup(struct net_device *dev)
2079 {
2080 struct vxlan_dev *vxlan = netdev_priv(dev);
2081 unsigned int h;
2082 int low, high;
2083
2084 eth_hw_addr_random(dev);
2085 ether_setup(dev);
2086 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2087 dev->hard_header_len = ETH_HLEN + VXLAN6_HEADROOM;
2088 else
2089 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
2090
2091 dev->netdev_ops = &vxlan_netdev_ops;
2092 dev->destructor = free_netdev;
2093 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2094
2095 dev->tx_queue_len = 0;
2096 dev->features |= NETIF_F_LLTX;
2097 dev->features |= NETIF_F_NETNS_LOCAL;
2098 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
2099 dev->features |= NETIF_F_RXCSUM;
2100 dev->features |= NETIF_F_GSO_SOFTWARE;
2101
2102 dev->vlan_features = dev->features;
2103 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2104 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2105 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2106 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2107 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
2108 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2109
2110 INIT_LIST_HEAD(&vxlan->next);
2111 spin_lock_init(&vxlan->hash_lock);
2112 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2113 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
2114 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
2115
2116 init_timer_deferrable(&vxlan->age_timer);
2117 vxlan->age_timer.function = vxlan_cleanup;
2118 vxlan->age_timer.data = (unsigned long) vxlan;
2119
2120 inet_get_local_port_range(dev_net(dev), &low, &high);
2121 vxlan->port_min = low;
2122 vxlan->port_max = high;
2123 vxlan->dst_port = htons(vxlan_port);
2124
2125 vxlan->dev = dev;
2126
2127 for (h = 0; h < FDB_HASH_SIZE; ++h)
2128 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2129 }
2130
2131 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2132 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
2133 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2134 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
2135 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2136 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2137 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
2138 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2139 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2140 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2141 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2142 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
2143 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
2144 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2145 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2146 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2147 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
2148 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
2149 };
2150
2151 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2152 {
2153 if (tb[IFLA_ADDRESS]) {
2154 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2155 pr_debug("invalid link address (not ethernet)\n");
2156 return -EINVAL;
2157 }
2158
2159 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2160 pr_debug("invalid all zero ethernet address\n");
2161 return -EADDRNOTAVAIL;
2162 }
2163 }
2164
2165 if (!data)
2166 return -EINVAL;
2167
2168 if (data[IFLA_VXLAN_ID]) {
2169 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2170 if (id >= VXLAN_VID_MASK)
2171 return -ERANGE;
2172 }
2173
2174 if (data[IFLA_VXLAN_PORT_RANGE]) {
2175 const struct ifla_vxlan_port_range *p
2176 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2177
2178 if (ntohs(p->high) < ntohs(p->low)) {
2179 pr_debug("port range %u .. %u not valid\n",
2180 ntohs(p->low), ntohs(p->high));
2181 return -EINVAL;
2182 }
2183 }
2184
2185 return 0;
2186 }
2187
2188 static void vxlan_get_drvinfo(struct net_device *netdev,
2189 struct ethtool_drvinfo *drvinfo)
2190 {
2191 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2192 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2193 }
2194
2195 static const struct ethtool_ops vxlan_ethtool_ops = {
2196 .get_drvinfo = vxlan_get_drvinfo,
2197 .get_link = ethtool_op_get_link,
2198 };
2199
2200 static void vxlan_del_work(struct work_struct *work)
2201 {
2202 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2203
2204 sk_release_kernel(vs->sock->sk);
2205 kfree_rcu(vs, rcu);
2206 }
2207
2208 #if IS_ENABLED(CONFIG_IPV6)
2209 /* Create UDP socket for encapsulation receive. AF_INET6 socket
2210 * could be used for both IPv4 and IPv6 communications, but
2211 * users may set bindv6only=1.
2212 */
2213 static struct socket *create_v6_sock(struct net *net, __be16 port)
2214 {
2215 struct sock *sk;
2216 struct socket *sock;
2217 struct sockaddr_in6 vxlan_addr = {
2218 .sin6_family = AF_INET6,
2219 .sin6_port = port,
2220 };
2221 int rc, val = 1;
2222
2223 rc = sock_create_kern(AF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
2224 if (rc < 0) {
2225 pr_debug("UDPv6 socket create failed\n");
2226 return ERR_PTR(rc);
2227 }
2228
2229 /* Put in proper namespace */
2230 sk = sock->sk;
2231 sk_change_net(sk, net);
2232
2233 kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
2234 (char *)&val, sizeof(val));
2235 rc = kernel_bind(sock, (struct sockaddr *)&vxlan_addr,
2236 sizeof(struct sockaddr_in6));
2237 if (rc < 0) {
2238 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2239 &vxlan_addr.sin6_addr, ntohs(vxlan_addr.sin6_port), rc);
2240 sk_release_kernel(sk);
2241 return ERR_PTR(rc);
2242 }
2243 /* At this point, IPv6 module should have been loaded in
2244 * sock_create_kern().
2245 */
2246 BUG_ON(!ipv6_stub);
2247
2248 /* Disable multicast loopback */
2249 inet_sk(sk)->mc_loop = 0;
2250 return sock;
2251 }
2252
2253 #else
2254
2255 static struct socket *create_v6_sock(struct net *net, __be16 port)
2256 {
2257 return ERR_PTR(-EPFNOSUPPORT);
2258 }
2259 #endif
2260
2261 static struct socket *create_v4_sock(struct net *net, __be16 port)
2262 {
2263 struct sock *sk;
2264 struct socket *sock;
2265 struct sockaddr_in vxlan_addr = {
2266 .sin_family = AF_INET,
2267 .sin_addr.s_addr = htonl(INADDR_ANY),
2268 .sin_port = port,
2269 };
2270 int rc;
2271
2272 /* Create UDP socket for encapsulation receive. */
2273 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
2274 if (rc < 0) {
2275 pr_debug("UDP socket create failed\n");
2276 return ERR_PTR(rc);
2277 }
2278
2279 /* Put in proper namespace */
2280 sk = sock->sk;
2281 sk_change_net(sk, net);
2282
2283 rc = kernel_bind(sock, (struct sockaddr *) &vxlan_addr,
2284 sizeof(vxlan_addr));
2285 if (rc < 0) {
2286 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2287 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
2288 sk_release_kernel(sk);
2289 return ERR_PTR(rc);
2290 }
2291
2292 /* Disable multicast loopback */
2293 inet_sk(sk)->mc_loop = 0;
2294 return sock;
2295 }
2296
2297 /* Create new listen socket if needed */
2298 static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
2299 vxlan_rcv_t *rcv, void *data, bool ipv6)
2300 {
2301 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2302 struct vxlan_sock *vs;
2303 struct socket *sock;
2304 struct sock *sk;
2305 unsigned int h;
2306
2307 vs = kmalloc(sizeof(*vs), GFP_KERNEL);
2308 if (!vs)
2309 return ERR_PTR(-ENOMEM);
2310
2311 for (h = 0; h < VNI_HASH_SIZE; ++h)
2312 INIT_HLIST_HEAD(&vs->vni_list[h]);
2313
2314 INIT_WORK(&vs->del_work, vxlan_del_work);
2315
2316 if (ipv6)
2317 sock = create_v6_sock(net, port);
2318 else
2319 sock = create_v4_sock(net, port);
2320 if (IS_ERR(sock)) {
2321 kfree(vs);
2322 return ERR_CAST(sock);
2323 }
2324
2325 vs->sock = sock;
2326 sk = sock->sk;
2327 atomic_set(&vs->refcnt, 1);
2328 vs->rcv = rcv;
2329 vs->data = data;
2330 rcu_assign_sk_user_data(vs->sock->sk, vs);
2331
2332 spin_lock(&vn->sock_lock);
2333 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2334 vxlan_notify_add_rx_port(sk);
2335 spin_unlock(&vn->sock_lock);
2336
2337 /* Mark socket as an encapsulation socket. */
2338 udp_sk(sk)->encap_type = 1;
2339 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
2340 #if IS_ENABLED(CONFIG_IPV6)
2341 if (ipv6)
2342 ipv6_stub->udpv6_encap_enable();
2343 else
2344 #endif
2345 udp_encap_enable();
2346
2347 return vs;
2348 }
2349
2350 struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2351 vxlan_rcv_t *rcv, void *data,
2352 bool no_share, bool ipv6)
2353 {
2354 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2355 struct vxlan_sock *vs;
2356
2357 vs = vxlan_socket_create(net, port, rcv, data, ipv6);
2358 if (!IS_ERR(vs))
2359 return vs;
2360
2361 if (no_share) /* Return error if sharing is not allowed. */
2362 return vs;
2363
2364 spin_lock(&vn->sock_lock);
2365 vs = vxlan_find_sock(net, port);
2366 if (vs) {
2367 if (vs->rcv == rcv)
2368 atomic_inc(&vs->refcnt);
2369 else
2370 vs = ERR_PTR(-EBUSY);
2371 }
2372 spin_unlock(&vn->sock_lock);
2373
2374 if (!vs)
2375 vs = ERR_PTR(-EINVAL);
2376
2377 return vs;
2378 }
2379 EXPORT_SYMBOL_GPL(vxlan_sock_add);
2380
2381 /* Scheduled at device creation to bind to a socket */
2382 static void vxlan_sock_work(struct work_struct *work)
2383 {
2384 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
2385 struct net *net = dev_net(vxlan->dev);
2386 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2387 __be16 port = vxlan->dst_port;
2388 struct vxlan_sock *nvs;
2389
2390 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags & VXLAN_F_IPV6);
2391 spin_lock(&vn->sock_lock);
2392 if (!IS_ERR(nvs))
2393 vxlan_vs_add_dev(nvs, vxlan);
2394 spin_unlock(&vn->sock_lock);
2395
2396 dev_put(vxlan->dev);
2397 }
2398
2399 static int vxlan_newlink(struct net *net, struct net_device *dev,
2400 struct nlattr *tb[], struct nlattr *data[])
2401 {
2402 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2403 struct vxlan_dev *vxlan = netdev_priv(dev);
2404 struct vxlan_rdst *dst = &vxlan->default_dst;
2405 __u32 vni;
2406 int err;
2407 bool use_ipv6 = false;
2408
2409 if (!data[IFLA_VXLAN_ID])
2410 return -EINVAL;
2411
2412 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
2413 dst->remote_vni = vni;
2414
2415 if (data[IFLA_VXLAN_GROUP]) {
2416 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
2417 dst->remote_ip.sa.sa_family = AF_INET;
2418 } else if (data[IFLA_VXLAN_GROUP6]) {
2419 if (!IS_ENABLED(CONFIG_IPV6))
2420 return -EPFNOSUPPORT;
2421
2422 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2423 sizeof(struct in6_addr));
2424 dst->remote_ip.sa.sa_family = AF_INET6;
2425 use_ipv6 = true;
2426 }
2427
2428 if (data[IFLA_VXLAN_LOCAL]) {
2429 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2430 vxlan->saddr.sa.sa_family = AF_INET;
2431 } else if (data[IFLA_VXLAN_LOCAL6]) {
2432 if (!IS_ENABLED(CONFIG_IPV6))
2433 return -EPFNOSUPPORT;
2434
2435 /* TODO: respect scope id */
2436 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2437 sizeof(struct in6_addr));
2438 vxlan->saddr.sa.sa_family = AF_INET6;
2439 use_ipv6 = true;
2440 }
2441
2442 if (data[IFLA_VXLAN_LINK] &&
2443 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
2444 struct net_device *lowerdev
2445 = __dev_get_by_index(net, dst->remote_ifindex);
2446
2447 if (!lowerdev) {
2448 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
2449 return -ENODEV;
2450 }
2451
2452 #if IS_ENABLED(CONFIG_IPV6)
2453 if (use_ipv6) {
2454 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2455 if (idev && idev->cnf.disable_ipv6) {
2456 pr_info("IPv6 is disabled via sysctl\n");
2457 return -EPERM;
2458 }
2459 vxlan->flags |= VXLAN_F_IPV6;
2460 }
2461 #endif
2462
2463 if (!tb[IFLA_MTU])
2464 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2465
2466 /* update header length based on lower device */
2467 dev->hard_header_len = lowerdev->hard_header_len +
2468 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2469 } else if (use_ipv6)
2470 vxlan->flags |= VXLAN_F_IPV6;
2471
2472 if (data[IFLA_VXLAN_TOS])
2473 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2474
2475 if (data[IFLA_VXLAN_TTL])
2476 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2477
2478 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
2479 vxlan->flags |= VXLAN_F_LEARN;
2480
2481 if (data[IFLA_VXLAN_AGEING])
2482 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2483 else
2484 vxlan->age_interval = FDB_AGE_DEFAULT;
2485
2486 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2487 vxlan->flags |= VXLAN_F_PROXY;
2488
2489 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2490 vxlan->flags |= VXLAN_F_RSC;
2491
2492 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2493 vxlan->flags |= VXLAN_F_L2MISS;
2494
2495 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2496 vxlan->flags |= VXLAN_F_L3MISS;
2497
2498 if (data[IFLA_VXLAN_LIMIT])
2499 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2500
2501 if (data[IFLA_VXLAN_PORT_RANGE]) {
2502 const struct ifla_vxlan_port_range *p
2503 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2504 vxlan->port_min = ntohs(p->low);
2505 vxlan->port_max = ntohs(p->high);
2506 }
2507
2508 if (data[IFLA_VXLAN_PORT])
2509 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2510
2511 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
2512 pr_info("duplicate VNI %u\n", vni);
2513 return -EEXIST;
2514 }
2515
2516 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
2517
2518 /* create an fdb entry for a valid default destination */
2519 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2520 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2521 &vxlan->default_dst.remote_ip,
2522 NUD_REACHABLE|NUD_PERMANENT,
2523 NLM_F_EXCL|NLM_F_CREATE,
2524 vxlan->dst_port,
2525 vxlan->default_dst.remote_vni,
2526 vxlan->default_dst.remote_ifindex,
2527 NTF_SELF);
2528 if (err)
2529 return err;
2530 }
2531
2532 err = register_netdevice(dev);
2533 if (err) {
2534 vxlan_fdb_delete_default(vxlan);
2535 return err;
2536 }
2537
2538 list_add(&vxlan->next, &vn->vxlan_list);
2539
2540 return 0;
2541 }
2542
2543 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2544 {
2545 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
2546 struct vxlan_dev *vxlan = netdev_priv(dev);
2547
2548 spin_lock(&vn->sock_lock);
2549 if (!hlist_unhashed(&vxlan->hlist))
2550 hlist_del_rcu(&vxlan->hlist);
2551 spin_unlock(&vn->sock_lock);
2552
2553 list_del(&vxlan->next);
2554 unregister_netdevice_queue(dev, head);
2555 }
2556
2557 static size_t vxlan_get_size(const struct net_device *dev)
2558 {
2559
2560 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
2561 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
2562 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
2563 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
2564 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2565 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2566 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
2567 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2568 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2569 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2570 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
2571 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2572 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
2573 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
2574 nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */
2575 0;
2576 }
2577
2578 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2579 {
2580 const struct vxlan_dev *vxlan = netdev_priv(dev);
2581 const struct vxlan_rdst *dst = &vxlan->default_dst;
2582 struct ifla_vxlan_port_range ports = {
2583 .low = htons(vxlan->port_min),
2584 .high = htons(vxlan->port_max),
2585 };
2586
2587 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
2588 goto nla_put_failure;
2589
2590 if (!vxlan_addr_any(&dst->remote_ip)) {
2591 if (dst->remote_ip.sa.sa_family == AF_INET) {
2592 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2593 dst->remote_ip.sin.sin_addr.s_addr))
2594 goto nla_put_failure;
2595 #if IS_ENABLED(CONFIG_IPV6)
2596 } else {
2597 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2598 &dst->remote_ip.sin6.sin6_addr))
2599 goto nla_put_failure;
2600 #endif
2601 }
2602 }
2603
2604 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
2605 goto nla_put_failure;
2606
2607 if (!vxlan_addr_any(&vxlan->saddr)) {
2608 if (vxlan->saddr.sa.sa_family == AF_INET) {
2609 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2610 vxlan->saddr.sin.sin_addr.s_addr))
2611 goto nla_put_failure;
2612 #if IS_ENABLED(CONFIG_IPV6)
2613 } else {
2614 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2615 &vxlan->saddr.sin6.sin6_addr))
2616 goto nla_put_failure;
2617 #endif
2618 }
2619 }
2620
2621 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2622 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
2623 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2624 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2625 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2626 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2627 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2628 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2629 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2630 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2631 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
2632 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
2633 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
2634 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port))
2635 goto nla_put_failure;
2636
2637 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2638 goto nla_put_failure;
2639
2640 return 0;
2641
2642 nla_put_failure:
2643 return -EMSGSIZE;
2644 }
2645
2646 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2647 .kind = "vxlan",
2648 .maxtype = IFLA_VXLAN_MAX,
2649 .policy = vxlan_policy,
2650 .priv_size = sizeof(struct vxlan_dev),
2651 .setup = vxlan_setup,
2652 .validate = vxlan_validate,
2653 .newlink = vxlan_newlink,
2654 .dellink = vxlan_dellink,
2655 .get_size = vxlan_get_size,
2656 .fill_info = vxlan_fill_info,
2657 };
2658
2659 static __net_init int vxlan_init_net(struct net *net)
2660 {
2661 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2662 unsigned int h;
2663
2664 INIT_LIST_HEAD(&vn->vxlan_list);
2665 spin_lock_init(&vn->sock_lock);
2666
2667 for (h = 0; h < PORT_HASH_SIZE; ++h)
2668 INIT_HLIST_HEAD(&vn->sock_list[h]);
2669
2670 return 0;
2671 }
2672
2673 static __net_exit void vxlan_exit_net(struct net *net)
2674 {
2675 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2676 struct vxlan_dev *vxlan;
2677 LIST_HEAD(list);
2678
2679 rtnl_lock();
2680 list_for_each_entry(vxlan, &vn->vxlan_list, next)
2681 unregister_netdevice_queue(vxlan->dev, &list);
2682 unregister_netdevice_many(&list);
2683 rtnl_unlock();
2684 }
2685
2686 static struct pernet_operations vxlan_net_ops = {
2687 .init = vxlan_init_net,
2688 .exit = vxlan_exit_net,
2689 .id = &vxlan_net_id,
2690 .size = sizeof(struct vxlan_net),
2691 };
2692
2693 static int __init vxlan_init_module(void)
2694 {
2695 int rc;
2696
2697 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2698 if (!vxlan_wq)
2699 return -ENOMEM;
2700
2701 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2702
2703 rc = register_pernet_device(&vxlan_net_ops);
2704 if (rc)
2705 goto out1;
2706
2707 rc = rtnl_link_register(&vxlan_link_ops);
2708 if (rc)
2709 goto out2;
2710
2711 return 0;
2712
2713 out2:
2714 unregister_pernet_device(&vxlan_net_ops);
2715 out1:
2716 destroy_workqueue(vxlan_wq);
2717 return rc;
2718 }
2719 late_initcall(vxlan_init_module);
2720
2721 static void __exit vxlan_cleanup_module(void)
2722 {
2723 rtnl_link_unregister(&vxlan_link_ops);
2724 destroy_workqueue(vxlan_wq);
2725 unregister_pernet_device(&vxlan_net_ops);
2726 rcu_barrier();
2727 }
2728 module_exit(vxlan_cleanup_module);
2729
2730 MODULE_LICENSE("GPL");
2731 MODULE_VERSION(VXLAN_VERSION);
2732 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
2733 MODULE_ALIAS_RTNL_LINK("vxlan");
Linux kernel - Deliverables
This page took 0.089483 seconds and 5 git commands to generate.