projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
bridge: fdb dumping takes a filter device
[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 #include <net/protocol.h>
44 #if IS_ENABLED(CONFIG_IPV6)
45 #include <net/ipv6.h>
46 #include <net/addrconf.h>
47 #include <net/ip6_tunnel.h>
48 #include <net/ip6_checksum.h>
49 #endif
50
51 #define VXLAN_VERSION "0.1"
52
53 #define PORT_HASH_BITS 8
54 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
55 #define VNI_HASH_BITS 10
56 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
57 #define FDB_HASH_BITS 8
58 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
59 #define FDB_AGE_DEFAULT 300 /* 5 min */
60 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
61
62 #define VXLAN_N_VID (1u << 24)
63 #define VXLAN_VID_MASK (VXLAN_N_VID - 1)
64 #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
65
66 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
67
68 /* VXLAN protocol header */
69 struct vxlanhdr {
70 __be32 vx_flags;
71 __be32 vx_vni;
72 };
73
74 /* UDP port for VXLAN traffic.
75 * The IANA assigned port is 4789, but the Linux default is 8472
76 * for compatibility with early adopters.
77 */
78 static unsigned short vxlan_port __read_mostly = 8472;
79 module_param_named(udp_port, vxlan_port, ushort, 0444);
80 MODULE_PARM_DESC(udp_port, "Destination UDP port");
81
82 static bool log_ecn_error = true;
83 module_param(log_ecn_error, bool, 0644);
84 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
85
86 static int vxlan_net_id;
87
88 static const u8 all_zeros_mac[ETH_ALEN];
89
90 /* per-network namespace private data for this module */
91 struct vxlan_net {
92 struct list_head vxlan_list;
93 struct hlist_head sock_list[PORT_HASH_SIZE];
94 spinlock_t sock_lock;
95 };
96
97 union vxlan_addr {
98 struct sockaddr_in sin;
99 struct sockaddr_in6 sin6;
100 struct sockaddr sa;
101 };
102
103 struct vxlan_rdst {
104 union vxlan_addr remote_ip;
105 __be16 remote_port;
106 u32 remote_vni;
107 u32 remote_ifindex;
108 struct list_head list;
109 struct rcu_head rcu;
110 };
111
112 /* Forwarding table entry */
113 struct vxlan_fdb {
114 struct hlist_node hlist; /* linked list of entries */
115 struct rcu_head rcu;
116 unsigned long updated; /* jiffies */
117 unsigned long used;
118 struct list_head remotes;
119 u16 state; /* see ndm_state */
120 u8 flags; /* see ndm_flags */
121 u8 eth_addr[ETH_ALEN];
122 };
123
124 /* Pseudo network device */
125 struct vxlan_dev {
126 struct hlist_node hlist; /* vni hash table */
127 struct list_head next; /* vxlan's per namespace list */
128 struct vxlan_sock *vn_sock; /* listening socket */
129 struct net_device *dev;
130 struct net *net; /* netns for packet i/o */
131 struct vxlan_rdst default_dst; /* default destination */
132 union vxlan_addr saddr; /* source address */
133 __be16 dst_port;
134 __u16 port_min; /* source port range */
135 __u16 port_max;
136 __u8 tos; /* TOS override */
137 __u8 ttl;
138 u32 flags; /* VXLAN_F_* in vxlan.h */
139
140 struct work_struct sock_work;
141 struct work_struct igmp_join;
142 struct work_struct igmp_leave;
143
144 unsigned long age_interval;
145 struct timer_list age_timer;
146 spinlock_t hash_lock;
147 unsigned int addrcnt;
148 unsigned int addrmax;
149
150 struct hlist_head fdb_head[FDB_HASH_SIZE];
151 };
152
153 /* salt for hash table */
154 static u32 vxlan_salt __read_mostly;
155 static struct workqueue_struct *vxlan_wq;
156
157 static void vxlan_sock_work(struct work_struct *work);
158
159 #if IS_ENABLED(CONFIG_IPV6)
160 static inline
161 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
162 {
163 if (a->sa.sa_family != b->sa.sa_family)
164 return false;
165 if (a->sa.sa_family == AF_INET6)
166 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
167 else
168 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
169 }
170
171 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
172 {
173 if (ipa->sa.sa_family == AF_INET6)
174 return ipv6_addr_any(&ipa->sin6.sin6_addr);
175 else
176 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
177 }
178
179 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
180 {
181 if (ipa->sa.sa_family == AF_INET6)
182 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
183 else
184 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
185 }
186
187 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
188 {
189 if (nla_len(nla) >= sizeof(struct in6_addr)) {
190 nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr));
191 ip->sa.sa_family = AF_INET6;
192 return 0;
193 } else if (nla_len(nla) >= sizeof(__be32)) {
194 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
195 ip->sa.sa_family = AF_INET;
196 return 0;
197 } else {
198 return -EAFNOSUPPORT;
199 }
200 }
201
202 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
203 const union vxlan_addr *ip)
204 {
205 if (ip->sa.sa_family == AF_INET6)
206 return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr);
207 else
208 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
209 }
210
211 #else /* !CONFIG_IPV6 */
212
213 static inline
214 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
215 {
216 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
217 }
218
219 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
220 {
221 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
222 }
223
224 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
225 {
226 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
227 }
228
229 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
230 {
231 if (nla_len(nla) >= sizeof(struct in6_addr)) {
232 return -EAFNOSUPPORT;
233 } else if (nla_len(nla) >= sizeof(__be32)) {
234 ip->sin.sin_addr.s_addr = nla_get_be32(nla);
235 ip->sa.sa_family = AF_INET;
236 return 0;
237 } else {
238 return -EAFNOSUPPORT;
239 }
240 }
241
242 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
243 const union vxlan_addr *ip)
244 {
245 return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr);
246 }
247 #endif
248
249 /* Virtual Network hash table head */
250 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
251 {
252 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
253 }
254
255 /* Socket hash table head */
256 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
257 {
258 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
259
260 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
261 }
262
263 /* First remote destination for a forwarding entry.
264 * Guaranteed to be non-NULL because remotes are never deleted.
265 */
266 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
267 {
268 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
269 }
270
271 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
272 {
273 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
274 }
275
276 /* Find VXLAN socket based on network namespace and UDP port */
277 static struct vxlan_sock *vxlan_find_sock(struct net *net, __be16 port)
278 {
279 struct vxlan_sock *vs;
280
281 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
282 if (inet_sk(vs->sock->sk)->inet_sport == port)
283 return vs;
284 }
285 return NULL;
286 }
287
288 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
289 {
290 struct vxlan_dev *vxlan;
291
292 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
293 if (vxlan->default_dst.remote_vni == id)
294 return vxlan;
295 }
296
297 return NULL;
298 }
299
300 /* Look up VNI in a per net namespace table */
301 static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port)
302 {
303 struct vxlan_sock *vs;
304
305 vs = vxlan_find_sock(net, port);
306 if (!vs)
307 return NULL;
308
309 return vxlan_vs_find_vni(vs, id);
310 }
311
312 /* Fill in neighbour message in skbuff. */
313 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
314 const struct vxlan_fdb *fdb,
315 u32 portid, u32 seq, int type, unsigned int flags,
316 const struct vxlan_rdst *rdst)
317 {
318 unsigned long now = jiffies;
319 struct nda_cacheinfo ci;
320 struct nlmsghdr *nlh;
321 struct ndmsg *ndm;
322 bool send_ip, send_eth;
323
324 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
325 if (nlh == NULL)
326 return -EMSGSIZE;
327
328 ndm = nlmsg_data(nlh);
329 memset(ndm, 0, sizeof(*ndm));
330
331 send_eth = send_ip = true;
332
333 if (type == RTM_GETNEIGH) {
334 ndm->ndm_family = AF_INET;
335 send_ip = !vxlan_addr_any(&rdst->remote_ip);
336 send_eth = !is_zero_ether_addr(fdb->eth_addr);
337 } else
338 ndm->ndm_family = AF_BRIDGE;
339 ndm->ndm_state = fdb->state;
340 ndm->ndm_ifindex = vxlan->dev->ifindex;
341 ndm->ndm_flags = fdb->flags;
342 ndm->ndm_type = NDA_DST;
343
344 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
345 goto nla_put_failure;
346
347 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
348 goto nla_put_failure;
349
350 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
351 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
352 goto nla_put_failure;
353 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
354 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
355 goto nla_put_failure;
356 if (rdst->remote_ifindex &&
357 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
358 goto nla_put_failure;
359
360 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
361 ci.ndm_confirmed = 0;
362 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
363 ci.ndm_refcnt = 0;
364
365 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
366 goto nla_put_failure;
367
368 return nlmsg_end(skb, nlh);
369
370 nla_put_failure:
371 nlmsg_cancel(skb, nlh);
372 return -EMSGSIZE;
373 }
374
375 static inline size_t vxlan_nlmsg_size(void)
376 {
377 return NLMSG_ALIGN(sizeof(struct ndmsg))
378 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
379 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
380 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
381 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
382 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
383 + nla_total_size(sizeof(struct nda_cacheinfo));
384 }
385
386 static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
387 struct vxlan_rdst *rd, int type)
388 {
389 struct net *net = dev_net(vxlan->dev);
390 struct sk_buff *skb;
391 int err = -ENOBUFS;
392
393 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
394 if (skb == NULL)
395 goto errout;
396
397 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
398 if (err < 0) {
399 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
400 WARN_ON(err == -EMSGSIZE);
401 kfree_skb(skb);
402 goto errout;
403 }
404
405 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
406 return;
407 errout:
408 if (err < 0)
409 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
410 }
411
412 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
413 {
414 struct vxlan_dev *vxlan = netdev_priv(dev);
415 struct vxlan_fdb f = {
416 .state = NUD_STALE,
417 };
418 struct vxlan_rdst remote = {
419 .remote_ip = *ipa, /* goes to NDA_DST */
420 .remote_vni = VXLAN_N_VID,
421 };
422
423 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
424 }
425
426 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
427 {
428 struct vxlan_fdb f = {
429 .state = NUD_STALE,
430 };
431 struct vxlan_rdst remote = { };
432
433 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
434
435 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
436 }
437
438 /* Hash Ethernet address */
439 static u32 eth_hash(const unsigned char *addr)
440 {
441 u64 value = get_unaligned((u64 *)addr);
442
443 /* only want 6 bytes */
444 #ifdef __BIG_ENDIAN
445 value >>= 16;
446 #else
447 value <<= 16;
448 #endif
449 return hash_64(value, FDB_HASH_BITS);
450 }
451
452 /* Hash chain to use given mac address */
453 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
454 const u8 *mac)
455 {
456 return &vxlan->fdb_head[eth_hash(mac)];
457 }
458
459 /* Look up Ethernet address in forwarding table */
460 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
461 const u8 *mac)
462 {
463 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
464 struct vxlan_fdb *f;
465
466 hlist_for_each_entry_rcu(f, head, hlist) {
467 if (ether_addr_equal(mac, f->eth_addr))
468 return f;
469 }
470
471 return NULL;
472 }
473
474 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
475 const u8 *mac)
476 {
477 struct vxlan_fdb *f;
478
479 f = __vxlan_find_mac(vxlan, mac);
480 if (f)
481 f->used = jiffies;
482
483 return f;
484 }
485
486 /* caller should hold vxlan->hash_lock */
487 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
488 union vxlan_addr *ip, __be16 port,
489 __u32 vni, __u32 ifindex)
490 {
491 struct vxlan_rdst *rd;
492
493 list_for_each_entry(rd, &f->remotes, list) {
494 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
495 rd->remote_port == port &&
496 rd->remote_vni == vni &&
497 rd->remote_ifindex == ifindex)
498 return rd;
499 }
500
501 return NULL;
502 }
503
504 /* Replace destination of unicast mac */
505 static int vxlan_fdb_replace(struct vxlan_fdb *f,
506 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
507 {
508 struct vxlan_rdst *rd;
509
510 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
511 if (rd)
512 return 0;
513
514 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
515 if (!rd)
516 return 0;
517 rd->remote_ip = *ip;
518 rd->remote_port = port;
519 rd->remote_vni = vni;
520 rd->remote_ifindex = ifindex;
521 return 1;
522 }
523
524 /* Add/update destinations for multicast */
525 static int vxlan_fdb_append(struct vxlan_fdb *f,
526 union vxlan_addr *ip, __be16 port, __u32 vni,
527 __u32 ifindex, struct vxlan_rdst **rdp)
528 {
529 struct vxlan_rdst *rd;
530
531 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
532 if (rd)
533 return 0;
534
535 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
536 if (rd == NULL)
537 return -ENOBUFS;
538 rd->remote_ip = *ip;
539 rd->remote_port = port;
540 rd->remote_vni = vni;
541 rd->remote_ifindex = ifindex;
542
543 list_add_tail_rcu(&rd->list, &f->remotes);
544
545 *rdp = rd;
546 return 1;
547 }
548
549 static struct sk_buff **vxlan_gro_receive(struct sk_buff **head, struct sk_buff *skb)
550 {
551 struct sk_buff *p, **pp = NULL;
552 struct vxlanhdr *vh, *vh2;
553 struct ethhdr *eh, *eh2;
554 unsigned int hlen, off_vx, off_eth;
555 const struct packet_offload *ptype;
556 __be16 type;
557 int flush = 1;
558
559 off_vx = skb_gro_offset(skb);
560 hlen = off_vx + sizeof(*vh);
561 vh = skb_gro_header_fast(skb, off_vx);
562 if (skb_gro_header_hard(skb, hlen)) {
563 vh = skb_gro_header_slow(skb, hlen, off_vx);
564 if (unlikely(!vh))
565 goto out;
566 }
567 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
568 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
569
570 off_eth = skb_gro_offset(skb);
571 hlen = off_eth + sizeof(*eh);
572 eh = skb_gro_header_fast(skb, off_eth);
573 if (skb_gro_header_hard(skb, hlen)) {
574 eh = skb_gro_header_slow(skb, hlen, off_eth);
575 if (unlikely(!eh))
576 goto out;
577 }
578
579 flush = 0;
580
581 for (p = *head; p; p = p->next) {
582 if (!NAPI_GRO_CB(p)->same_flow)
583 continue;
584
585 vh2 = (struct vxlanhdr *)(p->data + off_vx);
586 eh2 = (struct ethhdr *)(p->data + off_eth);
587 if (vh->vx_vni != vh2->vx_vni || compare_ether_header(eh, eh2)) {
588 NAPI_GRO_CB(p)->same_flow = 0;
589 continue;
590 }
591 }
592
593 type = eh->h_proto;
594
595 rcu_read_lock();
596 ptype = gro_find_receive_by_type(type);
597 if (ptype == NULL) {
598 flush = 1;
599 goto out_unlock;
600 }
601
602 skb_gro_pull(skb, sizeof(*eh)); /* pull inner eth header */
603 skb_gro_postpull_rcsum(skb, eh, sizeof(*eh));
604 pp = ptype->callbacks.gro_receive(head, skb);
605
606 out_unlock:
607 rcu_read_unlock();
608 out:
609 NAPI_GRO_CB(skb)->flush |= flush;
610
611 return pp;
612 }
613
614 static int vxlan_gro_complete(struct sk_buff *skb, int nhoff)
615 {
616 struct ethhdr *eh;
617 struct packet_offload *ptype;
618 __be16 type;
619 int vxlan_len = sizeof(struct vxlanhdr) + sizeof(struct ethhdr);
620 int err = -ENOSYS;
621
622 eh = (struct ethhdr *)(skb->data + nhoff + sizeof(struct vxlanhdr));
623 type = eh->h_proto;
624
625 rcu_read_lock();
626 ptype = gro_find_complete_by_type(type);
627 if (ptype != NULL)
628 err = ptype->callbacks.gro_complete(skb, nhoff + vxlan_len);
629
630 rcu_read_unlock();
631 return err;
632 }
633
634 /* Notify netdevs that UDP port started listening */
635 static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
636 {
637 struct net_device *dev;
638 struct sock *sk = vs->sock->sk;
639 struct net *net = sock_net(sk);
640 sa_family_t sa_family = sk->sk_family;
641 __be16 port = inet_sk(sk)->inet_sport;
642 int err;
643
644 if (sa_family == AF_INET) {
645 err = udp_add_offload(&vs->udp_offloads);
646 if (err)
647 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
648 }
649
650 rcu_read_lock();
651 for_each_netdev_rcu(net, dev) {
652 if (dev->netdev_ops->ndo_add_vxlan_port)
653 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
654 port);
655 }
656 rcu_read_unlock();
657 }
658
659 /* Notify netdevs that UDP port is no more listening */
660 static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
661 {
662 struct net_device *dev;
663 struct sock *sk = vs->sock->sk;
664 struct net *net = sock_net(sk);
665 sa_family_t sa_family = sk->sk_family;
666 __be16 port = inet_sk(sk)->inet_sport;
667
668 rcu_read_lock();
669 for_each_netdev_rcu(net, dev) {
670 if (dev->netdev_ops->ndo_del_vxlan_port)
671 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
672 port);
673 }
674 rcu_read_unlock();
675
676 if (sa_family == AF_INET)
677 udp_del_offload(&vs->udp_offloads);
678 }
679
680 /* Add new entry to forwarding table -- assumes lock held */
681 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
682 const u8 *mac, union vxlan_addr *ip,
683 __u16 state, __u16 flags,
684 __be16 port, __u32 vni, __u32 ifindex,
685 __u8 ndm_flags)
686 {
687 struct vxlan_rdst *rd = NULL;
688 struct vxlan_fdb *f;
689 int notify = 0;
690
691 f = __vxlan_find_mac(vxlan, mac);
692 if (f) {
693 if (flags & NLM_F_EXCL) {
694 netdev_dbg(vxlan->dev,
695 "lost race to create %pM\n", mac);
696 return -EEXIST;
697 }
698 if (f->state != state) {
699 f->state = state;
700 f->updated = jiffies;
701 notify = 1;
702 }
703 if (f->flags != ndm_flags) {
704 f->flags = ndm_flags;
705 f->updated = jiffies;
706 notify = 1;
707 }
708 if ((flags & NLM_F_REPLACE)) {
709 /* Only change unicasts */
710 if (!(is_multicast_ether_addr(f->eth_addr) ||
711 is_zero_ether_addr(f->eth_addr))) {
712 int rc = vxlan_fdb_replace(f, ip, port, vni,
713 ifindex);
714
715 if (rc < 0)
716 return rc;
717 notify |= rc;
718 } else
719 return -EOPNOTSUPP;
720 }
721 if ((flags & NLM_F_APPEND) &&
722 (is_multicast_ether_addr(f->eth_addr) ||
723 is_zero_ether_addr(f->eth_addr))) {
724 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
725 &rd);
726
727 if (rc < 0)
728 return rc;
729 notify |= rc;
730 }
731 } else {
732 if (!(flags & NLM_F_CREATE))
733 return -ENOENT;
734
735 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
736 return -ENOSPC;
737
738 /* Disallow replace to add a multicast entry */
739 if ((flags & NLM_F_REPLACE) &&
740 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
741 return -EOPNOTSUPP;
742
743 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
744 f = kmalloc(sizeof(*f), GFP_ATOMIC);
745 if (!f)
746 return -ENOMEM;
747
748 notify = 1;
749 f->state = state;
750 f->flags = ndm_flags;
751 f->updated = f->used = jiffies;
752 INIT_LIST_HEAD(&f->remotes);
753 memcpy(f->eth_addr, mac, ETH_ALEN);
754
755 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
756
757 ++vxlan->addrcnt;
758 hlist_add_head_rcu(&f->hlist,
759 vxlan_fdb_head(vxlan, mac));
760 }
761
762 if (notify) {
763 if (rd == NULL)
764 rd = first_remote_rtnl(f);
765 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
766 }
767
768 return 0;
769 }
770
771 static void vxlan_fdb_free(struct rcu_head *head)
772 {
773 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
774 struct vxlan_rdst *rd, *nd;
775
776 list_for_each_entry_safe(rd, nd, &f->remotes, list)
777 kfree(rd);
778 kfree(f);
779 }
780
781 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
782 {
783 netdev_dbg(vxlan->dev,
784 "delete %pM\n", f->eth_addr);
785
786 --vxlan->addrcnt;
787 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
788
789 hlist_del_rcu(&f->hlist);
790 call_rcu(&f->rcu, vxlan_fdb_free);
791 }
792
793 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
794 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
795 {
796 struct net *net = dev_net(vxlan->dev);
797 int err;
798
799 if (tb[NDA_DST]) {
800 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
801 if (err)
802 return err;
803 } else {
804 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
805 if (remote->sa.sa_family == AF_INET) {
806 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
807 ip->sa.sa_family = AF_INET;
808 #if IS_ENABLED(CONFIG_IPV6)
809 } else {
810 ip->sin6.sin6_addr = in6addr_any;
811 ip->sa.sa_family = AF_INET6;
812 #endif
813 }
814 }
815
816 if (tb[NDA_PORT]) {
817 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
818 return -EINVAL;
819 *port = nla_get_be16(tb[NDA_PORT]);
820 } else {
821 *port = vxlan->dst_port;
822 }
823
824 if (tb[NDA_VNI]) {
825 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
826 return -EINVAL;
827 *vni = nla_get_u32(tb[NDA_VNI]);
828 } else {
829 *vni = vxlan->default_dst.remote_vni;
830 }
831
832 if (tb[NDA_IFINDEX]) {
833 struct net_device *tdev;
834
835 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
836 return -EINVAL;
837 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
838 tdev = __dev_get_by_index(net, *ifindex);
839 if (!tdev)
840 return -EADDRNOTAVAIL;
841 } else {
842 *ifindex = 0;
843 }
844
845 return 0;
846 }
847
848 /* Add static entry (via netlink) */
849 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
850 struct net_device *dev,
851 const unsigned char *addr, u16 flags)
852 {
853 struct vxlan_dev *vxlan = netdev_priv(dev);
854 /* struct net *net = dev_net(vxlan->dev); */
855 union vxlan_addr ip;
856 __be16 port;
857 u32 vni, ifindex;
858 int err;
859
860 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
861 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
862 ndm->ndm_state);
863 return -EINVAL;
864 }
865
866 if (tb[NDA_DST] == NULL)
867 return -EINVAL;
868
869 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
870 if (err)
871 return err;
872
873 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
874 return -EAFNOSUPPORT;
875
876 spin_lock_bh(&vxlan->hash_lock);
877 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
878 port, vni, ifindex, ndm->ndm_flags);
879 spin_unlock_bh(&vxlan->hash_lock);
880
881 return err;
882 }
883
884 /* Delete entry (via netlink) */
885 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
886 struct net_device *dev,
887 const unsigned char *addr)
888 {
889 struct vxlan_dev *vxlan = netdev_priv(dev);
890 struct vxlan_fdb *f;
891 struct vxlan_rdst *rd = NULL;
892 union vxlan_addr ip;
893 __be16 port;
894 u32 vni, ifindex;
895 int err;
896
897 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
898 if (err)
899 return err;
900
901 err = -ENOENT;
902
903 spin_lock_bh(&vxlan->hash_lock);
904 f = vxlan_find_mac(vxlan, addr);
905 if (!f)
906 goto out;
907
908 if (!vxlan_addr_any(&ip)) {
909 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
910 if (!rd)
911 goto out;
912 }
913
914 err = 0;
915
916 /* remove a destination if it's not the only one on the list,
917 * otherwise destroy the fdb entry
918 */
919 if (rd && !list_is_singular(&f->remotes)) {
920 list_del_rcu(&rd->list);
921 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
922 kfree_rcu(rd, rcu);
923 goto out;
924 }
925
926 vxlan_fdb_destroy(vxlan, f);
927
928 out:
929 spin_unlock_bh(&vxlan->hash_lock);
930
931 return err;
932 }
933
934 /* Dump forwarding table */
935 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
936 struct net_device *dev,
937 struct net_device *filter_dev, int idx)
938 {
939 struct vxlan_dev *vxlan = netdev_priv(dev);
940 unsigned int h;
941
942 for (h = 0; h < FDB_HASH_SIZE; ++h) {
943 struct vxlan_fdb *f;
944 int err;
945
946 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
947 struct vxlan_rdst *rd;
948
949 if (idx < cb->args[0])
950 goto skip;
951
952 list_for_each_entry_rcu(rd, &f->remotes, list) {
953 err = vxlan_fdb_info(skb, vxlan, f,
954 NETLINK_CB(cb->skb).portid,
955 cb->nlh->nlmsg_seq,
956 RTM_NEWNEIGH,
957 NLM_F_MULTI, rd);
958 if (err < 0)
959 goto out;
960 }
961 skip:
962 ++idx;
963 }
964 }
965 out:
966 return idx;
967 }
968
969 /* Watch incoming packets to learn mapping between Ethernet address
970 * and Tunnel endpoint.
971 * Return true if packet is bogus and should be droppped.
972 */
973 static bool vxlan_snoop(struct net_device *dev,
974 union vxlan_addr *src_ip, const u8 *src_mac)
975 {
976 struct vxlan_dev *vxlan = netdev_priv(dev);
977 struct vxlan_fdb *f;
978
979 f = vxlan_find_mac(vxlan, src_mac);
980 if (likely(f)) {
981 struct vxlan_rdst *rdst = first_remote_rcu(f);
982
983 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
984 return false;
985
986 /* Don't migrate static entries, drop packets */
987 if (f->state & NUD_NOARP)
988 return true;
989
990 if (net_ratelimit())
991 netdev_info(dev,
992 "%pM migrated from %pIS to %pIS\n",
993 src_mac, &rdst->remote_ip, &src_ip);
994
995 rdst->remote_ip = *src_ip;
996 f->updated = jiffies;
997 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
998 } else {
999 /* learned new entry */
1000 spin_lock(&vxlan->hash_lock);
1001
1002 /* close off race between vxlan_flush and incoming packets */
1003 if (netif_running(dev))
1004 vxlan_fdb_create(vxlan, src_mac, src_ip,
1005 NUD_REACHABLE,
1006 NLM_F_EXCL|NLM_F_CREATE,
1007 vxlan->dst_port,
1008 vxlan->default_dst.remote_vni,
1009 0, NTF_SELF);
1010 spin_unlock(&vxlan->hash_lock);
1011 }
1012
1013 return false;
1014 }
1015
1016 /* See if multicast group is already in use by other ID */
1017 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1018 {
1019 struct vxlan_dev *vxlan;
1020
1021 /* The vxlan_sock is only used by dev, leaving group has
1022 * no effect on other vxlan devices.
1023 */
1024 if (atomic_read(&dev->vn_sock->refcnt) == 1)
1025 return false;
1026
1027 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1028 if (!netif_running(vxlan->dev) || vxlan == dev)
1029 continue;
1030
1031 if (vxlan->vn_sock != dev->vn_sock)
1032 continue;
1033
1034 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1035 &dev->default_dst.remote_ip))
1036 continue;
1037
1038 if (vxlan->default_dst.remote_ifindex !=
1039 dev->default_dst.remote_ifindex)
1040 continue;
1041
1042 return true;
1043 }
1044
1045 return false;
1046 }
1047
1048 static void vxlan_sock_hold(struct vxlan_sock *vs)
1049 {
1050 atomic_inc(&vs->refcnt);
1051 }
1052
1053 void vxlan_sock_release(struct vxlan_sock *vs)
1054 {
1055 struct sock *sk = vs->sock->sk;
1056 struct net *net = sock_net(sk);
1057 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1058
1059 if (!atomic_dec_and_test(&vs->refcnt))
1060 return;
1061
1062 spin_lock(&vn->sock_lock);
1063 hlist_del_rcu(&vs->hlist);
1064 rcu_assign_sk_user_data(vs->sock->sk, NULL);
1065 vxlan_notify_del_rx_port(vs);
1066 spin_unlock(&vn->sock_lock);
1067
1068 queue_work(vxlan_wq, &vs->del_work);
1069 }
1070 EXPORT_SYMBOL_GPL(vxlan_sock_release);
1071
1072 /* Callback to update multicast group membership when first VNI on
1073 * multicast asddress is brought up
1074 * Done as workqueue because ip_mc_join_group acquires RTNL.
1075 */
1076 static void vxlan_igmp_join(struct work_struct *work)
1077 {
1078 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
1079 struct vxlan_sock *vs = vxlan->vn_sock;
1080 struct sock *sk = vs->sock->sk;
1081 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1082 int ifindex = vxlan->default_dst.remote_ifindex;
1083
1084 lock_sock(sk);
1085 if (ip->sa.sa_family == AF_INET) {
1086 struct ip_mreqn mreq = {
1087 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1088 .imr_ifindex = ifindex,
1089 };
1090
1091 ip_mc_join_group(sk, &mreq);
1092 #if IS_ENABLED(CONFIG_IPV6)
1093 } else {
1094 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1095 &ip->sin6.sin6_addr);
1096 #endif
1097 }
1098 release_sock(sk);
1099
1100 vxlan_sock_release(vs);
1101 dev_put(vxlan->dev);
1102 }
1103
1104 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1105 static void vxlan_igmp_leave(struct work_struct *work)
1106 {
1107 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
1108 struct vxlan_sock *vs = vxlan->vn_sock;
1109 struct sock *sk = vs->sock->sk;
1110 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1111 int ifindex = vxlan->default_dst.remote_ifindex;
1112
1113 lock_sock(sk);
1114 if (ip->sa.sa_family == AF_INET) {
1115 struct ip_mreqn mreq = {
1116 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1117 .imr_ifindex = ifindex,
1118 };
1119
1120 ip_mc_leave_group(sk, &mreq);
1121 #if IS_ENABLED(CONFIG_IPV6)
1122 } else {
1123 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1124 &ip->sin6.sin6_addr);
1125 #endif
1126 }
1127
1128 release_sock(sk);
1129
1130 vxlan_sock_release(vs);
1131 dev_put(vxlan->dev);
1132 }
1133
1134 /* Callback from net/ipv4/udp.c to receive packets */
1135 static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1136 {
1137 struct vxlan_sock *vs;
1138 struct vxlanhdr *vxh;
1139
1140 /* Need Vxlan and inner Ethernet header to be present */
1141 if (!pskb_may_pull(skb, VXLAN_HLEN))
1142 goto error;
1143
1144 /* Return packets with reserved bits set */
1145 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1146 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
1147 (vxh->vx_vni & htonl(0xff))) {
1148 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1149 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1150 goto error;
1151 }
1152
1153 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1154 goto drop;
1155
1156 vs = rcu_dereference_sk_user_data(sk);
1157 if (!vs)
1158 goto drop;
1159
1160 skb_pop_rcv_encapsulation(skb);
1161
1162 vs->rcv(vs, skb, vxh->vx_vni);
1163 return 0;
1164
1165 drop:
1166 /* Consume bad packet */
1167 kfree_skb(skb);
1168 return 0;
1169
1170 error:
1171 /* Return non vxlan pkt */
1172 return 1;
1173 }
1174
1175 static void vxlan_rcv(struct vxlan_sock *vs,
1176 struct sk_buff *skb, __be32 vx_vni)
1177 {
1178 struct iphdr *oip = NULL;
1179 struct ipv6hdr *oip6 = NULL;
1180 struct vxlan_dev *vxlan;
1181 struct pcpu_sw_netstats *stats;
1182 union vxlan_addr saddr;
1183 __u32 vni;
1184 int err = 0;
1185 union vxlan_addr *remote_ip;
1186
1187 vni = ntohl(vx_vni) >> 8;
1188 /* Is this VNI defined? */
1189 vxlan = vxlan_vs_find_vni(vs, vni);
1190 if (!vxlan)
1191 goto drop;
1192
1193 remote_ip = &vxlan->default_dst.remote_ip;
1194 skb_reset_mac_header(skb);
1195 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
1196 skb->protocol = eth_type_trans(skb, vxlan->dev);
1197 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1198
1199 /* Ignore packet loops (and multicast echo) */
1200 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1201 goto drop;
1202
1203 /* Re-examine inner Ethernet packet */
1204 if (remote_ip->sa.sa_family == AF_INET) {
1205 oip = ip_hdr(skb);
1206 saddr.sin.sin_addr.s_addr = oip->saddr;
1207 saddr.sa.sa_family = AF_INET;
1208 #if IS_ENABLED(CONFIG_IPV6)
1209 } else {
1210 oip6 = ipv6_hdr(skb);
1211 saddr.sin6.sin6_addr = oip6->saddr;
1212 saddr.sa.sa_family = AF_INET6;
1213 #endif
1214 }
1215
1216 if ((vxlan->flags & VXLAN_F_LEARN) &&
1217 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1218 goto drop;
1219
1220 skb_reset_network_header(skb);
1221
1222 if (oip6)
1223 err = IP6_ECN_decapsulate(oip6, skb);
1224 if (oip)
1225 err = IP_ECN_decapsulate(oip, skb);
1226
1227 if (unlikely(err)) {
1228 if (log_ecn_error) {
1229 if (oip6)
1230 net_info_ratelimited("non-ECT from %pI6\n",
1231 &oip6->saddr);
1232 if (oip)
1233 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1234 &oip->saddr, oip->tos);
1235 }
1236 if (err > 1) {
1237 ++vxlan->dev->stats.rx_frame_errors;
1238 ++vxlan->dev->stats.rx_errors;
1239 goto drop;
1240 }
1241 }
1242
1243 stats = this_cpu_ptr(vxlan->dev->tstats);
1244 u64_stats_update_begin(&stats->syncp);
1245 stats->rx_packets++;
1246 stats->rx_bytes += skb->len;
1247 u64_stats_update_end(&stats->syncp);
1248
1249 netif_rx(skb);
1250
1251 return;
1252 drop:
1253 /* Consume bad packet */
1254 kfree_skb(skb);
1255 }
1256
1257 static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1258 {
1259 struct vxlan_dev *vxlan = netdev_priv(dev);
1260 struct arphdr *parp;
1261 u8 *arpptr, *sha;
1262 __be32 sip, tip;
1263 struct neighbour *n;
1264
1265 if (dev->flags & IFF_NOARP)
1266 goto out;
1267
1268 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1269 dev->stats.tx_dropped++;
1270 goto out;
1271 }
1272 parp = arp_hdr(skb);
1273
1274 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1275 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1276 parp->ar_pro != htons(ETH_P_IP) ||
1277 parp->ar_op != htons(ARPOP_REQUEST) ||
1278 parp->ar_hln != dev->addr_len ||
1279 parp->ar_pln != 4)
1280 goto out;
1281 arpptr = (u8 *)parp + sizeof(struct arphdr);
1282 sha = arpptr;
1283 arpptr += dev->addr_len; /* sha */
1284 memcpy(&sip, arpptr, sizeof(sip));
1285 arpptr += sizeof(sip);
1286 arpptr += dev->addr_len; /* tha */
1287 memcpy(&tip, arpptr, sizeof(tip));
1288
1289 if (ipv4_is_loopback(tip) ||
1290 ipv4_is_multicast(tip))
1291 goto out;
1292
1293 n = neigh_lookup(&arp_tbl, &tip, dev);
1294
1295 if (n) {
1296 struct vxlan_fdb *f;
1297 struct sk_buff *reply;
1298
1299 if (!(n->nud_state & NUD_CONNECTED)) {
1300 neigh_release(n);
1301 goto out;
1302 }
1303
1304 f = vxlan_find_mac(vxlan, n->ha);
1305 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1306 /* bridge-local neighbor */
1307 neigh_release(n);
1308 goto out;
1309 }
1310
1311 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1312 n->ha, sha);
1313
1314 neigh_release(n);
1315
1316 if (reply == NULL)
1317 goto out;
1318
1319 skb_reset_mac_header(reply);
1320 __skb_pull(reply, skb_network_offset(reply));
1321 reply->ip_summed = CHECKSUM_UNNECESSARY;
1322 reply->pkt_type = PACKET_HOST;
1323
1324 if (netif_rx_ni(reply) == NET_RX_DROP)
1325 dev->stats.rx_dropped++;
1326 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1327 union vxlan_addr ipa = {
1328 .sin.sin_addr.s_addr = tip,
1329 .sa.sa_family = AF_INET,
1330 };
1331
1332 vxlan_ip_miss(dev, &ipa);
1333 }
1334 out:
1335 consume_skb(skb);
1336 return NETDEV_TX_OK;
1337 }
1338
1339 #if IS_ENABLED(CONFIG_IPV6)
1340
1341 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1342 struct neighbour *n, bool isrouter)
1343 {
1344 struct net_device *dev = request->dev;
1345 struct sk_buff *reply;
1346 struct nd_msg *ns, *na;
1347 struct ipv6hdr *pip6;
1348 u8 *daddr;
1349 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1350 int ns_olen;
1351 int i, len;
1352
1353 if (dev == NULL)
1354 return NULL;
1355
1356 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1357 sizeof(*na) + na_olen + dev->needed_tailroom;
1358 reply = alloc_skb(len, GFP_ATOMIC);
1359 if (reply == NULL)
1360 return NULL;
1361
1362 reply->protocol = htons(ETH_P_IPV6);
1363 reply->dev = dev;
1364 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1365 skb_push(reply, sizeof(struct ethhdr));
1366 skb_set_mac_header(reply, 0);
1367
1368 ns = (struct nd_msg *)skb_transport_header(request);
1369
1370 daddr = eth_hdr(request)->h_source;
1371 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1372 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1373 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1374 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1375 break;
1376 }
1377 }
1378
1379 /* Ethernet header */
1380 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1381 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1382 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1383 reply->protocol = htons(ETH_P_IPV6);
1384
1385 skb_pull(reply, sizeof(struct ethhdr));
1386 skb_set_network_header(reply, 0);
1387 skb_put(reply, sizeof(struct ipv6hdr));
1388
1389 /* IPv6 header */
1390
1391 pip6 = ipv6_hdr(reply);
1392 memset(pip6, 0, sizeof(struct ipv6hdr));
1393 pip6->version = 6;
1394 pip6->priority = ipv6_hdr(request)->priority;
1395 pip6->nexthdr = IPPROTO_ICMPV6;
1396 pip6->hop_limit = 255;
1397 pip6->daddr = ipv6_hdr(request)->saddr;
1398 pip6->saddr = *(struct in6_addr *)n->primary_key;
1399
1400 skb_pull(reply, sizeof(struct ipv6hdr));
1401 skb_set_transport_header(reply, 0);
1402
1403 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1404
1405 /* Neighbor Advertisement */
1406 memset(na, 0, sizeof(*na)+na_olen);
1407 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1408 na->icmph.icmp6_router = isrouter;
1409 na->icmph.icmp6_override = 1;
1410 na->icmph.icmp6_solicited = 1;
1411 na->target = ns->target;
1412 ether_addr_copy(&na->opt[2], n->ha);
1413 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1414 na->opt[1] = na_olen >> 3;
1415
1416 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1417 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1418 csum_partial(na, sizeof(*na)+na_olen, 0));
1419
1420 pip6->payload_len = htons(sizeof(*na)+na_olen);
1421
1422 skb_push(reply, sizeof(struct ipv6hdr));
1423
1424 reply->ip_summed = CHECKSUM_UNNECESSARY;
1425
1426 return reply;
1427 }
1428
1429 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1430 {
1431 struct vxlan_dev *vxlan = netdev_priv(dev);
1432 struct nd_msg *msg;
1433 const struct ipv6hdr *iphdr;
1434 const struct in6_addr *saddr, *daddr;
1435 struct neighbour *n;
1436 struct inet6_dev *in6_dev;
1437
1438 in6_dev = __in6_dev_get(dev);
1439 if (!in6_dev)
1440 goto out;
1441
1442 if (!pskb_may_pull(skb, skb->len))
1443 goto out;
1444
1445 iphdr = ipv6_hdr(skb);
1446 saddr = &iphdr->saddr;
1447 daddr = &iphdr->daddr;
1448
1449 msg = (struct nd_msg *)skb_transport_header(skb);
1450 if (msg->icmph.icmp6_code != 0 ||
1451 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1452 goto out;
1453
1454 if (ipv6_addr_loopback(daddr) ||
1455 ipv6_addr_is_multicast(&msg->target))
1456 goto out;
1457
1458 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
1459
1460 if (n) {
1461 struct vxlan_fdb *f;
1462 struct sk_buff *reply;
1463
1464 if (!(n->nud_state & NUD_CONNECTED)) {
1465 neigh_release(n);
1466 goto out;
1467 }
1468
1469 f = vxlan_find_mac(vxlan, n->ha);
1470 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1471 /* bridge-local neighbor */
1472 neigh_release(n);
1473 goto out;
1474 }
1475
1476 reply = vxlan_na_create(skb, n,
1477 !!(f ? f->flags & NTF_ROUTER : 0));
1478
1479 neigh_release(n);
1480
1481 if (reply == NULL)
1482 goto out;
1483
1484 if (netif_rx_ni(reply) == NET_RX_DROP)
1485 dev->stats.rx_dropped++;
1486
1487 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1488 union vxlan_addr ipa = {
1489 .sin6.sin6_addr = msg->target,
1490 .sa.sa_family = AF_INET6,
1491 };
1492
1493 vxlan_ip_miss(dev, &ipa);
1494 }
1495
1496 out:
1497 consume_skb(skb);
1498 return NETDEV_TX_OK;
1499 }
1500 #endif
1501
1502 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1503 {
1504 struct vxlan_dev *vxlan = netdev_priv(dev);
1505 struct neighbour *n;
1506
1507 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1508 return false;
1509
1510 n = NULL;
1511 switch (ntohs(eth_hdr(skb)->h_proto)) {
1512 case ETH_P_IP:
1513 {
1514 struct iphdr *pip;
1515
1516 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1517 return false;
1518 pip = ip_hdr(skb);
1519 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1520 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1521 union vxlan_addr ipa = {
1522 .sin.sin_addr.s_addr = pip->daddr,
1523 .sa.sa_family = AF_INET,
1524 };
1525
1526 vxlan_ip_miss(dev, &ipa);
1527 return false;
1528 }
1529
1530 break;
1531 }
1532 #if IS_ENABLED(CONFIG_IPV6)
1533 case ETH_P_IPV6:
1534 {
1535 struct ipv6hdr *pip6;
1536
1537 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1538 return false;
1539 pip6 = ipv6_hdr(skb);
1540 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1541 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1542 union vxlan_addr ipa = {
1543 .sin6.sin6_addr = pip6->daddr,
1544 .sa.sa_family = AF_INET6,
1545 };
1546
1547 vxlan_ip_miss(dev, &ipa);
1548 return false;
1549 }
1550
1551 break;
1552 }
1553 #endif
1554 default:
1555 return false;
1556 }
1557
1558 if (n) {
1559 bool diff;
1560
1561 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1562 if (diff) {
1563 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1564 dev->addr_len);
1565 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1566 }
1567 neigh_release(n);
1568 return diff;
1569 }
1570
1571 return false;
1572 }
1573
1574 static inline struct sk_buff *vxlan_handle_offloads(struct sk_buff *skb,
1575 bool udp_csum)
1576 {
1577 int type = udp_csum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1578 return iptunnel_handle_offloads(skb, udp_csum, type);
1579 }
1580
1581 #if IS_ENABLED(CONFIG_IPV6)
1582 static int vxlan6_xmit_skb(struct vxlan_sock *vs,
1583 struct dst_entry *dst, struct sk_buff *skb,
1584 struct net_device *dev, struct in6_addr *saddr,
1585 struct in6_addr *daddr, __u8 prio, __u8 ttl,
1586 __be16 src_port, __be16 dst_port, __be32 vni,
1587 bool xnet)
1588 {
1589 struct ipv6hdr *ip6h;
1590 struct vxlanhdr *vxh;
1591 struct udphdr *uh;
1592 int min_headroom;
1593 int err;
1594
1595 skb = vxlan_handle_offloads(skb, !udp_get_no_check6_tx(vs->sock->sk));
1596 if (IS_ERR(skb))
1597 return -EINVAL;
1598
1599 skb_scrub_packet(skb, xnet);
1600
1601 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1602 + VXLAN_HLEN + sizeof(struct ipv6hdr)
1603 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1604
1605 /* Need space for new headers (invalidates iph ptr) */
1606 err = skb_cow_head(skb, min_headroom);
1607 if (unlikely(err))
1608 return err;
1609
1610 if (vlan_tx_tag_present(skb)) {
1611 if (WARN_ON(!__vlan_put_tag(skb,
1612 skb->vlan_proto,
1613 vlan_tx_tag_get(skb))))
1614 return -ENOMEM;
1615
1616 skb->vlan_tci = 0;
1617 }
1618
1619 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1620 vxh->vx_flags = htonl(VXLAN_FLAGS);
1621 vxh->vx_vni = vni;
1622
1623 __skb_push(skb, sizeof(*uh));
1624 skb_reset_transport_header(skb);
1625 uh = udp_hdr(skb);
1626
1627 uh->dest = dst_port;
1628 uh->source = src_port;
1629
1630 uh->len = htons(skb->len);
1631
1632 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1633 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1634 IPSKB_REROUTED);
1635 skb_dst_set(skb, dst);
1636
1637 udp6_set_csum(udp_get_no_check6_tx(vs->sock->sk), skb,
1638 saddr, daddr, skb->len);
1639
1640 __skb_push(skb, sizeof(*ip6h));
1641 skb_reset_network_header(skb);
1642 ip6h = ipv6_hdr(skb);
1643 ip6h->version = 6;
1644 ip6h->priority = prio;
1645 ip6h->flow_lbl[0] = 0;
1646 ip6h->flow_lbl[1] = 0;
1647 ip6h->flow_lbl[2] = 0;
1648 ip6h->payload_len = htons(skb->len);
1649 ip6h->nexthdr = IPPROTO_UDP;
1650 ip6h->hop_limit = ttl;
1651 ip6h->daddr = *daddr;
1652 ip6h->saddr = *saddr;
1653
1654 ip6tunnel_xmit(skb, dev);
1655 return 0;
1656 }
1657 #endif
1658
1659 int vxlan_xmit_skb(struct vxlan_sock *vs,
1660 struct rtable *rt, struct sk_buff *skb,
1661 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
1662 __be16 src_port, __be16 dst_port, __be32 vni, bool xnet)
1663 {
1664 struct vxlanhdr *vxh;
1665 struct udphdr *uh;
1666 int min_headroom;
1667 int err;
1668
1669 skb = vxlan_handle_offloads(skb, !vs->sock->sk->sk_no_check_tx);
1670 if (IS_ERR(skb))
1671 return -EINVAL;
1672
1673 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1674 + VXLAN_HLEN + sizeof(struct iphdr)
1675 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1676
1677 /* Need space for new headers (invalidates iph ptr) */
1678 err = skb_cow_head(skb, min_headroom);
1679 if (unlikely(err))
1680 return err;
1681
1682 if (vlan_tx_tag_present(skb)) {
1683 if (WARN_ON(!__vlan_put_tag(skb,
1684 skb->vlan_proto,
1685 vlan_tx_tag_get(skb))))
1686 return -ENOMEM;
1687
1688 skb->vlan_tci = 0;
1689 }
1690
1691 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1692 vxh->vx_flags = htonl(VXLAN_FLAGS);
1693 vxh->vx_vni = vni;
1694
1695 __skb_push(skb, sizeof(*uh));
1696 skb_reset_transport_header(skb);
1697 uh = udp_hdr(skb);
1698
1699 uh->dest = dst_port;
1700 uh->source = src_port;
1701
1702 uh->len = htons(skb->len);
1703
1704 udp_set_csum(vs->sock->sk->sk_no_check_tx, skb,
1705 src, dst, skb->len);
1706
1707 return iptunnel_xmit(vs->sock->sk, rt, skb, src, dst, IPPROTO_UDP,
1708 tos, ttl, df, xnet);
1709 }
1710 EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1711
1712 /* Bypass encapsulation if the destination is local */
1713 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1714 struct vxlan_dev *dst_vxlan)
1715 {
1716 struct pcpu_sw_netstats *tx_stats, *rx_stats;
1717 union vxlan_addr loopback;
1718 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
1719
1720 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1721 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1722 skb->pkt_type = PACKET_HOST;
1723 skb->encapsulation = 0;
1724 skb->dev = dst_vxlan->dev;
1725 __skb_pull(skb, skb_network_offset(skb));
1726
1727 if (remote_ip->sa.sa_family == AF_INET) {
1728 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1729 loopback.sa.sa_family = AF_INET;
1730 #if IS_ENABLED(CONFIG_IPV6)
1731 } else {
1732 loopback.sin6.sin6_addr = in6addr_loopback;
1733 loopback.sa.sa_family = AF_INET6;
1734 #endif
1735 }
1736
1737 if (dst_vxlan->flags & VXLAN_F_LEARN)
1738 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
1739
1740 u64_stats_update_begin(&tx_stats->syncp);
1741 tx_stats->tx_packets++;
1742 tx_stats->tx_bytes += skb->len;
1743 u64_stats_update_end(&tx_stats->syncp);
1744
1745 if (netif_rx(skb) == NET_RX_SUCCESS) {
1746 u64_stats_update_begin(&rx_stats->syncp);
1747 rx_stats->rx_packets++;
1748 rx_stats->rx_bytes += skb->len;
1749 u64_stats_update_end(&rx_stats->syncp);
1750 } else {
1751 skb->dev->stats.rx_dropped++;
1752 }
1753 }
1754
1755 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1756 struct vxlan_rdst *rdst, bool did_rsc)
1757 {
1758 struct vxlan_dev *vxlan = netdev_priv(dev);
1759 struct rtable *rt = NULL;
1760 const struct iphdr *old_iph;
1761 struct flowi4 fl4;
1762 union vxlan_addr *dst;
1763 __be16 src_port = 0, dst_port;
1764 u32 vni;
1765 __be16 df = 0;
1766 __u8 tos, ttl;
1767 int err;
1768
1769 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
1770 vni = rdst->remote_vni;
1771 dst = &rdst->remote_ip;
1772
1773 if (vxlan_addr_any(dst)) {
1774 if (did_rsc) {
1775 /* short-circuited back to local bridge */
1776 vxlan_encap_bypass(skb, vxlan, vxlan);
1777 return;
1778 }
1779 goto drop;
1780 }
1781
1782 old_iph = ip_hdr(skb);
1783
1784 ttl = vxlan->ttl;
1785 if (!ttl && vxlan_addr_multicast(dst))
1786 ttl = 1;
1787
1788 tos = vxlan->tos;
1789 if (tos == 1)
1790 tos = ip_tunnel_get_dsfield(old_iph, skb);
1791
1792 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->port_min,
1793 vxlan->port_max, true);
1794
1795 if (dst->sa.sa_family == AF_INET) {
1796 memset(&fl4, 0, sizeof(fl4));
1797 fl4.flowi4_oif = rdst->remote_ifindex;
1798 fl4.flowi4_tos = RT_TOS(tos);
1799 fl4.daddr = dst->sin.sin_addr.s_addr;
1800 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1801
1802 rt = ip_route_output_key(vxlan->net, &fl4);
1803 if (IS_ERR(rt)) {
1804 netdev_dbg(dev, "no route to %pI4\n",
1805 &dst->sin.sin_addr.s_addr);
1806 dev->stats.tx_carrier_errors++;
1807 goto tx_error;
1808 }
1809
1810 if (rt->dst.dev == dev) {
1811 netdev_dbg(dev, "circular route to %pI4\n",
1812 &dst->sin.sin_addr.s_addr);
1813 dev->stats.collisions++;
1814 goto rt_tx_error;
1815 }
1816
1817 /* Bypass encapsulation if the destination is local */
1818 if (rt->rt_flags & RTCF_LOCAL &&
1819 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1820 struct vxlan_dev *dst_vxlan;
1821
1822 ip_rt_put(rt);
1823 dst_vxlan = vxlan_find_vni(vxlan->net, vni, dst_port);
1824 if (!dst_vxlan)
1825 goto tx_error;
1826 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1827 return;
1828 }
1829
1830 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1831 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
1832
1833 err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
1834 fl4.saddr, dst->sin.sin_addr.s_addr,
1835 tos, ttl, df, src_port, dst_port,
1836 htonl(vni << 8),
1837 !net_eq(vxlan->net, dev_net(vxlan->dev)));
1838
1839 if (err < 0)
1840 goto rt_tx_error;
1841 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1842 #if IS_ENABLED(CONFIG_IPV6)
1843 } else {
1844 struct sock *sk = vxlan->vn_sock->sock->sk;
1845 struct dst_entry *ndst;
1846 struct flowi6 fl6;
1847 u32 flags;
1848
1849 memset(&fl6, 0, sizeof(fl6));
1850 fl6.flowi6_oif = rdst->remote_ifindex;
1851 fl6.daddr = dst->sin6.sin6_addr;
1852 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
1853 fl6.flowi6_proto = IPPROTO_UDP;
1854
1855 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1856 netdev_dbg(dev, "no route to %pI6\n",
1857 &dst->sin6.sin6_addr);
1858 dev->stats.tx_carrier_errors++;
1859 goto tx_error;
1860 }
1861
1862 if (ndst->dev == dev) {
1863 netdev_dbg(dev, "circular route to %pI6\n",
1864 &dst->sin6.sin6_addr);
1865 dst_release(ndst);
1866 dev->stats.collisions++;
1867 goto tx_error;
1868 }
1869
1870 /* Bypass encapsulation if the destination is local */
1871 flags = ((struct rt6_info *)ndst)->rt6i_flags;
1872 if (flags & RTF_LOCAL &&
1873 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1874 struct vxlan_dev *dst_vxlan;
1875
1876 dst_release(ndst);
1877 dst_vxlan = vxlan_find_vni(vxlan->net, vni, dst_port);
1878 if (!dst_vxlan)
1879 goto tx_error;
1880 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1881 return;
1882 }
1883
1884 ttl = ttl ? : ip6_dst_hoplimit(ndst);
1885
1886 err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
1887 dev, &fl6.saddr, &fl6.daddr, 0, ttl,
1888 src_port, dst_port, htonl(vni << 8),
1889 !net_eq(vxlan->net, dev_net(vxlan->dev)));
1890 #endif
1891 }
1892
1893 return;
1894
1895 drop:
1896 dev->stats.tx_dropped++;
1897 goto tx_free;
1898
1899 rt_tx_error:
1900 ip_rt_put(rt);
1901 tx_error:
1902 dev->stats.tx_errors++;
1903 tx_free:
1904 dev_kfree_skb(skb);
1905 }
1906
1907 /* Transmit local packets over Vxlan
1908 *
1909 * Outer IP header inherits ECN and DF from inner header.
1910 * Outer UDP destination is the VXLAN assigned port.
1911 * source port is based on hash of flow
1912 */
1913 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1914 {
1915 struct vxlan_dev *vxlan = netdev_priv(dev);
1916 struct ethhdr *eth;
1917 bool did_rsc = false;
1918 struct vxlan_rdst *rdst, *fdst = NULL;
1919 struct vxlan_fdb *f;
1920
1921 skb_reset_mac_header(skb);
1922 eth = eth_hdr(skb);
1923
1924 if ((vxlan->flags & VXLAN_F_PROXY)) {
1925 if (ntohs(eth->h_proto) == ETH_P_ARP)
1926 return arp_reduce(dev, skb);
1927 #if IS_ENABLED(CONFIG_IPV6)
1928 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
1929 skb->len >= sizeof(struct ipv6hdr) + sizeof(struct nd_msg) &&
1930 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
1931 struct nd_msg *msg;
1932
1933 msg = (struct nd_msg *)skb_transport_header(skb);
1934 if (msg->icmph.icmp6_code == 0 &&
1935 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
1936 return neigh_reduce(dev, skb);
1937 }
1938 #endif
1939 }
1940
1941 f = vxlan_find_mac(vxlan, eth->h_dest);
1942 did_rsc = false;
1943
1944 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
1945 (ntohs(eth->h_proto) == ETH_P_IP ||
1946 ntohs(eth->h_proto) == ETH_P_IPV6)) {
1947 did_rsc = route_shortcircuit(dev, skb);
1948 if (did_rsc)
1949 f = vxlan_find_mac(vxlan, eth->h_dest);
1950 }
1951
1952 if (f == NULL) {
1953 f = vxlan_find_mac(vxlan, all_zeros_mac);
1954 if (f == NULL) {
1955 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1956 !is_multicast_ether_addr(eth->h_dest))
1957 vxlan_fdb_miss(vxlan, eth->h_dest);
1958
1959 dev->stats.tx_dropped++;
1960 kfree_skb(skb);
1961 return NETDEV_TX_OK;
1962 }
1963 }
1964
1965 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1966 struct sk_buff *skb1;
1967
1968 if (!fdst) {
1969 fdst = rdst;
1970 continue;
1971 }
1972 skb1 = skb_clone(skb, GFP_ATOMIC);
1973 if (skb1)
1974 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
1975 }
1976
1977 if (fdst)
1978 vxlan_xmit_one(skb, dev, fdst, did_rsc);
1979 else
1980 kfree_skb(skb);
1981 return NETDEV_TX_OK;
1982 }
1983
1984 /* Walk the forwarding table and purge stale entries */
1985 static void vxlan_cleanup(unsigned long arg)
1986 {
1987 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1988 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1989 unsigned int h;
1990
1991 if (!netif_running(vxlan->dev))
1992 return;
1993
1994 spin_lock_bh(&vxlan->hash_lock);
1995 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1996 struct hlist_node *p, *n;
1997 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1998 struct vxlan_fdb *f
1999 = container_of(p, struct vxlan_fdb, hlist);
2000 unsigned long timeout;
2001
2002 if (f->state & NUD_PERMANENT)
2003 continue;
2004
2005 timeout = f->used + vxlan->age_interval * HZ;
2006 if (time_before_eq(timeout, jiffies)) {
2007 netdev_dbg(vxlan->dev,
2008 "garbage collect %pM\n",
2009 f->eth_addr);
2010 f->state = NUD_STALE;
2011 vxlan_fdb_destroy(vxlan, f);
2012 } else if (time_before(timeout, next_timer))
2013 next_timer = timeout;
2014 }
2015 }
2016 spin_unlock_bh(&vxlan->hash_lock);
2017
2018 mod_timer(&vxlan->age_timer, next_timer);
2019 }
2020
2021 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2022 {
2023 __u32 vni = vxlan->default_dst.remote_vni;
2024
2025 vxlan->vn_sock = vs;
2026 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
2027 }
2028
2029 /* Setup stats when device is created */
2030 static int vxlan_init(struct net_device *dev)
2031 {
2032 struct vxlan_dev *vxlan = netdev_priv(dev);
2033 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2034 struct vxlan_sock *vs;
2035
2036 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2037 if (!dev->tstats)
2038 return -ENOMEM;
2039
2040 spin_lock(&vn->sock_lock);
2041 vs = vxlan_find_sock(vxlan->net, vxlan->dst_port);
2042 if (vs) {
2043 /* If we have a socket with same port already, reuse it */
2044 atomic_inc(&vs->refcnt);
2045 vxlan_vs_add_dev(vs, vxlan);
2046 } else {
2047 /* otherwise make new socket outside of RTNL */
2048 dev_hold(dev);
2049 queue_work(vxlan_wq, &vxlan->sock_work);
2050 }
2051 spin_unlock(&vn->sock_lock);
2052
2053 return 0;
2054 }
2055
2056 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
2057 {
2058 struct vxlan_fdb *f;
2059
2060 spin_lock_bh(&vxlan->hash_lock);
2061 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2062 if (f)
2063 vxlan_fdb_destroy(vxlan, f);
2064 spin_unlock_bh(&vxlan->hash_lock);
2065 }
2066
2067 static void vxlan_uninit(struct net_device *dev)
2068 {
2069 struct vxlan_dev *vxlan = netdev_priv(dev);
2070 struct vxlan_sock *vs = vxlan->vn_sock;
2071
2072 vxlan_fdb_delete_default(vxlan);
2073
2074 if (vs)
2075 vxlan_sock_release(vs);
2076 free_percpu(dev->tstats);
2077 }
2078
2079 /* Start ageing timer and join group when device is brought up */
2080 static int vxlan_open(struct net_device *dev)
2081 {
2082 struct vxlan_dev *vxlan = netdev_priv(dev);
2083 struct vxlan_sock *vs = vxlan->vn_sock;
2084
2085 /* socket hasn't been created */
2086 if (!vs)
2087 return -ENOTCONN;
2088
2089 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
2090 vxlan_sock_hold(vs);
2091 dev_hold(dev);
2092 queue_work(vxlan_wq, &vxlan->igmp_join);
2093 }
2094
2095 if (vxlan->age_interval)
2096 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2097
2098 return 0;
2099 }
2100
2101 /* Purge the forwarding table */
2102 static void vxlan_flush(struct vxlan_dev *vxlan)
2103 {
2104 unsigned int h;
2105
2106 spin_lock_bh(&vxlan->hash_lock);
2107 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2108 struct hlist_node *p, *n;
2109 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2110 struct vxlan_fdb *f
2111 = container_of(p, struct vxlan_fdb, hlist);
2112 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2113 if (!is_zero_ether_addr(f->eth_addr))
2114 vxlan_fdb_destroy(vxlan, f);
2115 }
2116 }
2117 spin_unlock_bh(&vxlan->hash_lock);
2118 }
2119
2120 /* Cleanup timer and forwarding table on shutdown */
2121 static int vxlan_stop(struct net_device *dev)
2122 {
2123 struct vxlan_dev *vxlan = netdev_priv(dev);
2124 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2125 struct vxlan_sock *vs = vxlan->vn_sock;
2126
2127 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2128 !vxlan_group_used(vn, vxlan)) {
2129 vxlan_sock_hold(vs);
2130 dev_hold(dev);
2131 queue_work(vxlan_wq, &vxlan->igmp_leave);
2132 }
2133
2134 del_timer_sync(&vxlan->age_timer);
2135
2136 vxlan_flush(vxlan);
2137
2138 return 0;
2139 }
2140
2141 /* Stub, nothing needs to be done. */
2142 static void vxlan_set_multicast_list(struct net_device *dev)
2143 {
2144 }
2145
2146 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2147 {
2148 struct vxlan_dev *vxlan = netdev_priv(dev);
2149 struct vxlan_rdst *dst = &vxlan->default_dst;
2150 struct net_device *lowerdev;
2151 int max_mtu;
2152
2153 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
2154 if (lowerdev == NULL)
2155 return eth_change_mtu(dev, new_mtu);
2156
2157 if (dst->remote_ip.sa.sa_family == AF_INET6)
2158 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2159 else
2160 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2161
2162 if (new_mtu < 68 || new_mtu > max_mtu)
2163 return -EINVAL;
2164
2165 dev->mtu = new_mtu;
2166 return 0;
2167 }
2168
2169 static const struct net_device_ops vxlan_netdev_ops = {
2170 .ndo_init = vxlan_init,
2171 .ndo_uninit = vxlan_uninit,
2172 .ndo_open = vxlan_open,
2173 .ndo_stop = vxlan_stop,
2174 .ndo_start_xmit = vxlan_xmit,
2175 .ndo_get_stats64 = ip_tunnel_get_stats64,
2176 .ndo_set_rx_mode = vxlan_set_multicast_list,
2177 .ndo_change_mtu = vxlan_change_mtu,
2178 .ndo_validate_addr = eth_validate_addr,
2179 .ndo_set_mac_address = eth_mac_addr,
2180 .ndo_fdb_add = vxlan_fdb_add,
2181 .ndo_fdb_del = vxlan_fdb_delete,
2182 .ndo_fdb_dump = vxlan_fdb_dump,
2183 };
2184
2185 /* Info for udev, that this is a virtual tunnel endpoint */
2186 static struct device_type vxlan_type = {
2187 .name = "vxlan",
2188 };
2189
2190 /* Calls the ndo_add_vxlan_port of the caller in order to
2191 * supply the listening VXLAN udp ports. Callers are expected
2192 * to implement the ndo_add_vxlan_port.
2193 */
2194 void vxlan_get_rx_port(struct net_device *dev)
2195 {
2196 struct vxlan_sock *vs;
2197 struct net *net = dev_net(dev);
2198 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2199 sa_family_t sa_family;
2200 __be16 port;
2201 unsigned int i;
2202
2203 spin_lock(&vn->sock_lock);
2204 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2205 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2206 port = inet_sk(vs->sock->sk)->inet_sport;
2207 sa_family = vs->sock->sk->sk_family;
2208 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2209 port);
2210 }
2211 }
2212 spin_unlock(&vn->sock_lock);
2213 }
2214 EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2215
2216 /* Initialize the device structure. */
2217 static void vxlan_setup(struct net_device *dev)
2218 {
2219 struct vxlan_dev *vxlan = netdev_priv(dev);
2220 unsigned int h;
2221
2222 eth_hw_addr_random(dev);
2223 ether_setup(dev);
2224 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2225 dev->needed_headroom = ETH_HLEN + VXLAN6_HEADROOM;
2226 else
2227 dev->needed_headroom = ETH_HLEN + VXLAN_HEADROOM;
2228
2229 dev->netdev_ops = &vxlan_netdev_ops;
2230 dev->destructor = free_netdev;
2231 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2232
2233 dev->tx_queue_len = 0;
2234 dev->features |= NETIF_F_LLTX;
2235 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
2236 dev->features |= NETIF_F_RXCSUM;
2237 dev->features |= NETIF_F_GSO_SOFTWARE;
2238
2239 dev->vlan_features = dev->features;
2240 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2241 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2242 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2243 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2244 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
2245 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2246
2247 INIT_LIST_HEAD(&vxlan->next);
2248 spin_lock_init(&vxlan->hash_lock);
2249 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2250 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
2251 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
2252
2253 init_timer_deferrable(&vxlan->age_timer);
2254 vxlan->age_timer.function = vxlan_cleanup;
2255 vxlan->age_timer.data = (unsigned long) vxlan;
2256
2257 vxlan->dst_port = htons(vxlan_port);
2258
2259 vxlan->dev = dev;
2260
2261 for (h = 0; h < FDB_HASH_SIZE; ++h)
2262 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2263 }
2264
2265 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2266 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
2267 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2268 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
2269 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2270 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2271 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
2272 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2273 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2274 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2275 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2276 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
2277 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
2278 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2279 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2280 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2281 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
2282 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
2283 };
2284
2285 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2286 {
2287 if (tb[IFLA_ADDRESS]) {
2288 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2289 pr_debug("invalid link address (not ethernet)\n");
2290 return -EINVAL;
2291 }
2292
2293 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2294 pr_debug("invalid all zero ethernet address\n");
2295 return -EADDRNOTAVAIL;
2296 }
2297 }
2298
2299 if (!data)
2300 return -EINVAL;
2301
2302 if (data[IFLA_VXLAN_ID]) {
2303 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2304 if (id >= VXLAN_VID_MASK)
2305 return -ERANGE;
2306 }
2307
2308 if (data[IFLA_VXLAN_PORT_RANGE]) {
2309 const struct ifla_vxlan_port_range *p
2310 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2311
2312 if (ntohs(p->high) < ntohs(p->low)) {
2313 pr_debug("port range %u .. %u not valid\n",
2314 ntohs(p->low), ntohs(p->high));
2315 return -EINVAL;
2316 }
2317 }
2318
2319 return 0;
2320 }
2321
2322 static void vxlan_get_drvinfo(struct net_device *netdev,
2323 struct ethtool_drvinfo *drvinfo)
2324 {
2325 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2326 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2327 }
2328
2329 static const struct ethtool_ops vxlan_ethtool_ops = {
2330 .get_drvinfo = vxlan_get_drvinfo,
2331 .get_link = ethtool_op_get_link,
2332 };
2333
2334 static void vxlan_del_work(struct work_struct *work)
2335 {
2336 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2337
2338 sk_release_kernel(vs->sock->sk);
2339 kfree_rcu(vs, rcu);
2340 }
2341
2342 #if IS_ENABLED(CONFIG_IPV6)
2343 /* Create UDP socket for encapsulation receive. AF_INET6 socket
2344 * could be used for both IPv4 and IPv6 communications, but
2345 * users may set bindv6only=1.
2346 */
2347 static struct socket *create_v6_sock(struct net *net, __be16 port, u32 flags)
2348 {
2349 struct sock *sk;
2350 struct socket *sock;
2351 struct sockaddr_in6 vxlan_addr = {
2352 .sin6_family = AF_INET6,
2353 .sin6_port = port,
2354 };
2355 int rc, val = 1;
2356
2357 rc = sock_create_kern(AF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
2358 if (rc < 0) {
2359 pr_debug("UDPv6 socket create failed\n");
2360 return ERR_PTR(rc);
2361 }
2362
2363 /* Put in proper namespace */
2364 sk = sock->sk;
2365 sk_change_net(sk, net);
2366
2367 kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
2368 (char *)&val, sizeof(val));
2369 rc = kernel_bind(sock, (struct sockaddr *)&vxlan_addr,
2370 sizeof(struct sockaddr_in6));
2371 if (rc < 0) {
2372 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2373 &vxlan_addr.sin6_addr, ntohs(vxlan_addr.sin6_port), rc);
2374 sk_release_kernel(sk);
2375 return ERR_PTR(rc);
2376 }
2377 /* At this point, IPv6 module should have been loaded in
2378 * sock_create_kern().
2379 */
2380 BUG_ON(!ipv6_stub);
2381
2382 /* Disable multicast loopback */
2383 inet_sk(sk)->mc_loop = 0;
2384
2385 if (flags & VXLAN_F_UDP_ZERO_CSUM6_TX)
2386 udp_set_no_check6_tx(sk, true);
2387
2388 if (flags & VXLAN_F_UDP_ZERO_CSUM6_RX)
2389 udp_set_no_check6_rx(sk, true);
2390
2391 return sock;
2392 }
2393
2394 #else
2395
2396 static struct socket *create_v6_sock(struct net *net, __be16 port, u32 flags)
2397 {
2398 return ERR_PTR(-EPFNOSUPPORT);
2399 }
2400 #endif
2401
2402 static struct socket *create_v4_sock(struct net *net, __be16 port, u32 flags)
2403 {
2404 struct sock *sk;
2405 struct socket *sock;
2406 struct sockaddr_in vxlan_addr = {
2407 .sin_family = AF_INET,
2408 .sin_addr.s_addr = htonl(INADDR_ANY),
2409 .sin_port = port,
2410 };
2411 int rc;
2412
2413 /* Create UDP socket for encapsulation receive. */
2414 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
2415 if (rc < 0) {
2416 pr_debug("UDP socket create failed\n");
2417 return ERR_PTR(rc);
2418 }
2419
2420 /* Put in proper namespace */
2421 sk = sock->sk;
2422 sk_change_net(sk, net);
2423
2424 rc = kernel_bind(sock, (struct sockaddr *) &vxlan_addr,
2425 sizeof(vxlan_addr));
2426 if (rc < 0) {
2427 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2428 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
2429 sk_release_kernel(sk);
2430 return ERR_PTR(rc);
2431 }
2432
2433 /* Disable multicast loopback */
2434 inet_sk(sk)->mc_loop = 0;
2435
2436 if (!(flags & VXLAN_F_UDP_CSUM))
2437 sock->sk->sk_no_check_tx = 1;
2438
2439 return sock;
2440 }
2441
2442 /* Create new listen socket if needed */
2443 static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
2444 vxlan_rcv_t *rcv, void *data,
2445 u32 flags)
2446 {
2447 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2448 struct vxlan_sock *vs;
2449 struct socket *sock;
2450 struct sock *sk;
2451 unsigned int h;
2452 bool ipv6 = !!(flags & VXLAN_F_IPV6);
2453
2454 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
2455 if (!vs)
2456 return ERR_PTR(-ENOMEM);
2457
2458 for (h = 0; h < VNI_HASH_SIZE; ++h)
2459 INIT_HLIST_HEAD(&vs->vni_list[h]);
2460
2461 INIT_WORK(&vs->del_work, vxlan_del_work);
2462
2463 if (ipv6)
2464 sock = create_v6_sock(net, port, flags);
2465 else
2466 sock = create_v4_sock(net, port, flags);
2467 if (IS_ERR(sock)) {
2468 kfree(vs);
2469 return ERR_CAST(sock);
2470 }
2471
2472 vs->sock = sock;
2473 sk = sock->sk;
2474 atomic_set(&vs->refcnt, 1);
2475 vs->rcv = rcv;
2476 vs->data = data;
2477 rcu_assign_sk_user_data(vs->sock->sk, vs);
2478
2479 /* Initialize the vxlan udp offloads structure */
2480 vs->udp_offloads.port = port;
2481 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2482 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2483
2484 spin_lock(&vn->sock_lock);
2485 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2486 vxlan_notify_add_rx_port(vs);
2487 spin_unlock(&vn->sock_lock);
2488
2489 /* Mark socket as an encapsulation socket. */
2490 udp_sk(sk)->encap_type = 1;
2491 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
2492 #if IS_ENABLED(CONFIG_IPV6)
2493 if (ipv6)
2494 ipv6_stub->udpv6_encap_enable();
2495 else
2496 #endif
2497 udp_encap_enable();
2498
2499 return vs;
2500 }
2501
2502 struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2503 vxlan_rcv_t *rcv, void *data,
2504 bool no_share, u32 flags)
2505 {
2506 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2507 struct vxlan_sock *vs;
2508
2509 vs = vxlan_socket_create(net, port, rcv, data, flags);
2510 if (!IS_ERR(vs))
2511 return vs;
2512
2513 if (no_share) /* Return error if sharing is not allowed. */
2514 return vs;
2515
2516 spin_lock(&vn->sock_lock);
2517 vs = vxlan_find_sock(net, port);
2518 if (vs) {
2519 if (vs->rcv == rcv)
2520 atomic_inc(&vs->refcnt);
2521 else
2522 vs = ERR_PTR(-EBUSY);
2523 }
2524 spin_unlock(&vn->sock_lock);
2525
2526 if (!vs)
2527 vs = ERR_PTR(-EINVAL);
2528
2529 return vs;
2530 }
2531 EXPORT_SYMBOL_GPL(vxlan_sock_add);
2532
2533 /* Scheduled at device creation to bind to a socket */
2534 static void vxlan_sock_work(struct work_struct *work)
2535 {
2536 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
2537 struct net *net = vxlan->net;
2538 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2539 __be16 port = vxlan->dst_port;
2540 struct vxlan_sock *nvs;
2541
2542 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags);
2543 spin_lock(&vn->sock_lock);
2544 if (!IS_ERR(nvs))
2545 vxlan_vs_add_dev(nvs, vxlan);
2546 spin_unlock(&vn->sock_lock);
2547
2548 dev_put(vxlan->dev);
2549 }
2550
2551 static int vxlan_newlink(struct net *net, struct net_device *dev,
2552 struct nlattr *tb[], struct nlattr *data[])
2553 {
2554 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2555 struct vxlan_dev *vxlan = netdev_priv(dev);
2556 struct vxlan_rdst *dst = &vxlan->default_dst;
2557 __u32 vni;
2558 int err;
2559 bool use_ipv6 = false;
2560
2561 if (!data[IFLA_VXLAN_ID])
2562 return -EINVAL;
2563
2564 vxlan->net = dev_net(dev);
2565
2566 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
2567 dst->remote_vni = vni;
2568
2569 /* Unless IPv6 is explicitly requested, assume IPv4 */
2570 dst->remote_ip.sa.sa_family = AF_INET;
2571 if (data[IFLA_VXLAN_GROUP]) {
2572 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
2573 } else if (data[IFLA_VXLAN_GROUP6]) {
2574 if (!IS_ENABLED(CONFIG_IPV6))
2575 return -EPFNOSUPPORT;
2576
2577 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2578 sizeof(struct in6_addr));
2579 dst->remote_ip.sa.sa_family = AF_INET6;
2580 use_ipv6 = true;
2581 }
2582
2583 if (data[IFLA_VXLAN_LOCAL]) {
2584 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2585 vxlan->saddr.sa.sa_family = AF_INET;
2586 } else if (data[IFLA_VXLAN_LOCAL6]) {
2587 if (!IS_ENABLED(CONFIG_IPV6))
2588 return -EPFNOSUPPORT;
2589
2590 /* TODO: respect scope id */
2591 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2592 sizeof(struct in6_addr));
2593 vxlan->saddr.sa.sa_family = AF_INET6;
2594 use_ipv6 = true;
2595 }
2596
2597 if (data[IFLA_VXLAN_LINK] &&
2598 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
2599 struct net_device *lowerdev
2600 = __dev_get_by_index(net, dst->remote_ifindex);
2601
2602 if (!lowerdev) {
2603 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
2604 return -ENODEV;
2605 }
2606
2607 #if IS_ENABLED(CONFIG_IPV6)
2608 if (use_ipv6) {
2609 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2610 if (idev && idev->cnf.disable_ipv6) {
2611 pr_info("IPv6 is disabled via sysctl\n");
2612 return -EPERM;
2613 }
2614 vxlan->flags |= VXLAN_F_IPV6;
2615 }
2616 #endif
2617
2618 if (!tb[IFLA_MTU])
2619 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2620
2621 dev->needed_headroom = lowerdev->hard_header_len +
2622 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2623 } else if (use_ipv6)
2624 vxlan->flags |= VXLAN_F_IPV6;
2625
2626 if (data[IFLA_VXLAN_TOS])
2627 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2628
2629 if (data[IFLA_VXLAN_TTL])
2630 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2631
2632 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
2633 vxlan->flags |= VXLAN_F_LEARN;
2634
2635 if (data[IFLA_VXLAN_AGEING])
2636 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2637 else
2638 vxlan->age_interval = FDB_AGE_DEFAULT;
2639
2640 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2641 vxlan->flags |= VXLAN_F_PROXY;
2642
2643 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2644 vxlan->flags |= VXLAN_F_RSC;
2645
2646 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2647 vxlan->flags |= VXLAN_F_L2MISS;
2648
2649 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2650 vxlan->flags |= VXLAN_F_L3MISS;
2651
2652 if (data[IFLA_VXLAN_LIMIT])
2653 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2654
2655 if (data[IFLA_VXLAN_PORT_RANGE]) {
2656 const struct ifla_vxlan_port_range *p
2657 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2658 vxlan->port_min = ntohs(p->low);
2659 vxlan->port_max = ntohs(p->high);
2660 }
2661
2662 if (data[IFLA_VXLAN_PORT])
2663 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2664
2665 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2666 vxlan->flags |= VXLAN_F_UDP_CSUM;
2667
2668 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2669 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2670 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2671
2672 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2673 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2674 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2675
2676 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
2677 pr_info("duplicate VNI %u\n", vni);
2678 return -EEXIST;
2679 }
2680
2681 dev->ethtool_ops = &vxlan_ethtool_ops;
2682
2683 /* create an fdb entry for a valid default destination */
2684 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2685 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2686 &vxlan->default_dst.remote_ip,
2687 NUD_REACHABLE|NUD_PERMANENT,
2688 NLM_F_EXCL|NLM_F_CREATE,
2689 vxlan->dst_port,
2690 vxlan->default_dst.remote_vni,
2691 vxlan->default_dst.remote_ifindex,
2692 NTF_SELF);
2693 if (err)
2694 return err;
2695 }
2696
2697 err = register_netdevice(dev);
2698 if (err) {
2699 vxlan_fdb_delete_default(vxlan);
2700 return err;
2701 }
2702
2703 list_add(&vxlan->next, &vn->vxlan_list);
2704
2705 return 0;
2706 }
2707
2708 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2709 {
2710 struct vxlan_dev *vxlan = netdev_priv(dev);
2711 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2712
2713 spin_lock(&vn->sock_lock);
2714 if (!hlist_unhashed(&vxlan->hlist))
2715 hlist_del_rcu(&vxlan->hlist);
2716 spin_unlock(&vn->sock_lock);
2717
2718 list_del(&vxlan->next);
2719 unregister_netdevice_queue(dev, head);
2720 }
2721
2722 static size_t vxlan_get_size(const struct net_device *dev)
2723 {
2724
2725 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
2726 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
2727 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
2728 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
2729 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2730 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2731 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
2732 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2733 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2734 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2735 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
2736 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2737 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
2738 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
2739 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2740 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2741 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2742 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
2743 0;
2744 }
2745
2746 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2747 {
2748 const struct vxlan_dev *vxlan = netdev_priv(dev);
2749 const struct vxlan_rdst *dst = &vxlan->default_dst;
2750 struct ifla_vxlan_port_range ports = {
2751 .low = htons(vxlan->port_min),
2752 .high = htons(vxlan->port_max),
2753 };
2754
2755 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
2756 goto nla_put_failure;
2757
2758 if (!vxlan_addr_any(&dst->remote_ip)) {
2759 if (dst->remote_ip.sa.sa_family == AF_INET) {
2760 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2761 dst->remote_ip.sin.sin_addr.s_addr))
2762 goto nla_put_failure;
2763 #if IS_ENABLED(CONFIG_IPV6)
2764 } else {
2765 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2766 &dst->remote_ip.sin6.sin6_addr))
2767 goto nla_put_failure;
2768 #endif
2769 }
2770 }
2771
2772 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
2773 goto nla_put_failure;
2774
2775 if (!vxlan_addr_any(&vxlan->saddr)) {
2776 if (vxlan->saddr.sa.sa_family == AF_INET) {
2777 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2778 vxlan->saddr.sin.sin_addr.s_addr))
2779 goto nla_put_failure;
2780 #if IS_ENABLED(CONFIG_IPV6)
2781 } else {
2782 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2783 &vxlan->saddr.sin6.sin6_addr))
2784 goto nla_put_failure;
2785 #endif
2786 }
2787 }
2788
2789 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2790 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
2791 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2792 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2793 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2794 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2795 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2796 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2797 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2798 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2799 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
2800 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
2801 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
2802 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port) ||
2803 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
2804 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
2805 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
2806 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
2807 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
2808 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)))
2809 goto nla_put_failure;
2810
2811 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2812 goto nla_put_failure;
2813
2814 return 0;
2815
2816 nla_put_failure:
2817 return -EMSGSIZE;
2818 }
2819
2820 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2821 .kind = "vxlan",
2822 .maxtype = IFLA_VXLAN_MAX,
2823 .policy = vxlan_policy,
2824 .priv_size = sizeof(struct vxlan_dev),
2825 .setup = vxlan_setup,
2826 .validate = vxlan_validate,
2827 .newlink = vxlan_newlink,
2828 .dellink = vxlan_dellink,
2829 .get_size = vxlan_get_size,
2830 .fill_info = vxlan_fill_info,
2831 };
2832
2833 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
2834 struct net_device *dev)
2835 {
2836 struct vxlan_dev *vxlan, *next;
2837 LIST_HEAD(list_kill);
2838
2839 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2840 struct vxlan_rdst *dst = &vxlan->default_dst;
2841
2842 /* In case we created vxlan device with carrier
2843 * and we loose the carrier due to module unload
2844 * we also need to remove vxlan device. In other
2845 * cases, it's not necessary and remote_ifindex
2846 * is 0 here, so no matches.
2847 */
2848 if (dst->remote_ifindex == dev->ifindex)
2849 vxlan_dellink(vxlan->dev, &list_kill);
2850 }
2851
2852 unregister_netdevice_many(&list_kill);
2853 }
2854
2855 static int vxlan_lowerdev_event(struct notifier_block *unused,
2856 unsigned long event, void *ptr)
2857 {
2858 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2859 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
2860
2861 if (event == NETDEV_UNREGISTER)
2862 vxlan_handle_lowerdev_unregister(vn, dev);
2863
2864 return NOTIFY_DONE;
2865 }
2866
2867 static struct notifier_block vxlan_notifier_block __read_mostly = {
2868 .notifier_call = vxlan_lowerdev_event,
2869 };
2870
2871 static __net_init int vxlan_init_net(struct net *net)
2872 {
2873 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2874 unsigned int h;
2875
2876 INIT_LIST_HEAD(&vn->vxlan_list);
2877 spin_lock_init(&vn->sock_lock);
2878
2879 for (h = 0; h < PORT_HASH_SIZE; ++h)
2880 INIT_HLIST_HEAD(&vn->sock_list[h]);
2881
2882 return 0;
2883 }
2884
2885 static void __net_exit vxlan_exit_net(struct net *net)
2886 {
2887 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2888 struct vxlan_dev *vxlan, *next;
2889 struct net_device *dev, *aux;
2890 LIST_HEAD(list);
2891
2892 rtnl_lock();
2893 for_each_netdev_safe(net, dev, aux)
2894 if (dev->rtnl_link_ops == &vxlan_link_ops)
2895 unregister_netdevice_queue(dev, &list);
2896
2897 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2898 /* If vxlan->dev is in the same netns, it has already been added
2899 * to the list by the previous loop.
2900 */
2901 if (!net_eq(dev_net(vxlan->dev), net))
2902 unregister_netdevice_queue(dev, &list);
2903 }
2904
2905 unregister_netdevice_many(&list);
2906 rtnl_unlock();
2907 }
2908
2909 static struct pernet_operations vxlan_net_ops = {
2910 .init = vxlan_init_net,
2911 .exit = vxlan_exit_net,
2912 .id = &vxlan_net_id,
2913 .size = sizeof(struct vxlan_net),
2914 };
2915
2916 static int __init vxlan_init_module(void)
2917 {
2918 int rc;
2919
2920 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2921 if (!vxlan_wq)
2922 return -ENOMEM;
2923
2924 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2925
2926 rc = register_pernet_subsys(&vxlan_net_ops);
2927 if (rc)
2928 goto out1;
2929
2930 rc = register_netdevice_notifier(&vxlan_notifier_block);
2931 if (rc)
2932 goto out2;
2933
2934 rc = rtnl_link_register(&vxlan_link_ops);
2935 if (rc)
2936 goto out3;
2937
2938 return 0;
2939 out3:
2940 unregister_netdevice_notifier(&vxlan_notifier_block);
2941 out2:
2942 unregister_pernet_subsys(&vxlan_net_ops);
2943 out1:
2944 destroy_workqueue(vxlan_wq);
2945 return rc;
2946 }
2947 late_initcall(vxlan_init_module);
2948
2949 static void __exit vxlan_cleanup_module(void)
2950 {
2951 rtnl_link_unregister(&vxlan_link_ops);
2952 unregister_netdevice_notifier(&vxlan_notifier_block);
2953 destroy_workqueue(vxlan_wq);
2954 unregister_pernet_subsys(&vxlan_net_ops);
2955 /* rcu_barrier() is called by netns */
2956 }
2957 module_exit(vxlan_cleanup_module);
2958
2959 MODULE_LICENSE("GPL");
2960 MODULE_VERSION(VXLAN_VERSION);
2961 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
2962 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
2963 MODULE_ALIAS_RTNL_LINK("vxlan");
Linux kernel - Deliverables
This page took 0.163324 seconds and 5 git commands to generate.