projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
vxlan: Call udp_flow_src_port
[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, int idx)
937 {
938 struct vxlan_dev *vxlan = netdev_priv(dev);
939 unsigned int h;
940
941 for (h = 0; h < FDB_HASH_SIZE; ++h) {
942 struct vxlan_fdb *f;
943 int err;
944
945 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
946 struct vxlan_rdst *rd;
947
948 if (idx < cb->args[0])
949 goto skip;
950
951 list_for_each_entry_rcu(rd, &f->remotes, list) {
952 err = vxlan_fdb_info(skb, vxlan, f,
953 NETLINK_CB(cb->skb).portid,
954 cb->nlh->nlmsg_seq,
955 RTM_NEWNEIGH,
956 NLM_F_MULTI, rd);
957 if (err < 0)
958 goto out;
959 }
960 skip:
961 ++idx;
962 }
963 }
964 out:
965 return idx;
966 }
967
968 /* Watch incoming packets to learn mapping between Ethernet address
969 * and Tunnel endpoint.
970 * Return true if packet is bogus and should be droppped.
971 */
972 static bool vxlan_snoop(struct net_device *dev,
973 union vxlan_addr *src_ip, const u8 *src_mac)
974 {
975 struct vxlan_dev *vxlan = netdev_priv(dev);
976 struct vxlan_fdb *f;
977
978 f = vxlan_find_mac(vxlan, src_mac);
979 if (likely(f)) {
980 struct vxlan_rdst *rdst = first_remote_rcu(f);
981
982 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
983 return false;
984
985 /* Don't migrate static entries, drop packets */
986 if (f->state & NUD_NOARP)
987 return true;
988
989 if (net_ratelimit())
990 netdev_info(dev,
991 "%pM migrated from %pIS to %pIS\n",
992 src_mac, &rdst->remote_ip, &src_ip);
993
994 rdst->remote_ip = *src_ip;
995 f->updated = jiffies;
996 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
997 } else {
998 /* learned new entry */
999 spin_lock(&vxlan->hash_lock);
1000
1001 /* close off race between vxlan_flush and incoming packets */
1002 if (netif_running(dev))
1003 vxlan_fdb_create(vxlan, src_mac, src_ip,
1004 NUD_REACHABLE,
1005 NLM_F_EXCL|NLM_F_CREATE,
1006 vxlan->dst_port,
1007 vxlan->default_dst.remote_vni,
1008 0, NTF_SELF);
1009 spin_unlock(&vxlan->hash_lock);
1010 }
1011
1012 return false;
1013 }
1014
1015 /* See if multicast group is already in use by other ID */
1016 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1017 {
1018 struct vxlan_dev *vxlan;
1019
1020 /* The vxlan_sock is only used by dev, leaving group has
1021 * no effect on other vxlan devices.
1022 */
1023 if (atomic_read(&dev->vn_sock->refcnt) == 1)
1024 return false;
1025
1026 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1027 if (!netif_running(vxlan->dev) || vxlan == dev)
1028 continue;
1029
1030 if (vxlan->vn_sock != dev->vn_sock)
1031 continue;
1032
1033 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1034 &dev->default_dst.remote_ip))
1035 continue;
1036
1037 if (vxlan->default_dst.remote_ifindex !=
1038 dev->default_dst.remote_ifindex)
1039 continue;
1040
1041 return true;
1042 }
1043
1044 return false;
1045 }
1046
1047 static void vxlan_sock_hold(struct vxlan_sock *vs)
1048 {
1049 atomic_inc(&vs->refcnt);
1050 }
1051
1052 void vxlan_sock_release(struct vxlan_sock *vs)
1053 {
1054 struct sock *sk = vs->sock->sk;
1055 struct net *net = sock_net(sk);
1056 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1057
1058 if (!atomic_dec_and_test(&vs->refcnt))
1059 return;
1060
1061 spin_lock(&vn->sock_lock);
1062 hlist_del_rcu(&vs->hlist);
1063 rcu_assign_sk_user_data(vs->sock->sk, NULL);
1064 vxlan_notify_del_rx_port(vs);
1065 spin_unlock(&vn->sock_lock);
1066
1067 queue_work(vxlan_wq, &vs->del_work);
1068 }
1069 EXPORT_SYMBOL_GPL(vxlan_sock_release);
1070
1071 /* Callback to update multicast group membership when first VNI on
1072 * multicast asddress is brought up
1073 * Done as workqueue because ip_mc_join_group acquires RTNL.
1074 */
1075 static void vxlan_igmp_join(struct work_struct *work)
1076 {
1077 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
1078 struct vxlan_sock *vs = vxlan->vn_sock;
1079 struct sock *sk = vs->sock->sk;
1080 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1081 int ifindex = vxlan->default_dst.remote_ifindex;
1082
1083 lock_sock(sk);
1084 if (ip->sa.sa_family == AF_INET) {
1085 struct ip_mreqn mreq = {
1086 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1087 .imr_ifindex = ifindex,
1088 };
1089
1090 ip_mc_join_group(sk, &mreq);
1091 #if IS_ENABLED(CONFIG_IPV6)
1092 } else {
1093 ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1094 &ip->sin6.sin6_addr);
1095 #endif
1096 }
1097 release_sock(sk);
1098
1099 vxlan_sock_release(vs);
1100 dev_put(vxlan->dev);
1101 }
1102
1103 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1104 static void vxlan_igmp_leave(struct work_struct *work)
1105 {
1106 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
1107 struct vxlan_sock *vs = vxlan->vn_sock;
1108 struct sock *sk = vs->sock->sk;
1109 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1110 int ifindex = vxlan->default_dst.remote_ifindex;
1111
1112 lock_sock(sk);
1113 if (ip->sa.sa_family == AF_INET) {
1114 struct ip_mreqn mreq = {
1115 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1116 .imr_ifindex = ifindex,
1117 };
1118
1119 ip_mc_leave_group(sk, &mreq);
1120 #if IS_ENABLED(CONFIG_IPV6)
1121 } else {
1122 ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1123 &ip->sin6.sin6_addr);
1124 #endif
1125 }
1126
1127 release_sock(sk);
1128
1129 vxlan_sock_release(vs);
1130 dev_put(vxlan->dev);
1131 }
1132
1133 /* Callback from net/ipv4/udp.c to receive packets */
1134 static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1135 {
1136 struct vxlan_sock *vs;
1137 struct vxlanhdr *vxh;
1138
1139 /* Need Vxlan and inner Ethernet header to be present */
1140 if (!pskb_may_pull(skb, VXLAN_HLEN))
1141 goto error;
1142
1143 /* Return packets with reserved bits set */
1144 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1145 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
1146 (vxh->vx_vni & htonl(0xff))) {
1147 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1148 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1149 goto error;
1150 }
1151
1152 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1153 goto drop;
1154
1155 vs = rcu_dereference_sk_user_data(sk);
1156 if (!vs)
1157 goto drop;
1158
1159 skb_pop_rcv_encapsulation(skb);
1160
1161 vs->rcv(vs, skb, vxh->vx_vni);
1162 return 0;
1163
1164 drop:
1165 /* Consume bad packet */
1166 kfree_skb(skb);
1167 return 0;
1168
1169 error:
1170 /* Return non vxlan pkt */
1171 return 1;
1172 }
1173
1174 static void vxlan_rcv(struct vxlan_sock *vs,
1175 struct sk_buff *skb, __be32 vx_vni)
1176 {
1177 struct iphdr *oip = NULL;
1178 struct ipv6hdr *oip6 = NULL;
1179 struct vxlan_dev *vxlan;
1180 struct pcpu_sw_netstats *stats;
1181 union vxlan_addr saddr;
1182 __u32 vni;
1183 int err = 0;
1184 union vxlan_addr *remote_ip;
1185
1186 vni = ntohl(vx_vni) >> 8;
1187 /* Is this VNI defined? */
1188 vxlan = vxlan_vs_find_vni(vs, vni);
1189 if (!vxlan)
1190 goto drop;
1191
1192 remote_ip = &vxlan->default_dst.remote_ip;
1193 skb_reset_mac_header(skb);
1194 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
1195 skb->protocol = eth_type_trans(skb, vxlan->dev);
1196 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1197
1198 /* Ignore packet loops (and multicast echo) */
1199 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1200 goto drop;
1201
1202 /* Re-examine inner Ethernet packet */
1203 if (remote_ip->sa.sa_family == AF_INET) {
1204 oip = ip_hdr(skb);
1205 saddr.sin.sin_addr.s_addr = oip->saddr;
1206 saddr.sa.sa_family = AF_INET;
1207 #if IS_ENABLED(CONFIG_IPV6)
1208 } else {
1209 oip6 = ipv6_hdr(skb);
1210 saddr.sin6.sin6_addr = oip6->saddr;
1211 saddr.sa.sa_family = AF_INET6;
1212 #endif
1213 }
1214
1215 if ((vxlan->flags & VXLAN_F_LEARN) &&
1216 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1217 goto drop;
1218
1219 skb_reset_network_header(skb);
1220
1221 if (oip6)
1222 err = IP6_ECN_decapsulate(oip6, skb);
1223 if (oip)
1224 err = IP_ECN_decapsulate(oip, skb);
1225
1226 if (unlikely(err)) {
1227 if (log_ecn_error) {
1228 if (oip6)
1229 net_info_ratelimited("non-ECT from %pI6\n",
1230 &oip6->saddr);
1231 if (oip)
1232 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1233 &oip->saddr, oip->tos);
1234 }
1235 if (err > 1) {
1236 ++vxlan->dev->stats.rx_frame_errors;
1237 ++vxlan->dev->stats.rx_errors;
1238 goto drop;
1239 }
1240 }
1241
1242 stats = this_cpu_ptr(vxlan->dev->tstats);
1243 u64_stats_update_begin(&stats->syncp);
1244 stats->rx_packets++;
1245 stats->rx_bytes += skb->len;
1246 u64_stats_update_end(&stats->syncp);
1247
1248 netif_rx(skb);
1249
1250 return;
1251 drop:
1252 /* Consume bad packet */
1253 kfree_skb(skb);
1254 }
1255
1256 static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1257 {
1258 struct vxlan_dev *vxlan = netdev_priv(dev);
1259 struct arphdr *parp;
1260 u8 *arpptr, *sha;
1261 __be32 sip, tip;
1262 struct neighbour *n;
1263
1264 if (dev->flags & IFF_NOARP)
1265 goto out;
1266
1267 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1268 dev->stats.tx_dropped++;
1269 goto out;
1270 }
1271 parp = arp_hdr(skb);
1272
1273 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1274 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1275 parp->ar_pro != htons(ETH_P_IP) ||
1276 parp->ar_op != htons(ARPOP_REQUEST) ||
1277 parp->ar_hln != dev->addr_len ||
1278 parp->ar_pln != 4)
1279 goto out;
1280 arpptr = (u8 *)parp + sizeof(struct arphdr);
1281 sha = arpptr;
1282 arpptr += dev->addr_len; /* sha */
1283 memcpy(&sip, arpptr, sizeof(sip));
1284 arpptr += sizeof(sip);
1285 arpptr += dev->addr_len; /* tha */
1286 memcpy(&tip, arpptr, sizeof(tip));
1287
1288 if (ipv4_is_loopback(tip) ||
1289 ipv4_is_multicast(tip))
1290 goto out;
1291
1292 n = neigh_lookup(&arp_tbl, &tip, dev);
1293
1294 if (n) {
1295 struct vxlan_fdb *f;
1296 struct sk_buff *reply;
1297
1298 if (!(n->nud_state & NUD_CONNECTED)) {
1299 neigh_release(n);
1300 goto out;
1301 }
1302
1303 f = vxlan_find_mac(vxlan, n->ha);
1304 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1305 /* bridge-local neighbor */
1306 neigh_release(n);
1307 goto out;
1308 }
1309
1310 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1311 n->ha, sha);
1312
1313 neigh_release(n);
1314
1315 if (reply == NULL)
1316 goto out;
1317
1318 skb_reset_mac_header(reply);
1319 __skb_pull(reply, skb_network_offset(reply));
1320 reply->ip_summed = CHECKSUM_UNNECESSARY;
1321 reply->pkt_type = PACKET_HOST;
1322
1323 if (netif_rx_ni(reply) == NET_RX_DROP)
1324 dev->stats.rx_dropped++;
1325 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1326 union vxlan_addr ipa = {
1327 .sin.sin_addr.s_addr = tip,
1328 .sa.sa_family = AF_INET,
1329 };
1330
1331 vxlan_ip_miss(dev, &ipa);
1332 }
1333 out:
1334 consume_skb(skb);
1335 return NETDEV_TX_OK;
1336 }
1337
1338 #if IS_ENABLED(CONFIG_IPV6)
1339
1340 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1341 struct neighbour *n, bool isrouter)
1342 {
1343 struct net_device *dev = request->dev;
1344 struct sk_buff *reply;
1345 struct nd_msg *ns, *na;
1346 struct ipv6hdr *pip6;
1347 u8 *daddr;
1348 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1349 int ns_olen;
1350 int i, len;
1351
1352 if (dev == NULL)
1353 return NULL;
1354
1355 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1356 sizeof(*na) + na_olen + dev->needed_tailroom;
1357 reply = alloc_skb(len, GFP_ATOMIC);
1358 if (reply == NULL)
1359 return NULL;
1360
1361 reply->protocol = htons(ETH_P_IPV6);
1362 reply->dev = dev;
1363 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1364 skb_push(reply, sizeof(struct ethhdr));
1365 skb_set_mac_header(reply, 0);
1366
1367 ns = (struct nd_msg *)skb_transport_header(request);
1368
1369 daddr = eth_hdr(request)->h_source;
1370 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1371 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1372 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1373 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1374 break;
1375 }
1376 }
1377
1378 /* Ethernet header */
1379 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1380 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1381 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1382 reply->protocol = htons(ETH_P_IPV6);
1383
1384 skb_pull(reply, sizeof(struct ethhdr));
1385 skb_set_network_header(reply, 0);
1386 skb_put(reply, sizeof(struct ipv6hdr));
1387
1388 /* IPv6 header */
1389
1390 pip6 = ipv6_hdr(reply);
1391 memset(pip6, 0, sizeof(struct ipv6hdr));
1392 pip6->version = 6;
1393 pip6->priority = ipv6_hdr(request)->priority;
1394 pip6->nexthdr = IPPROTO_ICMPV6;
1395 pip6->hop_limit = 255;
1396 pip6->daddr = ipv6_hdr(request)->saddr;
1397 pip6->saddr = *(struct in6_addr *)n->primary_key;
1398
1399 skb_pull(reply, sizeof(struct ipv6hdr));
1400 skb_set_transport_header(reply, 0);
1401
1402 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1403
1404 /* Neighbor Advertisement */
1405 memset(na, 0, sizeof(*na)+na_olen);
1406 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1407 na->icmph.icmp6_router = isrouter;
1408 na->icmph.icmp6_override = 1;
1409 na->icmph.icmp6_solicited = 1;
1410 na->target = ns->target;
1411 ether_addr_copy(&na->opt[2], n->ha);
1412 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1413 na->opt[1] = na_olen >> 3;
1414
1415 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1416 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1417 csum_partial(na, sizeof(*na)+na_olen, 0));
1418
1419 pip6->payload_len = htons(sizeof(*na)+na_olen);
1420
1421 skb_push(reply, sizeof(struct ipv6hdr));
1422
1423 reply->ip_summed = CHECKSUM_UNNECESSARY;
1424
1425 return reply;
1426 }
1427
1428 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1429 {
1430 struct vxlan_dev *vxlan = netdev_priv(dev);
1431 struct nd_msg *msg;
1432 const struct ipv6hdr *iphdr;
1433 const struct in6_addr *saddr, *daddr;
1434 struct neighbour *n;
1435 struct inet6_dev *in6_dev;
1436
1437 in6_dev = __in6_dev_get(dev);
1438 if (!in6_dev)
1439 goto out;
1440
1441 if (!pskb_may_pull(skb, skb->len))
1442 goto out;
1443
1444 iphdr = ipv6_hdr(skb);
1445 saddr = &iphdr->saddr;
1446 daddr = &iphdr->daddr;
1447
1448 msg = (struct nd_msg *)skb_transport_header(skb);
1449 if (msg->icmph.icmp6_code != 0 ||
1450 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1451 goto out;
1452
1453 if (ipv6_addr_loopback(daddr) ||
1454 ipv6_addr_is_multicast(&msg->target))
1455 goto out;
1456
1457 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
1458
1459 if (n) {
1460 struct vxlan_fdb *f;
1461 struct sk_buff *reply;
1462
1463 if (!(n->nud_state & NUD_CONNECTED)) {
1464 neigh_release(n);
1465 goto out;
1466 }
1467
1468 f = vxlan_find_mac(vxlan, n->ha);
1469 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1470 /* bridge-local neighbor */
1471 neigh_release(n);
1472 goto out;
1473 }
1474
1475 reply = vxlan_na_create(skb, n,
1476 !!(f ? f->flags & NTF_ROUTER : 0));
1477
1478 neigh_release(n);
1479
1480 if (reply == NULL)
1481 goto out;
1482
1483 if (netif_rx_ni(reply) == NET_RX_DROP)
1484 dev->stats.rx_dropped++;
1485
1486 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1487 union vxlan_addr ipa = {
1488 .sin6.sin6_addr = msg->target,
1489 .sa.sa_family = AF_INET6,
1490 };
1491
1492 vxlan_ip_miss(dev, &ipa);
1493 }
1494
1495 out:
1496 consume_skb(skb);
1497 return NETDEV_TX_OK;
1498 }
1499 #endif
1500
1501 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1502 {
1503 struct vxlan_dev *vxlan = netdev_priv(dev);
1504 struct neighbour *n;
1505
1506 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1507 return false;
1508
1509 n = NULL;
1510 switch (ntohs(eth_hdr(skb)->h_proto)) {
1511 case ETH_P_IP:
1512 {
1513 struct iphdr *pip;
1514
1515 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1516 return false;
1517 pip = ip_hdr(skb);
1518 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1519 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1520 union vxlan_addr ipa = {
1521 .sin.sin_addr.s_addr = pip->daddr,
1522 .sa.sa_family = AF_INET,
1523 };
1524
1525 vxlan_ip_miss(dev, &ipa);
1526 return false;
1527 }
1528
1529 break;
1530 }
1531 #if IS_ENABLED(CONFIG_IPV6)
1532 case ETH_P_IPV6:
1533 {
1534 struct ipv6hdr *pip6;
1535
1536 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1537 return false;
1538 pip6 = ipv6_hdr(skb);
1539 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1540 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1541 union vxlan_addr ipa = {
1542 .sin6.sin6_addr = pip6->daddr,
1543 .sa.sa_family = AF_INET6,
1544 };
1545
1546 vxlan_ip_miss(dev, &ipa);
1547 return false;
1548 }
1549
1550 break;
1551 }
1552 #endif
1553 default:
1554 return false;
1555 }
1556
1557 if (n) {
1558 bool diff;
1559
1560 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1561 if (diff) {
1562 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1563 dev->addr_len);
1564 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1565 }
1566 neigh_release(n);
1567 return diff;
1568 }
1569
1570 return false;
1571 }
1572
1573 static inline struct sk_buff *vxlan_handle_offloads(struct sk_buff *skb,
1574 bool udp_csum)
1575 {
1576 int type = udp_csum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1577 return iptunnel_handle_offloads(skb, udp_csum, type);
1578 }
1579
1580 #if IS_ENABLED(CONFIG_IPV6)
1581 static int vxlan6_xmit_skb(struct vxlan_sock *vs,
1582 struct dst_entry *dst, struct sk_buff *skb,
1583 struct net_device *dev, struct in6_addr *saddr,
1584 struct in6_addr *daddr, __u8 prio, __u8 ttl,
1585 __be16 src_port, __be16 dst_port, __be32 vni,
1586 bool xnet)
1587 {
1588 struct ipv6hdr *ip6h;
1589 struct vxlanhdr *vxh;
1590 struct udphdr *uh;
1591 int min_headroom;
1592 int err;
1593
1594 skb = vxlan_handle_offloads(skb, !udp_get_no_check6_tx(vs->sock->sk));
1595 if (IS_ERR(skb))
1596 return -EINVAL;
1597
1598 skb_scrub_packet(skb, xnet);
1599
1600 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1601 + VXLAN_HLEN + sizeof(struct ipv6hdr)
1602 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1603
1604 /* Need space for new headers (invalidates iph ptr) */
1605 err = skb_cow_head(skb, min_headroom);
1606 if (unlikely(err))
1607 return err;
1608
1609 if (vlan_tx_tag_present(skb)) {
1610 if (WARN_ON(!__vlan_put_tag(skb,
1611 skb->vlan_proto,
1612 vlan_tx_tag_get(skb))))
1613 return -ENOMEM;
1614
1615 skb->vlan_tci = 0;
1616 }
1617
1618 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1619 vxh->vx_flags = htonl(VXLAN_FLAGS);
1620 vxh->vx_vni = vni;
1621
1622 __skb_push(skb, sizeof(*uh));
1623 skb_reset_transport_header(skb);
1624 uh = udp_hdr(skb);
1625
1626 uh->dest = dst_port;
1627 uh->source = src_port;
1628
1629 uh->len = htons(skb->len);
1630
1631 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1632 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1633 IPSKB_REROUTED);
1634 skb_dst_set(skb, dst);
1635
1636 udp6_set_csum(udp_get_no_check6_tx(vs->sock->sk), skb,
1637 saddr, daddr, skb->len);
1638
1639 __skb_push(skb, sizeof(*ip6h));
1640 skb_reset_network_header(skb);
1641 ip6h = ipv6_hdr(skb);
1642 ip6h->version = 6;
1643 ip6h->priority = prio;
1644 ip6h->flow_lbl[0] = 0;
1645 ip6h->flow_lbl[1] = 0;
1646 ip6h->flow_lbl[2] = 0;
1647 ip6h->payload_len = htons(skb->len);
1648 ip6h->nexthdr = IPPROTO_UDP;
1649 ip6h->hop_limit = ttl;
1650 ip6h->daddr = *daddr;
1651 ip6h->saddr = *saddr;
1652
1653 ip6tunnel_xmit(skb, dev);
1654 return 0;
1655 }
1656 #endif
1657
1658 int vxlan_xmit_skb(struct vxlan_sock *vs,
1659 struct rtable *rt, struct sk_buff *skb,
1660 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
1661 __be16 src_port, __be16 dst_port, __be32 vni, bool xnet)
1662 {
1663 struct vxlanhdr *vxh;
1664 struct udphdr *uh;
1665 int min_headroom;
1666 int err;
1667
1668 skb = vxlan_handle_offloads(skb, !vs->sock->sk->sk_no_check_tx);
1669 if (IS_ERR(skb))
1670 return -EINVAL;
1671
1672 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1673 + VXLAN_HLEN + sizeof(struct iphdr)
1674 + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
1675
1676 /* Need space for new headers (invalidates iph ptr) */
1677 err = skb_cow_head(skb, min_headroom);
1678 if (unlikely(err))
1679 return err;
1680
1681 if (vlan_tx_tag_present(skb)) {
1682 if (WARN_ON(!__vlan_put_tag(skb,
1683 skb->vlan_proto,
1684 vlan_tx_tag_get(skb))))
1685 return -ENOMEM;
1686
1687 skb->vlan_tci = 0;
1688 }
1689
1690 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1691 vxh->vx_flags = htonl(VXLAN_FLAGS);
1692 vxh->vx_vni = vni;
1693
1694 __skb_push(skb, sizeof(*uh));
1695 skb_reset_transport_header(skb);
1696 uh = udp_hdr(skb);
1697
1698 uh->dest = dst_port;
1699 uh->source = src_port;
1700
1701 uh->len = htons(skb->len);
1702
1703 udp_set_csum(vs->sock->sk->sk_no_check_tx, skb,
1704 src, dst, skb->len);
1705
1706 return iptunnel_xmit(vs->sock->sk, rt, skb, src, dst, IPPROTO_UDP,
1707 tos, ttl, df, xnet);
1708 }
1709 EXPORT_SYMBOL_GPL(vxlan_xmit_skb);
1710
1711 /* Bypass encapsulation if the destination is local */
1712 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1713 struct vxlan_dev *dst_vxlan)
1714 {
1715 struct pcpu_sw_netstats *tx_stats, *rx_stats;
1716 union vxlan_addr loopback;
1717 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
1718
1719 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1720 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1721 skb->pkt_type = PACKET_HOST;
1722 skb->encapsulation = 0;
1723 skb->dev = dst_vxlan->dev;
1724 __skb_pull(skb, skb_network_offset(skb));
1725
1726 if (remote_ip->sa.sa_family == AF_INET) {
1727 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1728 loopback.sa.sa_family = AF_INET;
1729 #if IS_ENABLED(CONFIG_IPV6)
1730 } else {
1731 loopback.sin6.sin6_addr = in6addr_loopback;
1732 loopback.sa.sa_family = AF_INET6;
1733 #endif
1734 }
1735
1736 if (dst_vxlan->flags & VXLAN_F_LEARN)
1737 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
1738
1739 u64_stats_update_begin(&tx_stats->syncp);
1740 tx_stats->tx_packets++;
1741 tx_stats->tx_bytes += skb->len;
1742 u64_stats_update_end(&tx_stats->syncp);
1743
1744 if (netif_rx(skb) == NET_RX_SUCCESS) {
1745 u64_stats_update_begin(&rx_stats->syncp);
1746 rx_stats->rx_packets++;
1747 rx_stats->rx_bytes += skb->len;
1748 u64_stats_update_end(&rx_stats->syncp);
1749 } else {
1750 skb->dev->stats.rx_dropped++;
1751 }
1752 }
1753
1754 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1755 struct vxlan_rdst *rdst, bool did_rsc)
1756 {
1757 struct vxlan_dev *vxlan = netdev_priv(dev);
1758 struct rtable *rt = NULL;
1759 const struct iphdr *old_iph;
1760 struct flowi4 fl4;
1761 union vxlan_addr *dst;
1762 __be16 src_port = 0, dst_port;
1763 u32 vni;
1764 __be16 df = 0;
1765 __u8 tos, ttl;
1766 int err;
1767
1768 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
1769 vni = rdst->remote_vni;
1770 dst = &rdst->remote_ip;
1771
1772 if (vxlan_addr_any(dst)) {
1773 if (did_rsc) {
1774 /* short-circuited back to local bridge */
1775 vxlan_encap_bypass(skb, vxlan, vxlan);
1776 return;
1777 }
1778 goto drop;
1779 }
1780
1781 old_iph = ip_hdr(skb);
1782
1783 ttl = vxlan->ttl;
1784 if (!ttl && vxlan_addr_multicast(dst))
1785 ttl = 1;
1786
1787 tos = vxlan->tos;
1788 if (tos == 1)
1789 tos = ip_tunnel_get_dsfield(old_iph, skb);
1790
1791 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->port_min,
1792 vxlan->port_max, true);
1793
1794 if (dst->sa.sa_family == AF_INET) {
1795 memset(&fl4, 0, sizeof(fl4));
1796 fl4.flowi4_oif = rdst->remote_ifindex;
1797 fl4.flowi4_tos = RT_TOS(tos);
1798 fl4.daddr = dst->sin.sin_addr.s_addr;
1799 fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr;
1800
1801 rt = ip_route_output_key(vxlan->net, &fl4);
1802 if (IS_ERR(rt)) {
1803 netdev_dbg(dev, "no route to %pI4\n",
1804 &dst->sin.sin_addr.s_addr);
1805 dev->stats.tx_carrier_errors++;
1806 goto tx_error;
1807 }
1808
1809 if (rt->dst.dev == dev) {
1810 netdev_dbg(dev, "circular route to %pI4\n",
1811 &dst->sin.sin_addr.s_addr);
1812 dev->stats.collisions++;
1813 goto rt_tx_error;
1814 }
1815
1816 /* Bypass encapsulation if the destination is local */
1817 if (rt->rt_flags & RTCF_LOCAL &&
1818 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1819 struct vxlan_dev *dst_vxlan;
1820
1821 ip_rt_put(rt);
1822 dst_vxlan = vxlan_find_vni(vxlan->net, vni, dst_port);
1823 if (!dst_vxlan)
1824 goto tx_error;
1825 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1826 return;
1827 }
1828
1829 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1830 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
1831
1832 err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
1833 fl4.saddr, dst->sin.sin_addr.s_addr,
1834 tos, ttl, df, src_port, dst_port,
1835 htonl(vni << 8),
1836 !net_eq(vxlan->net, dev_net(vxlan->dev)));
1837
1838 if (err < 0)
1839 goto rt_tx_error;
1840 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1841 #if IS_ENABLED(CONFIG_IPV6)
1842 } else {
1843 struct sock *sk = vxlan->vn_sock->sock->sk;
1844 struct dst_entry *ndst;
1845 struct flowi6 fl6;
1846 u32 flags;
1847
1848 memset(&fl6, 0, sizeof(fl6));
1849 fl6.flowi6_oif = rdst->remote_ifindex;
1850 fl6.daddr = dst->sin6.sin6_addr;
1851 fl6.saddr = vxlan->saddr.sin6.sin6_addr;
1852 fl6.flowi6_proto = IPPROTO_UDP;
1853
1854 if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) {
1855 netdev_dbg(dev, "no route to %pI6\n",
1856 &dst->sin6.sin6_addr);
1857 dev->stats.tx_carrier_errors++;
1858 goto tx_error;
1859 }
1860
1861 if (ndst->dev == dev) {
1862 netdev_dbg(dev, "circular route to %pI6\n",
1863 &dst->sin6.sin6_addr);
1864 dst_release(ndst);
1865 dev->stats.collisions++;
1866 goto tx_error;
1867 }
1868
1869 /* Bypass encapsulation if the destination is local */
1870 flags = ((struct rt6_info *)ndst)->rt6i_flags;
1871 if (flags & RTF_LOCAL &&
1872 !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1873 struct vxlan_dev *dst_vxlan;
1874
1875 dst_release(ndst);
1876 dst_vxlan = vxlan_find_vni(vxlan->net, vni, dst_port);
1877 if (!dst_vxlan)
1878 goto tx_error;
1879 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1880 return;
1881 }
1882
1883 ttl = ttl ? : ip6_dst_hoplimit(ndst);
1884
1885 err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
1886 dev, &fl6.saddr, &fl6.daddr, 0, ttl,
1887 src_port, dst_port, htonl(vni << 8),
1888 !net_eq(vxlan->net, dev_net(vxlan->dev)));
1889 #endif
1890 }
1891
1892 return;
1893
1894 drop:
1895 dev->stats.tx_dropped++;
1896 goto tx_free;
1897
1898 rt_tx_error:
1899 ip_rt_put(rt);
1900 tx_error:
1901 dev->stats.tx_errors++;
1902 tx_free:
1903 dev_kfree_skb(skb);
1904 }
1905
1906 /* Transmit local packets over Vxlan
1907 *
1908 * Outer IP header inherits ECN and DF from inner header.
1909 * Outer UDP destination is the VXLAN assigned port.
1910 * source port is based on hash of flow
1911 */
1912 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1913 {
1914 struct vxlan_dev *vxlan = netdev_priv(dev);
1915 struct ethhdr *eth;
1916 bool did_rsc = false;
1917 struct vxlan_rdst *rdst, *fdst = NULL;
1918 struct vxlan_fdb *f;
1919
1920 skb_reset_mac_header(skb);
1921 eth = eth_hdr(skb);
1922
1923 if ((vxlan->flags & VXLAN_F_PROXY)) {
1924 if (ntohs(eth->h_proto) == ETH_P_ARP)
1925 return arp_reduce(dev, skb);
1926 #if IS_ENABLED(CONFIG_IPV6)
1927 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
1928 skb->len >= sizeof(struct ipv6hdr) + sizeof(struct nd_msg) &&
1929 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
1930 struct nd_msg *msg;
1931
1932 msg = (struct nd_msg *)skb_transport_header(skb);
1933 if (msg->icmph.icmp6_code == 0 &&
1934 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
1935 return neigh_reduce(dev, skb);
1936 }
1937 #endif
1938 }
1939
1940 f = vxlan_find_mac(vxlan, eth->h_dest);
1941 did_rsc = false;
1942
1943 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
1944 (ntohs(eth->h_proto) == ETH_P_IP ||
1945 ntohs(eth->h_proto) == ETH_P_IPV6)) {
1946 did_rsc = route_shortcircuit(dev, skb);
1947 if (did_rsc)
1948 f = vxlan_find_mac(vxlan, eth->h_dest);
1949 }
1950
1951 if (f == NULL) {
1952 f = vxlan_find_mac(vxlan, all_zeros_mac);
1953 if (f == NULL) {
1954 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1955 !is_multicast_ether_addr(eth->h_dest))
1956 vxlan_fdb_miss(vxlan, eth->h_dest);
1957
1958 dev->stats.tx_dropped++;
1959 kfree_skb(skb);
1960 return NETDEV_TX_OK;
1961 }
1962 }
1963
1964 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1965 struct sk_buff *skb1;
1966
1967 if (!fdst) {
1968 fdst = rdst;
1969 continue;
1970 }
1971 skb1 = skb_clone(skb, GFP_ATOMIC);
1972 if (skb1)
1973 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
1974 }
1975
1976 if (fdst)
1977 vxlan_xmit_one(skb, dev, fdst, did_rsc);
1978 else
1979 kfree_skb(skb);
1980 return NETDEV_TX_OK;
1981 }
1982
1983 /* Walk the forwarding table and purge stale entries */
1984 static void vxlan_cleanup(unsigned long arg)
1985 {
1986 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1987 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1988 unsigned int h;
1989
1990 if (!netif_running(vxlan->dev))
1991 return;
1992
1993 spin_lock_bh(&vxlan->hash_lock);
1994 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1995 struct hlist_node *p, *n;
1996 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1997 struct vxlan_fdb *f
1998 = container_of(p, struct vxlan_fdb, hlist);
1999 unsigned long timeout;
2000
2001 if (f->state & NUD_PERMANENT)
2002 continue;
2003
2004 timeout = f->used + vxlan->age_interval * HZ;
2005 if (time_before_eq(timeout, jiffies)) {
2006 netdev_dbg(vxlan->dev,
2007 "garbage collect %pM\n",
2008 f->eth_addr);
2009 f->state = NUD_STALE;
2010 vxlan_fdb_destroy(vxlan, f);
2011 } else if (time_before(timeout, next_timer))
2012 next_timer = timeout;
2013 }
2014 }
2015 spin_unlock_bh(&vxlan->hash_lock);
2016
2017 mod_timer(&vxlan->age_timer, next_timer);
2018 }
2019
2020 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2021 {
2022 __u32 vni = vxlan->default_dst.remote_vni;
2023
2024 vxlan->vn_sock = vs;
2025 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
2026 }
2027
2028 /* Setup stats when device is created */
2029 static int vxlan_init(struct net_device *dev)
2030 {
2031 struct vxlan_dev *vxlan = netdev_priv(dev);
2032 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2033 struct vxlan_sock *vs;
2034
2035 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2036 if (!dev->tstats)
2037 return -ENOMEM;
2038
2039 spin_lock(&vn->sock_lock);
2040 vs = vxlan_find_sock(vxlan->net, vxlan->dst_port);
2041 if (vs) {
2042 /* If we have a socket with same port already, reuse it */
2043 atomic_inc(&vs->refcnt);
2044 vxlan_vs_add_dev(vs, vxlan);
2045 } else {
2046 /* otherwise make new socket outside of RTNL */
2047 dev_hold(dev);
2048 queue_work(vxlan_wq, &vxlan->sock_work);
2049 }
2050 spin_unlock(&vn->sock_lock);
2051
2052 return 0;
2053 }
2054
2055 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
2056 {
2057 struct vxlan_fdb *f;
2058
2059 spin_lock_bh(&vxlan->hash_lock);
2060 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2061 if (f)
2062 vxlan_fdb_destroy(vxlan, f);
2063 spin_unlock_bh(&vxlan->hash_lock);
2064 }
2065
2066 static void vxlan_uninit(struct net_device *dev)
2067 {
2068 struct vxlan_dev *vxlan = netdev_priv(dev);
2069 struct vxlan_sock *vs = vxlan->vn_sock;
2070
2071 vxlan_fdb_delete_default(vxlan);
2072
2073 if (vs)
2074 vxlan_sock_release(vs);
2075 free_percpu(dev->tstats);
2076 }
2077
2078 /* Start ageing timer and join group when device is brought up */
2079 static int vxlan_open(struct net_device *dev)
2080 {
2081 struct vxlan_dev *vxlan = netdev_priv(dev);
2082 struct vxlan_sock *vs = vxlan->vn_sock;
2083
2084 /* socket hasn't been created */
2085 if (!vs)
2086 return -ENOTCONN;
2087
2088 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
2089 vxlan_sock_hold(vs);
2090 dev_hold(dev);
2091 queue_work(vxlan_wq, &vxlan->igmp_join);
2092 }
2093
2094 if (vxlan->age_interval)
2095 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2096
2097 return 0;
2098 }
2099
2100 /* Purge the forwarding table */
2101 static void vxlan_flush(struct vxlan_dev *vxlan)
2102 {
2103 unsigned int h;
2104
2105 spin_lock_bh(&vxlan->hash_lock);
2106 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2107 struct hlist_node *p, *n;
2108 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2109 struct vxlan_fdb *f
2110 = container_of(p, struct vxlan_fdb, hlist);
2111 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2112 if (!is_zero_ether_addr(f->eth_addr))
2113 vxlan_fdb_destroy(vxlan, f);
2114 }
2115 }
2116 spin_unlock_bh(&vxlan->hash_lock);
2117 }
2118
2119 /* Cleanup timer and forwarding table on shutdown */
2120 static int vxlan_stop(struct net_device *dev)
2121 {
2122 struct vxlan_dev *vxlan = netdev_priv(dev);
2123 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2124 struct vxlan_sock *vs = vxlan->vn_sock;
2125
2126 if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2127 !vxlan_group_used(vn, vxlan)) {
2128 vxlan_sock_hold(vs);
2129 dev_hold(dev);
2130 queue_work(vxlan_wq, &vxlan->igmp_leave);
2131 }
2132
2133 del_timer_sync(&vxlan->age_timer);
2134
2135 vxlan_flush(vxlan);
2136
2137 return 0;
2138 }
2139
2140 /* Stub, nothing needs to be done. */
2141 static void vxlan_set_multicast_list(struct net_device *dev)
2142 {
2143 }
2144
2145 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2146 {
2147 struct vxlan_dev *vxlan = netdev_priv(dev);
2148 struct vxlan_rdst *dst = &vxlan->default_dst;
2149 struct net_device *lowerdev;
2150 int max_mtu;
2151
2152 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
2153 if (lowerdev == NULL)
2154 return eth_change_mtu(dev, new_mtu);
2155
2156 if (dst->remote_ip.sa.sa_family == AF_INET6)
2157 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2158 else
2159 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2160
2161 if (new_mtu < 68 || new_mtu > max_mtu)
2162 return -EINVAL;
2163
2164 dev->mtu = new_mtu;
2165 return 0;
2166 }
2167
2168 static const struct net_device_ops vxlan_netdev_ops = {
2169 .ndo_init = vxlan_init,
2170 .ndo_uninit = vxlan_uninit,
2171 .ndo_open = vxlan_open,
2172 .ndo_stop = vxlan_stop,
2173 .ndo_start_xmit = vxlan_xmit,
2174 .ndo_get_stats64 = ip_tunnel_get_stats64,
2175 .ndo_set_rx_mode = vxlan_set_multicast_list,
2176 .ndo_change_mtu = vxlan_change_mtu,
2177 .ndo_validate_addr = eth_validate_addr,
2178 .ndo_set_mac_address = eth_mac_addr,
2179 .ndo_fdb_add = vxlan_fdb_add,
2180 .ndo_fdb_del = vxlan_fdb_delete,
2181 .ndo_fdb_dump = vxlan_fdb_dump,
2182 };
2183
2184 /* Info for udev, that this is a virtual tunnel endpoint */
2185 static struct device_type vxlan_type = {
2186 .name = "vxlan",
2187 };
2188
2189 /* Calls the ndo_add_vxlan_port of the caller in order to
2190 * supply the listening VXLAN udp ports. Callers are expected
2191 * to implement the ndo_add_vxlan_port.
2192 */
2193 void vxlan_get_rx_port(struct net_device *dev)
2194 {
2195 struct vxlan_sock *vs;
2196 struct net *net = dev_net(dev);
2197 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2198 sa_family_t sa_family;
2199 __be16 port;
2200 unsigned int i;
2201
2202 spin_lock(&vn->sock_lock);
2203 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2204 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2205 port = inet_sk(vs->sock->sk)->inet_sport;
2206 sa_family = vs->sock->sk->sk_family;
2207 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2208 port);
2209 }
2210 }
2211 spin_unlock(&vn->sock_lock);
2212 }
2213 EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2214
2215 /* Initialize the device structure. */
2216 static void vxlan_setup(struct net_device *dev)
2217 {
2218 struct vxlan_dev *vxlan = netdev_priv(dev);
2219 unsigned int h;
2220
2221 eth_hw_addr_random(dev);
2222 ether_setup(dev);
2223 if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6)
2224 dev->needed_headroom = ETH_HLEN + VXLAN6_HEADROOM;
2225 else
2226 dev->needed_headroom = ETH_HLEN + VXLAN_HEADROOM;
2227
2228 dev->netdev_ops = &vxlan_netdev_ops;
2229 dev->destructor = free_netdev;
2230 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2231
2232 dev->tx_queue_len = 0;
2233 dev->features |= NETIF_F_LLTX;
2234 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
2235 dev->features |= NETIF_F_RXCSUM;
2236 dev->features |= NETIF_F_GSO_SOFTWARE;
2237
2238 dev->vlan_features = dev->features;
2239 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2240 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2241 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2242 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2243 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
2244 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2245
2246 INIT_LIST_HEAD(&vxlan->next);
2247 spin_lock_init(&vxlan->hash_lock);
2248 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
2249 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
2250 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
2251
2252 init_timer_deferrable(&vxlan->age_timer);
2253 vxlan->age_timer.function = vxlan_cleanup;
2254 vxlan->age_timer.data = (unsigned long) vxlan;
2255
2256 vxlan->dst_port = htons(vxlan_port);
2257
2258 vxlan->dev = dev;
2259
2260 for (h = 0; h < FDB_HASH_SIZE; ++h)
2261 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2262 }
2263
2264 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2265 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
2266 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2267 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
2268 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2269 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2270 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
2271 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2272 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2273 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2274 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2275 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
2276 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
2277 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2278 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2279 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2280 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
2281 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
2282 };
2283
2284 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2285 {
2286 if (tb[IFLA_ADDRESS]) {
2287 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2288 pr_debug("invalid link address (not ethernet)\n");
2289 return -EINVAL;
2290 }
2291
2292 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2293 pr_debug("invalid all zero ethernet address\n");
2294 return -EADDRNOTAVAIL;
2295 }
2296 }
2297
2298 if (!data)
2299 return -EINVAL;
2300
2301 if (data[IFLA_VXLAN_ID]) {
2302 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2303 if (id >= VXLAN_VID_MASK)
2304 return -ERANGE;
2305 }
2306
2307 if (data[IFLA_VXLAN_PORT_RANGE]) {
2308 const struct ifla_vxlan_port_range *p
2309 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2310
2311 if (ntohs(p->high) < ntohs(p->low)) {
2312 pr_debug("port range %u .. %u not valid\n",
2313 ntohs(p->low), ntohs(p->high));
2314 return -EINVAL;
2315 }
2316 }
2317
2318 return 0;
2319 }
2320
2321 static void vxlan_get_drvinfo(struct net_device *netdev,
2322 struct ethtool_drvinfo *drvinfo)
2323 {
2324 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2325 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2326 }
2327
2328 static const struct ethtool_ops vxlan_ethtool_ops = {
2329 .get_drvinfo = vxlan_get_drvinfo,
2330 .get_link = ethtool_op_get_link,
2331 };
2332
2333 static void vxlan_del_work(struct work_struct *work)
2334 {
2335 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2336
2337 sk_release_kernel(vs->sock->sk);
2338 kfree_rcu(vs, rcu);
2339 }
2340
2341 #if IS_ENABLED(CONFIG_IPV6)
2342 /* Create UDP socket for encapsulation receive. AF_INET6 socket
2343 * could be used for both IPv4 and IPv6 communications, but
2344 * users may set bindv6only=1.
2345 */
2346 static struct socket *create_v6_sock(struct net *net, __be16 port, u32 flags)
2347 {
2348 struct sock *sk;
2349 struct socket *sock;
2350 struct sockaddr_in6 vxlan_addr = {
2351 .sin6_family = AF_INET6,
2352 .sin6_port = port,
2353 };
2354 int rc, val = 1;
2355
2356 rc = sock_create_kern(AF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock);
2357 if (rc < 0) {
2358 pr_debug("UDPv6 socket create failed\n");
2359 return ERR_PTR(rc);
2360 }
2361
2362 /* Put in proper namespace */
2363 sk = sock->sk;
2364 sk_change_net(sk, net);
2365
2366 kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
2367 (char *)&val, sizeof(val));
2368 rc = kernel_bind(sock, (struct sockaddr *)&vxlan_addr,
2369 sizeof(struct sockaddr_in6));
2370 if (rc < 0) {
2371 pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n",
2372 &vxlan_addr.sin6_addr, ntohs(vxlan_addr.sin6_port), rc);
2373 sk_release_kernel(sk);
2374 return ERR_PTR(rc);
2375 }
2376 /* At this point, IPv6 module should have been loaded in
2377 * sock_create_kern().
2378 */
2379 BUG_ON(!ipv6_stub);
2380
2381 /* Disable multicast loopback */
2382 inet_sk(sk)->mc_loop = 0;
2383
2384 if (flags & VXLAN_F_UDP_ZERO_CSUM6_TX)
2385 udp_set_no_check6_tx(sk, true);
2386
2387 if (flags & VXLAN_F_UDP_ZERO_CSUM6_RX)
2388 udp_set_no_check6_rx(sk, true);
2389
2390 return sock;
2391 }
2392
2393 #else
2394
2395 static struct socket *create_v6_sock(struct net *net, __be16 port, u32 flags)
2396 {
2397 return ERR_PTR(-EPFNOSUPPORT);
2398 }
2399 #endif
2400
2401 static struct socket *create_v4_sock(struct net *net, __be16 port, u32 flags)
2402 {
2403 struct sock *sk;
2404 struct socket *sock;
2405 struct sockaddr_in vxlan_addr = {
2406 .sin_family = AF_INET,
2407 .sin_addr.s_addr = htonl(INADDR_ANY),
2408 .sin_port = port,
2409 };
2410 int rc;
2411
2412 /* Create UDP socket for encapsulation receive. */
2413 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock);
2414 if (rc < 0) {
2415 pr_debug("UDP socket create failed\n");
2416 return ERR_PTR(rc);
2417 }
2418
2419 /* Put in proper namespace */
2420 sk = sock->sk;
2421 sk_change_net(sk, net);
2422
2423 rc = kernel_bind(sock, (struct sockaddr *) &vxlan_addr,
2424 sizeof(vxlan_addr));
2425 if (rc < 0) {
2426 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
2427 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
2428 sk_release_kernel(sk);
2429 return ERR_PTR(rc);
2430 }
2431
2432 /* Disable multicast loopback */
2433 inet_sk(sk)->mc_loop = 0;
2434
2435 if (!(flags & VXLAN_F_UDP_CSUM))
2436 sock->sk->sk_no_check_tx = 1;
2437
2438 return sock;
2439 }
2440
2441 /* Create new listen socket if needed */
2442 static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
2443 vxlan_rcv_t *rcv, void *data,
2444 u32 flags)
2445 {
2446 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2447 struct vxlan_sock *vs;
2448 struct socket *sock;
2449 struct sock *sk;
2450 unsigned int h;
2451 bool ipv6 = !!(flags & VXLAN_F_IPV6);
2452
2453 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
2454 if (!vs)
2455 return ERR_PTR(-ENOMEM);
2456
2457 for (h = 0; h < VNI_HASH_SIZE; ++h)
2458 INIT_HLIST_HEAD(&vs->vni_list[h]);
2459
2460 INIT_WORK(&vs->del_work, vxlan_del_work);
2461
2462 if (ipv6)
2463 sock = create_v6_sock(net, port, flags);
2464 else
2465 sock = create_v4_sock(net, port, flags);
2466 if (IS_ERR(sock)) {
2467 kfree(vs);
2468 return ERR_CAST(sock);
2469 }
2470
2471 vs->sock = sock;
2472 sk = sock->sk;
2473 atomic_set(&vs->refcnt, 1);
2474 vs->rcv = rcv;
2475 vs->data = data;
2476 rcu_assign_sk_user_data(vs->sock->sk, vs);
2477
2478 /* Initialize the vxlan udp offloads structure */
2479 vs->udp_offloads.port = port;
2480 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2481 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2482
2483 spin_lock(&vn->sock_lock);
2484 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2485 vxlan_notify_add_rx_port(vs);
2486 spin_unlock(&vn->sock_lock);
2487
2488 /* Mark socket as an encapsulation socket. */
2489 udp_sk(sk)->encap_type = 1;
2490 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
2491 #if IS_ENABLED(CONFIG_IPV6)
2492 if (ipv6)
2493 ipv6_stub->udpv6_encap_enable();
2494 else
2495 #endif
2496 udp_encap_enable();
2497
2498 return vs;
2499 }
2500
2501 struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2502 vxlan_rcv_t *rcv, void *data,
2503 bool no_share, u32 flags)
2504 {
2505 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2506 struct vxlan_sock *vs;
2507
2508 vs = vxlan_socket_create(net, port, rcv, data, flags);
2509 if (!IS_ERR(vs))
2510 return vs;
2511
2512 if (no_share) /* Return error if sharing is not allowed. */
2513 return vs;
2514
2515 spin_lock(&vn->sock_lock);
2516 vs = vxlan_find_sock(net, port);
2517 if (vs) {
2518 if (vs->rcv == rcv)
2519 atomic_inc(&vs->refcnt);
2520 else
2521 vs = ERR_PTR(-EBUSY);
2522 }
2523 spin_unlock(&vn->sock_lock);
2524
2525 if (!vs)
2526 vs = ERR_PTR(-EINVAL);
2527
2528 return vs;
2529 }
2530 EXPORT_SYMBOL_GPL(vxlan_sock_add);
2531
2532 /* Scheduled at device creation to bind to a socket */
2533 static void vxlan_sock_work(struct work_struct *work)
2534 {
2535 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
2536 struct net *net = vxlan->net;
2537 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2538 __be16 port = vxlan->dst_port;
2539 struct vxlan_sock *nvs;
2540
2541 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags);
2542 spin_lock(&vn->sock_lock);
2543 if (!IS_ERR(nvs))
2544 vxlan_vs_add_dev(nvs, vxlan);
2545 spin_unlock(&vn->sock_lock);
2546
2547 dev_put(vxlan->dev);
2548 }
2549
2550 static int vxlan_newlink(struct net *net, struct net_device *dev,
2551 struct nlattr *tb[], struct nlattr *data[])
2552 {
2553 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2554 struct vxlan_dev *vxlan = netdev_priv(dev);
2555 struct vxlan_rdst *dst = &vxlan->default_dst;
2556 __u32 vni;
2557 int err;
2558 bool use_ipv6 = false;
2559
2560 if (!data[IFLA_VXLAN_ID])
2561 return -EINVAL;
2562
2563 vxlan->net = dev_net(dev);
2564
2565 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
2566 dst->remote_vni = vni;
2567
2568 /* Unless IPv6 is explicitly requested, assume IPv4 */
2569 dst->remote_ip.sa.sa_family = AF_INET;
2570 if (data[IFLA_VXLAN_GROUP]) {
2571 dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]);
2572 } else if (data[IFLA_VXLAN_GROUP6]) {
2573 if (!IS_ENABLED(CONFIG_IPV6))
2574 return -EPFNOSUPPORT;
2575
2576 nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6],
2577 sizeof(struct in6_addr));
2578 dst->remote_ip.sa.sa_family = AF_INET6;
2579 use_ipv6 = true;
2580 }
2581
2582 if (data[IFLA_VXLAN_LOCAL]) {
2583 vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
2584 vxlan->saddr.sa.sa_family = AF_INET;
2585 } else if (data[IFLA_VXLAN_LOCAL6]) {
2586 if (!IS_ENABLED(CONFIG_IPV6))
2587 return -EPFNOSUPPORT;
2588
2589 /* TODO: respect scope id */
2590 nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6],
2591 sizeof(struct in6_addr));
2592 vxlan->saddr.sa.sa_family = AF_INET6;
2593 use_ipv6 = true;
2594 }
2595
2596 if (data[IFLA_VXLAN_LINK] &&
2597 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
2598 struct net_device *lowerdev
2599 = __dev_get_by_index(net, dst->remote_ifindex);
2600
2601 if (!lowerdev) {
2602 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
2603 return -ENODEV;
2604 }
2605
2606 #if IS_ENABLED(CONFIG_IPV6)
2607 if (use_ipv6) {
2608 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2609 if (idev && idev->cnf.disable_ipv6) {
2610 pr_info("IPv6 is disabled via sysctl\n");
2611 return -EPERM;
2612 }
2613 vxlan->flags |= VXLAN_F_IPV6;
2614 }
2615 #endif
2616
2617 if (!tb[IFLA_MTU])
2618 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2619
2620 dev->needed_headroom = lowerdev->hard_header_len +
2621 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2622 } else if (use_ipv6)
2623 vxlan->flags |= VXLAN_F_IPV6;
2624
2625 if (data[IFLA_VXLAN_TOS])
2626 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2627
2628 if (data[IFLA_VXLAN_TTL])
2629 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2630
2631 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
2632 vxlan->flags |= VXLAN_F_LEARN;
2633
2634 if (data[IFLA_VXLAN_AGEING])
2635 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2636 else
2637 vxlan->age_interval = FDB_AGE_DEFAULT;
2638
2639 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2640 vxlan->flags |= VXLAN_F_PROXY;
2641
2642 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2643 vxlan->flags |= VXLAN_F_RSC;
2644
2645 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2646 vxlan->flags |= VXLAN_F_L2MISS;
2647
2648 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2649 vxlan->flags |= VXLAN_F_L3MISS;
2650
2651 if (data[IFLA_VXLAN_LIMIT])
2652 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2653
2654 if (data[IFLA_VXLAN_PORT_RANGE]) {
2655 const struct ifla_vxlan_port_range *p
2656 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2657 vxlan->port_min = ntohs(p->low);
2658 vxlan->port_max = ntohs(p->high);
2659 }
2660
2661 if (data[IFLA_VXLAN_PORT])
2662 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2663
2664 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2665 vxlan->flags |= VXLAN_F_UDP_CSUM;
2666
2667 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2668 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2669 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2670
2671 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2672 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2673 vxlan->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2674
2675 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
2676 pr_info("duplicate VNI %u\n", vni);
2677 return -EEXIST;
2678 }
2679
2680 dev->ethtool_ops = &vxlan_ethtool_ops;
2681
2682 /* create an fdb entry for a valid default destination */
2683 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2684 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2685 &vxlan->default_dst.remote_ip,
2686 NUD_REACHABLE|NUD_PERMANENT,
2687 NLM_F_EXCL|NLM_F_CREATE,
2688 vxlan->dst_port,
2689 vxlan->default_dst.remote_vni,
2690 vxlan->default_dst.remote_ifindex,
2691 NTF_SELF);
2692 if (err)
2693 return err;
2694 }
2695
2696 err = register_netdevice(dev);
2697 if (err) {
2698 vxlan_fdb_delete_default(vxlan);
2699 return err;
2700 }
2701
2702 list_add(&vxlan->next, &vn->vxlan_list);
2703
2704 return 0;
2705 }
2706
2707 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2708 {
2709 struct vxlan_dev *vxlan = netdev_priv(dev);
2710 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2711
2712 spin_lock(&vn->sock_lock);
2713 if (!hlist_unhashed(&vxlan->hlist))
2714 hlist_del_rcu(&vxlan->hlist);
2715 spin_unlock(&vn->sock_lock);
2716
2717 list_del(&vxlan->next);
2718 unregister_netdevice_queue(dev, head);
2719 }
2720
2721 static size_t vxlan_get_size(const struct net_device *dev)
2722 {
2723
2724 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
2725 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
2726 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
2727 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
2728 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2729 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2730 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
2731 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2732 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2733 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2734 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
2735 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2736 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
2737 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
2738 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2739 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2740 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2741 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
2742 0;
2743 }
2744
2745 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2746 {
2747 const struct vxlan_dev *vxlan = netdev_priv(dev);
2748 const struct vxlan_rdst *dst = &vxlan->default_dst;
2749 struct ifla_vxlan_port_range ports = {
2750 .low = htons(vxlan->port_min),
2751 .high = htons(vxlan->port_max),
2752 };
2753
2754 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
2755 goto nla_put_failure;
2756
2757 if (!vxlan_addr_any(&dst->remote_ip)) {
2758 if (dst->remote_ip.sa.sa_family == AF_INET) {
2759 if (nla_put_be32(skb, IFLA_VXLAN_GROUP,
2760 dst->remote_ip.sin.sin_addr.s_addr))
2761 goto nla_put_failure;
2762 #if IS_ENABLED(CONFIG_IPV6)
2763 } else {
2764 if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr),
2765 &dst->remote_ip.sin6.sin6_addr))
2766 goto nla_put_failure;
2767 #endif
2768 }
2769 }
2770
2771 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
2772 goto nla_put_failure;
2773
2774 if (!vxlan_addr_any(&vxlan->saddr)) {
2775 if (vxlan->saddr.sa.sa_family == AF_INET) {
2776 if (nla_put_be32(skb, IFLA_VXLAN_LOCAL,
2777 vxlan->saddr.sin.sin_addr.s_addr))
2778 goto nla_put_failure;
2779 #if IS_ENABLED(CONFIG_IPV6)
2780 } else {
2781 if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr),
2782 &vxlan->saddr.sin6.sin6_addr))
2783 goto nla_put_failure;
2784 #endif
2785 }
2786 }
2787
2788 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
2789 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
2790 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2791 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2792 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2793 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2794 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2795 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2796 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2797 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2798 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
2799 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
2800 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
2801 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port) ||
2802 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
2803 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
2804 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
2805 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
2806 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
2807 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)))
2808 goto nla_put_failure;
2809
2810 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2811 goto nla_put_failure;
2812
2813 return 0;
2814
2815 nla_put_failure:
2816 return -EMSGSIZE;
2817 }
2818
2819 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2820 .kind = "vxlan",
2821 .maxtype = IFLA_VXLAN_MAX,
2822 .policy = vxlan_policy,
2823 .priv_size = sizeof(struct vxlan_dev),
2824 .setup = vxlan_setup,
2825 .validate = vxlan_validate,
2826 .newlink = vxlan_newlink,
2827 .dellink = vxlan_dellink,
2828 .get_size = vxlan_get_size,
2829 .fill_info = vxlan_fill_info,
2830 };
2831
2832 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
2833 struct net_device *dev)
2834 {
2835 struct vxlan_dev *vxlan, *next;
2836 LIST_HEAD(list_kill);
2837
2838 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2839 struct vxlan_rdst *dst = &vxlan->default_dst;
2840
2841 /* In case we created vxlan device with carrier
2842 * and we loose the carrier due to module unload
2843 * we also need to remove vxlan device. In other
2844 * cases, it's not necessary and remote_ifindex
2845 * is 0 here, so no matches.
2846 */
2847 if (dst->remote_ifindex == dev->ifindex)
2848 vxlan_dellink(vxlan->dev, &list_kill);
2849 }
2850
2851 unregister_netdevice_many(&list_kill);
2852 }
2853
2854 static int vxlan_lowerdev_event(struct notifier_block *unused,
2855 unsigned long event, void *ptr)
2856 {
2857 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2858 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
2859
2860 if (event == NETDEV_UNREGISTER)
2861 vxlan_handle_lowerdev_unregister(vn, dev);
2862
2863 return NOTIFY_DONE;
2864 }
2865
2866 static struct notifier_block vxlan_notifier_block __read_mostly = {
2867 .notifier_call = vxlan_lowerdev_event,
2868 };
2869
2870 static __net_init int vxlan_init_net(struct net *net)
2871 {
2872 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2873 unsigned int h;
2874
2875 INIT_LIST_HEAD(&vn->vxlan_list);
2876 spin_lock_init(&vn->sock_lock);
2877
2878 for (h = 0; h < PORT_HASH_SIZE; ++h)
2879 INIT_HLIST_HEAD(&vn->sock_list[h]);
2880
2881 return 0;
2882 }
2883
2884 static void __net_exit vxlan_exit_net(struct net *net)
2885 {
2886 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2887 struct vxlan_dev *vxlan, *next;
2888 struct net_device *dev, *aux;
2889 LIST_HEAD(list);
2890
2891 rtnl_lock();
2892 for_each_netdev_safe(net, dev, aux)
2893 if (dev->rtnl_link_ops == &vxlan_link_ops)
2894 unregister_netdevice_queue(dev, &list);
2895
2896 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2897 /* If vxlan->dev is in the same netns, it has already been added
2898 * to the list by the previous loop.
2899 */
2900 if (!net_eq(dev_net(vxlan->dev), net))
2901 unregister_netdevice_queue(dev, &list);
2902 }
2903
2904 unregister_netdevice_many(&list);
2905 rtnl_unlock();
2906 }
2907
2908 static struct pernet_operations vxlan_net_ops = {
2909 .init = vxlan_init_net,
2910 .exit = vxlan_exit_net,
2911 .id = &vxlan_net_id,
2912 .size = sizeof(struct vxlan_net),
2913 };
2914
2915 static int __init vxlan_init_module(void)
2916 {
2917 int rc;
2918
2919 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
2920 if (!vxlan_wq)
2921 return -ENOMEM;
2922
2923 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2924
2925 rc = register_pernet_subsys(&vxlan_net_ops);
2926 if (rc)
2927 goto out1;
2928
2929 rc = register_netdevice_notifier(&vxlan_notifier_block);
2930 if (rc)
2931 goto out2;
2932
2933 rc = rtnl_link_register(&vxlan_link_ops);
2934 if (rc)
2935 goto out3;
2936
2937 return 0;
2938 out3:
2939 unregister_netdevice_notifier(&vxlan_notifier_block);
2940 out2:
2941 unregister_pernet_subsys(&vxlan_net_ops);
2942 out1:
2943 destroy_workqueue(vxlan_wq);
2944 return rc;
2945 }
2946 late_initcall(vxlan_init_module);
2947
2948 static void __exit vxlan_cleanup_module(void)
2949 {
2950 rtnl_link_unregister(&vxlan_link_ops);
2951 unregister_netdevice_notifier(&vxlan_notifier_block);
2952 destroy_workqueue(vxlan_wq);
2953 unregister_pernet_subsys(&vxlan_net_ops);
2954 /* rcu_barrier() is called by netns */
2955 }
2956 module_exit(vxlan_cleanup_module);
2957
2958 MODULE_LICENSE("GPL");
2959 MODULE_VERSION(VXLAN_VERSION);
2960 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
2961 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
2962 MODULE_ALIAS_RTNL_LINK("vxlan");
Linux kernel - Deliverables
This page took 0.091783 seconds and 6 git commands to generate.