?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/evalenti/linux...
[linux/fpc-iii.git] / drivers / net / vxlan.c
blob1c0fa364323e280f2ed25c86bdcb8f3508a26c34
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/udp_tunnel.h>
37 #include <net/rtnetlink.h>
38 #include <net/route.h>
39 #include <net/dsfield.h>
40 #include <net/inet_ecn.h>
41 #include <net/net_namespace.h>
42 #include <net/netns/generic.h>
43 #include <net/vxlan.h>
44 #include <net/protocol.h>
45
46 #if IS_ENABLED(CONFIG_IPV6)
47 #include <net/ipv6.h>
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
50 #include <net/ip6_checksum.h>
51 #endif
52 #include <net/dst_metadata.h>
53
54 #define VXLAN_VERSION "0.1"
55
56 #define PORT_HASH_BITS 8
57 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
58 #define FDB_AGE_DEFAULT 300 /* 5 min */
59 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
60
61 /* UDP port for VXLAN traffic.
62 * The IANA assigned port is 4789, but the Linux default is 8472
63 * for compatibility with early adopters.
64 */
65 static unsigned short vxlan_port __read_mostly = 8472;
66 module_param_named(udp_port, vxlan_port, ushort, 0444);
67 MODULE_PARM_DESC(udp_port, "Destination UDP port");
68
69 static bool log_ecn_error = true;
70 module_param(log_ecn_error, bool, 0644);
71 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
72
73 static int vxlan_net_id;
74 static struct rtnl_link_ops vxlan_link_ops;
75
76 static const u8 all_zeros_mac[ETH_ALEN + 2];
77
78 static int vxlan_sock_add(struct vxlan_dev *vxlan);
79
80 /* per-network namespace private data for this module */
81 struct vxlan_net {
82 struct list_head vxlan_list;
83 struct hlist_head sock_list[PORT_HASH_SIZE];
84 spinlock_t sock_lock;
85 };
86
87 /* Forwarding table entry */
88 struct vxlan_fdb {
89 struct hlist_node hlist; /* linked list of entries */
90 struct rcu_head rcu;
91 unsigned long updated; /* jiffies */
92 unsigned long used;
93 struct list_head remotes;
94 u8 eth_addr[ETH_ALEN];
95 u16 state; /* see ndm_state */
96 u8 flags; /* see ndm_flags */
97 };
98
99 /* salt for hash table */
100 static u32 vxlan_salt __read_mostly;
101 static struct workqueue_struct *vxlan_wq;
102
103 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
104 {
105 return vs->flags & VXLAN_F_COLLECT_METADATA ||
106 ip_tunnel_collect_metadata();
107 }
108
109 #if IS_ENABLED(CONFIG_IPV6)
110 static inline
111 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
112 {
113 if (a->sa.sa_family != b->sa.sa_family)
114 return false;
115 if (a->sa.sa_family == AF_INET6)
116 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
117 else
118 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
119 }
120
121 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
122 {
123 if (ipa->sa.sa_family == AF_INET6)
124 return ipv6_addr_any(&ipa->sin6.sin6_addr);
125 else
126 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
127 }
128
129 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
130 {
131 if (ipa->sa.sa_family == AF_INET6)
132 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
133 else
134 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
135 }
136
137 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
138 {
139 if (nla_len(nla) >= sizeof(struct in6_addr)) {
140 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
141 ip->sa.sa_family = AF_INET6;
142 return 0;
143 } else if (nla_len(nla) >= sizeof(__be32)) {
144 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
145 ip->sa.sa_family = AF_INET;
146 return 0;
147 } else {
148 return -EAFNOSUPPORT;
149 }
150 }
151
152 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
153 const union vxlan_addr *ip)
154 {
155 if (ip->sa.sa_family == AF_INET6)
156 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
157 else
158 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
159 }
160
161 #else /* !CONFIG_IPV6 */
162
163 static inline
164 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
165 {
166 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
167 }
168
169 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
170 {
171 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
172 }
173
174 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
175 {
176 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
177 }
178
179 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
180 {
181 if (nla_len(nla) >= sizeof(struct in6_addr)) {
182 return -EAFNOSUPPORT;
183 } else if (nla_len(nla) >= sizeof(__be32)) {
184 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
185 ip->sa.sa_family = AF_INET;
186 return 0;
187 } else {
188 return -EAFNOSUPPORT;
189 }
190 }
191
192 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
193 const union vxlan_addr *ip)
194 {
195 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
196 }
197 #endif
198
199 /* Virtual Network hash table head */
200 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
201 {
202 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
203 }
204
205 /* Socket hash table head */
206 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
207 {
208 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
209
210 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
211 }
212
213 /* First remote destination for a forwarding entry.
214 * Guaranteed to be non-NULL because remotes are never deleted.
215 */
216 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
217 {
218 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
219 }
220
221 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
222 {
223 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
224 }
225
226 /* Find VXLAN socket based on network namespace, address family and UDP port
227 * and enabled unshareable flags.
228 */
229 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
230 __be16 port, u32 flags)
231 {
232 struct vxlan_sock *vs;
233
234 flags &= VXLAN_F_RCV_FLAGS;
235
236 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
237 if (inet_sk(vs->sock->sk)->inet_sport == port &&
238 vxlan_get_sk_family(vs) == family &&
239 vs->flags == flags)
240 return vs;
241 }
242 return NULL;
243 }
244
245 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni)
246 {
247 struct vxlan_dev *vxlan;
248
249 /* For flow based devices, map all packets to VNI 0 */
250 if (vs->flags & VXLAN_F_COLLECT_METADATA)
251 vni = 0;
252
253 hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) {
254 if (vxlan->default_dst.remote_vni == vni)
255 return vxlan;
256 }
257
258 return NULL;
259 }
260
261 /* Look up VNI in a per net namespace table */
262 static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni,
263 sa_family_t family, __be16 port,
264 u32 flags)
265 {
266 struct vxlan_sock *vs;
267
268 vs = vxlan_find_sock(net, family, port, flags);
269 if (!vs)
270 return NULL;
271
272 return vxlan_vs_find_vni(vs, vni);
273 }
274
275 /* Fill in neighbour message in skbuff. */
276 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
277 const struct vxlan_fdb *fdb,
278 u32 portid, u32 seq, int type, unsigned int flags,
279 const struct vxlan_rdst *rdst)
280 {
281 unsigned long now = jiffies;
282 struct nda_cacheinfo ci;
283 struct nlmsghdr *nlh;
284 struct ndmsg *ndm;
285 bool send_ip, send_eth;
286
287 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
288 if (nlh == NULL)
289 return -EMSGSIZE;
290
291 ndm = nlmsg_data(nlh);
292 memset(ndm, 0, sizeof(*ndm));
293
294 send_eth = send_ip = true;
295
296 if (type == RTM_GETNEIGH) {
297 ndm->ndm_family = AF_INET;
298 send_ip = !vxlan_addr_any(&rdst->remote_ip);
299 send_eth = !is_zero_ether_addr(fdb->eth_addr);
300 } else
301 ndm->ndm_family = AF_BRIDGE;
302 ndm->ndm_state = fdb->state;
303 ndm->ndm_ifindex = vxlan->dev->ifindex;
304 ndm->ndm_flags = fdb->flags;
305 ndm->ndm_type = RTN_UNICAST;
306
307 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
308 nla_put_s32(skb, NDA_LINK_NETNSID,
309 peernet2id_alloc(dev_net(vxlan->dev), vxlan->net)))
310 goto nla_put_failure;
311
312 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
313 goto nla_put_failure;
314
315 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
316 goto nla_put_failure;
317
318 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
319 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
320 goto nla_put_failure;
321 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
322 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
323 goto nla_put_failure;
324 if (rdst->remote_ifindex &&
325 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
326 goto nla_put_failure;
327
328 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
329 ci.ndm_confirmed = 0;
330 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
331 ci.ndm_refcnt = 0;
332
333 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
334 goto nla_put_failure;
335
336 nlmsg_end(skb, nlh);
337 return 0;
338
339 nla_put_failure:
340 nlmsg_cancel(skb, nlh);
341 return -EMSGSIZE;
342 }
343
344 static inline size_t vxlan_nlmsg_size(void)
345 {
346 return NLMSG_ALIGN(sizeof(struct ndmsg))
347 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
348 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
349 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
350 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
351 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
352 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
353 + nla_total_size(sizeof(struct nda_cacheinfo));
354 }
355
356 static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
357 struct vxlan_rdst *rd, int type)
358 {
359 struct net *net = dev_net(vxlan->dev);
360 struct sk_buff *skb;
361 int err = -ENOBUFS;
362
363 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
364 if (skb == NULL)
365 goto errout;
366
367 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
368 if (err < 0) {
369 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
370 WARN_ON(err == -EMSGSIZE);
371 kfree_skb(skb);
372 goto errout;
373 }
374
375 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
376 return;
377 errout:
378 if (err < 0)
379 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
380 }
381
382 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
383 {
384 struct vxlan_dev *vxlan = netdev_priv(dev);
385 struct vxlan_fdb f = {
386 .state = NUD_STALE,
387 };
388 struct vxlan_rdst remote = {
389 .remote_ip = *ipa, /* goes to NDA_DST */
390 .remote_vni = cpu_to_be32(VXLAN_N_VID),
391 };
392
393 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
394 }
395
396 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
397 {
398 struct vxlan_fdb f = {
399 .state = NUD_STALE,
400 };
401 struct vxlan_rdst remote = { };
402
403 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
404
405 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
406 }
407
408 /* Hash Ethernet address */
409 static u32 eth_hash(const unsigned char *addr)
410 {
411 u64 value = get_unaligned((u64 *)addr);
412
413 /* only want 6 bytes */
414 #ifdef __BIG_ENDIAN
415 value >>= 16;
416 #else
417 value <<= 16;
418 #endif
419 return hash_64(value, FDB_HASH_BITS);
420 }
421
422 /* Hash chain to use given mac address */
423 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
424 const u8 *mac)
425 {
426 return &vxlan->fdb_head[eth_hash(mac)];
427 }
428
429 /* Look up Ethernet address in forwarding table */
430 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
431 const u8 *mac)
432 {
433 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
434 struct vxlan_fdb *f;
435
436 hlist_for_each_entry_rcu(f, head, hlist) {
437 if (ether_addr_equal(mac, f->eth_addr))
438 return f;
439 }
440
441 return NULL;
442 }
443
444 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
445 const u8 *mac)
446 {
447 struct vxlan_fdb *f;
448
449 f = __vxlan_find_mac(vxlan, mac);
450 if (f)
451 f->used = jiffies;
452
453 return f;
454 }
455
456 /* caller should hold vxlan->hash_lock */
457 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
458 union vxlan_addr *ip, __be16 port,
459 __be32 vni, __u32 ifindex)
460 {
461 struct vxlan_rdst *rd;
462
463 list_for_each_entry(rd, &f->remotes, list) {
464 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
465 rd->remote_port == port &&
466 rd->remote_vni == vni &&
467 rd->remote_ifindex == ifindex)
468 return rd;
469 }
470
471 return NULL;
472 }
473
474 /* Replace destination of unicast mac */
475 static int vxlan_fdb_replace(struct vxlan_fdb *f,
476 union vxlan_addr *ip, __be16 port, __be32 vni,
477 __u32 ifindex)
478 {
479 struct vxlan_rdst *rd;
480
481 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
482 if (rd)
483 return 0;
484
485 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
486 if (!rd)
487 return 0;
488
489 dst_cache_reset(&rd->dst_cache);
490 rd->remote_ip = *ip;
491 rd->remote_port = port;
492 rd->remote_vni = vni;
493 rd->remote_ifindex = ifindex;
494 return 1;
495 }
496
497 /* Add/update destinations for multicast */
498 static int vxlan_fdb_append(struct vxlan_fdb *f,
499 union vxlan_addr *ip, __be16 port, __be32 vni,
500 __u32 ifindex, struct vxlan_rdst **rdp)
501 {
502 struct vxlan_rdst *rd;
503
504 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
505 if (rd)
506 return 0;
507
508 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
509 if (rd == NULL)
510 return -ENOBUFS;
511
512 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
513 kfree(rd);
514 return -ENOBUFS;
515 }
516
517 rd->remote_ip = *ip;
518 rd->remote_port = port;
519 rd->remote_vni = vni;
520 rd->remote_ifindex = ifindex;
521
522 list_add_tail_rcu(&rd->list, &f->remotes);
523
524 *rdp = rd;
525 return 1;
526 }
527
528 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
529 unsigned int off,
530 struct vxlanhdr *vh, size_t hdrlen,
531 __be32 vni_field,
532 struct gro_remcsum *grc,
533 bool nopartial)
534 {
535 size_t start, offset;
536
537 if (skb->remcsum_offload)
538 return vh;
539
540 if (!NAPI_GRO_CB(skb)->csum_valid)
541 return NULL;
542
543 start = vxlan_rco_start(vni_field);
544 offset = start + vxlan_rco_offset(vni_field);
545
546 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
547 start, offset, grc, nopartial);
548
549 skb->remcsum_offload = 1;
550
551 return vh;
552 }
553
554 static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
555 struct sk_buff *skb,
556 struct udp_offload *uoff)
557 {
558 struct sk_buff *p, **pp = NULL;
559 struct vxlanhdr *vh, *vh2;
560 unsigned int hlen, off_vx;
561 int flush = 1;
562 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
563 udp_offloads);
564 __be32 flags;
565 struct gro_remcsum grc;
566
567 skb_gro_remcsum_init(&grc);
568
569 off_vx = skb_gro_offset(skb);
570 hlen = off_vx + sizeof(*vh);
571 vh = skb_gro_header_fast(skb, off_vx);
572 if (skb_gro_header_hard(skb, hlen)) {
573 vh = skb_gro_header_slow(skb, hlen, off_vx);
574 if (unlikely(!vh))
575 goto out;
576 }
577
578 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
579
580 flags = vh->vx_flags;
581
582 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
583 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
584 vh->vx_vni, &grc,
585 !!(vs->flags &
586 VXLAN_F_REMCSUM_NOPARTIAL));
587
588 if (!vh)
589 goto out;
590 }
591
592 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
593
594 for (p = *head; p; p = p->next) {
595 if (!NAPI_GRO_CB(p)->same_flow)
596 continue;
597
598 vh2 = (struct vxlanhdr *)(p->data + off_vx);
599 if (vh->vx_flags != vh2->vx_flags ||
600 vh->vx_vni != vh2->vx_vni) {
601 NAPI_GRO_CB(p)->same_flow = 0;
602 continue;
603 }
604 }
605
606 pp = eth_gro_receive(head, skb);
607 flush = 0;
608
609 out:
610 skb_gro_remcsum_cleanup(skb, &grc);
611 NAPI_GRO_CB(skb)->flush |= flush;
612
613 return pp;
614 }
615
616 static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
617 struct udp_offload *uoff)
618 {
619 udp_tunnel_gro_complete(skb, nhoff);
620
621 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
622 }
623
624 /* Notify netdevs that UDP port started listening */
625 static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
626 {
627 struct net_device *dev;
628 struct sock *sk = vs->sock->sk;
629 struct net *net = sock_net(sk);
630 sa_family_t sa_family = vxlan_get_sk_family(vs);
631 __be16 port = inet_sk(sk)->inet_sport;
632 int err;
633
634 if (sa_family == AF_INET) {
635 err = udp_add_offload(net, &vs->udp_offloads);
636 if (err)
637 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
638 }
639
640 rcu_read_lock();
641 for_each_netdev_rcu(net, dev) {
642 if (dev->netdev_ops->ndo_add_vxlan_port)
643 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
644 port);
645 }
646 rcu_read_unlock();
647 }
648
649 /* Notify netdevs that UDP port is no more listening */
650 static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
651 {
652 struct net_device *dev;
653 struct sock *sk = vs->sock->sk;
654 struct net *net = sock_net(sk);
655 sa_family_t sa_family = vxlan_get_sk_family(vs);
656 __be16 port = inet_sk(sk)->inet_sport;
657
658 rcu_read_lock();
659 for_each_netdev_rcu(net, dev) {
660 if (dev->netdev_ops->ndo_del_vxlan_port)
661 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
662 port);
663 }
664 rcu_read_unlock();
665
666 if (sa_family == AF_INET)
667 udp_del_offload(&vs->udp_offloads);
668 }
669
670 /* Add new entry to forwarding table -- assumes lock held */
671 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
672 const u8 *mac, union vxlan_addr *ip,
673 __u16 state, __u16 flags,
674 __be16 port, __be32 vni, __u32 ifindex,
675 __u8 ndm_flags)
676 {
677 struct vxlan_rdst *rd = NULL;
678 struct vxlan_fdb *f;
679 int notify = 0;
680
681 f = __vxlan_find_mac(vxlan, mac);
682 if (f) {
683 if (flags & NLM_F_EXCL) {
684 netdev_dbg(vxlan->dev,
685 "lost race to create %pM\n", mac);
686 return -EEXIST;
687 }
688 if (f->state != state) {
689 f->state = state;
690 f->updated = jiffies;
691 notify = 1;
692 }
693 if (f->flags != ndm_flags) {
694 f->flags = ndm_flags;
695 f->updated = jiffies;
696 notify = 1;
697 }
698 if ((flags & NLM_F_REPLACE)) {
699 /* Only change unicasts */
700 if (!(is_multicast_ether_addr(f->eth_addr) ||
701 is_zero_ether_addr(f->eth_addr))) {
702 notify |= vxlan_fdb_replace(f, ip, port, vni,
703 ifindex);
704 } else
705 return -EOPNOTSUPP;
706 }
707 if ((flags & NLM_F_APPEND) &&
708 (is_multicast_ether_addr(f->eth_addr) ||
709 is_zero_ether_addr(f->eth_addr))) {
710 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
711 &rd);
712
713 if (rc < 0)
714 return rc;
715 notify |= rc;
716 }
717 } else {
718 if (!(flags & NLM_F_CREATE))
719 return -ENOENT;
720
721 if (vxlan->cfg.addrmax &&
722 vxlan->addrcnt >= vxlan->cfg.addrmax)
723 return -ENOSPC;
724
725 /* Disallow replace to add a multicast entry */
726 if ((flags & NLM_F_REPLACE) &&
727 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
728 return -EOPNOTSUPP;
729
730 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
731 f = kmalloc(sizeof(*f), GFP_ATOMIC);
732 if (!f)
733 return -ENOMEM;
734
735 notify = 1;
736 f->state = state;
737 f->flags = ndm_flags;
738 f->updated = f->used = jiffies;
739 INIT_LIST_HEAD(&f->remotes);
740 memcpy(f->eth_addr, mac, ETH_ALEN);
741
742 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
743
744 ++vxlan->addrcnt;
745 hlist_add_head_rcu(&f->hlist,
746 vxlan_fdb_head(vxlan, mac));
747 }
748
749 if (notify) {
750 if (rd == NULL)
751 rd = first_remote_rtnl(f);
752 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
753 }
754
755 return 0;
756 }
757
758 static void vxlan_fdb_free(struct rcu_head *head)
759 {
760 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
761 struct vxlan_rdst *rd, *nd;
762
763 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
764 dst_cache_destroy(&rd->dst_cache);
765 kfree(rd);
766 }
767 kfree(f);
768 }
769
770 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
771 {
772 netdev_dbg(vxlan->dev,
773 "delete %pM\n", f->eth_addr);
774
775 --vxlan->addrcnt;
776 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
777
778 hlist_del_rcu(&f->hlist);
779 call_rcu(&f->rcu, vxlan_fdb_free);
780 }
781
782 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
783 union vxlan_addr *ip, __be16 *port, __be32 *vni,
784 u32 *ifindex)
785 {
786 struct net *net = dev_net(vxlan->dev);
787 int err;
788
789 if (tb[NDA_DST]) {
790 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
791 if (err)
792 return err;
793 } else {
794 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
795 if (remote->sa.sa_family == AF_INET) {
796 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
797 ip->sa.sa_family = AF_INET;
798 #if IS_ENABLED(CONFIG_IPV6)
799 } else {
800 ip->sin6.sin6_addr = in6addr_any;
801 ip->sa.sa_family = AF_INET6;
802 #endif
803 }
804 }
805
806 if (tb[NDA_PORT]) {
807 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
808 return -EINVAL;
809 *port = nla_get_be16(tb[NDA_PORT]);
810 } else {
811 *port = vxlan->cfg.dst_port;
812 }
813
814 if (tb[NDA_VNI]) {
815 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
816 return -EINVAL;
817 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
818 } else {
819 *vni = vxlan->default_dst.remote_vni;
820 }
821
822 if (tb[NDA_IFINDEX]) {
823 struct net_device *tdev;
824
825 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
826 return -EINVAL;
827 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
828 tdev = __dev_get_by_index(net, *ifindex);
829 if (!tdev)
830 return -EADDRNOTAVAIL;
831 } else {
832 *ifindex = 0;
833 }
834
835 return 0;
836 }
837
838 /* Add static entry (via netlink) */
839 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
840 struct net_device *dev,
841 const unsigned char *addr, u16 vid, u16 flags)
842 {
843 struct vxlan_dev *vxlan = netdev_priv(dev);
844 /* struct net *net = dev_net(vxlan->dev); */
845 union vxlan_addr ip;
846 __be16 port;
847 __be32 vni;
848 u32 ifindex;
849 int err;
850
851 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
852 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
853 ndm->ndm_state);
854 return -EINVAL;
855 }
856
857 if (tb[NDA_DST] == NULL)
858 return -EINVAL;
859
860 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
861 if (err)
862 return err;
863
864 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
865 return -EAFNOSUPPORT;
866
867 spin_lock_bh(&vxlan->hash_lock);
868 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
869 port, vni, ifindex, ndm->ndm_flags);
870 spin_unlock_bh(&vxlan->hash_lock);
871
872 return err;
873 }
874
875 /* Delete entry (via netlink) */
876 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
877 struct net_device *dev,
878 const unsigned char *addr, u16 vid)
879 {
880 struct vxlan_dev *vxlan = netdev_priv(dev);
881 struct vxlan_fdb *f;
882 struct vxlan_rdst *rd = NULL;
883 union vxlan_addr ip;
884 __be16 port;
885 __be32 vni;
886 u32 ifindex;
887 int err;
888
889 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
890 if (err)
891 return err;
892
893 err = -ENOENT;
894
895 spin_lock_bh(&vxlan->hash_lock);
896 f = vxlan_find_mac(vxlan, addr);
897 if (!f)
898 goto out;
899
900 if (!vxlan_addr_any(&ip)) {
901 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
902 if (!rd)
903 goto out;
904 }
905
906 err = 0;
907
908 /* remove a destination if it's not the only one on the list,
909 * otherwise destroy the fdb entry
910 */
911 if (rd && !list_is_singular(&f->remotes)) {
912 list_del_rcu(&rd->list);
913 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
914 kfree_rcu(rd, rcu);
915 goto out;
916 }
917
918 vxlan_fdb_destroy(vxlan, f);
919
920 out:
921 spin_unlock_bh(&vxlan->hash_lock);
922
923 return err;
924 }
925
926 /* Dump forwarding table */
927 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
928 struct net_device *dev,
929 struct net_device *filter_dev, int idx)
930 {
931 struct vxlan_dev *vxlan = netdev_priv(dev);
932 unsigned int h;
933
934 for (h = 0; h < FDB_HASH_SIZE; ++h) {
935 struct vxlan_fdb *f;
936 int err;
937
938 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
939 struct vxlan_rdst *rd;
940
941 list_for_each_entry_rcu(rd, &f->remotes, list) {
942 if (idx < cb->args[0])
943 goto skip;
944
945 err = vxlan_fdb_info(skb, vxlan, f,
946 NETLINK_CB(cb->skb).portid,
947 cb->nlh->nlmsg_seq,
948 RTM_NEWNEIGH,
949 NLM_F_MULTI, rd);
950 if (err < 0) {
951 cb->args[1] = err;
952 goto out;
953 }
954 skip:
955 ++idx;
956 }
957 }
958 }
959 out:
960 return idx;
961 }
962
963 /* Watch incoming packets to learn mapping between Ethernet address
964 * and Tunnel endpoint.
965 * Return true if packet is bogus and should be dropped.
966 */
967 static bool vxlan_snoop(struct net_device *dev,
968 union vxlan_addr *src_ip, const u8 *src_mac)
969 {
970 struct vxlan_dev *vxlan = netdev_priv(dev);
971 struct vxlan_fdb *f;
972
973 f = vxlan_find_mac(vxlan, src_mac);
974 if (likely(f)) {
975 struct vxlan_rdst *rdst = first_remote_rcu(f);
976
977 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
978 return false;
979
980 /* Don't migrate static entries, drop packets */
981 if (f->state & NUD_NOARP)
982 return true;
983
984 if (net_ratelimit())
985 netdev_info(dev,
986 "%pM migrated from %pIS to %pIS\n",
987 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
988
989 rdst->remote_ip = *src_ip;
990 f->updated = jiffies;
991 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
992 } else {
993 /* learned new entry */
994 spin_lock(&vxlan->hash_lock);
995
996 /* close off race between vxlan_flush and incoming packets */
997 if (netif_running(dev))
998 vxlan_fdb_create(vxlan, src_mac, src_ip,
999 NUD_REACHABLE,
1000 NLM_F_EXCL|NLM_F_CREATE,
1001 vxlan->cfg.dst_port,
1002 vxlan->default_dst.remote_vni,
1003 0, NTF_SELF);
1004 spin_unlock(&vxlan->hash_lock);
1005 }
1006
1007 return false;
1008 }
1009
1010 /* See if multicast group is already in use by other ID */
1011 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1012 {
1013 struct vxlan_dev *vxlan;
1014 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1015
1016 /* The vxlan_sock is only used by dev, leaving group has
1017 * no effect on other vxlan devices.
1018 */
1019 if (family == AF_INET && dev->vn4_sock &&
1020 atomic_read(&dev->vn4_sock->refcnt) == 1)
1021 return false;
1022 #if IS_ENABLED(CONFIG_IPV6)
1023 if (family == AF_INET6 && dev->vn6_sock &&
1024 atomic_read(&dev->vn6_sock->refcnt) == 1)
1025 return false;
1026 #endif
1027
1028 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1029 if (!netif_running(vxlan->dev) || vxlan == dev)
1030 continue;
1031
1032 if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock)
1033 continue;
1034 #if IS_ENABLED(CONFIG_IPV6)
1035 if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock)
1036 continue;
1037 #endif
1038
1039 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1040 &dev->default_dst.remote_ip))
1041 continue;
1042
1043 if (vxlan->default_dst.remote_ifindex !=
1044 dev->default_dst.remote_ifindex)
1045 continue;
1046
1047 return true;
1048 }
1049
1050 return false;
1051 }
1052
1053 static void __vxlan_sock_release(struct vxlan_sock *vs)
1054 {
1055 struct vxlan_net *vn;
1056
1057 if (!vs)
1058 return;
1059 if (!atomic_dec_and_test(&vs->refcnt))
1060 return;
1061
1062 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1063 spin_lock(&vn->sock_lock);
1064 hlist_del_rcu(&vs->hlist);
1065 vxlan_notify_del_rx_port(vs);
1066 spin_unlock(&vn->sock_lock);
1067
1068 queue_work(vxlan_wq, &vs->del_work);
1069 }
1070
1071 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1072 {
1073 __vxlan_sock_release(vxlan->vn4_sock);
1074 #if IS_ENABLED(CONFIG_IPV6)
1075 __vxlan_sock_release(vxlan->vn6_sock);
1076 #endif
1077 }
1078
1079 /* Update multicast group membership when first VNI on
1080 * multicast address is brought up
1081 */
1082 static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1083 {
1084 struct sock *sk;
1085 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1086 int ifindex = vxlan->default_dst.remote_ifindex;
1087 int ret = -EINVAL;
1088
1089 if (ip->sa.sa_family == AF_INET) {
1090 struct ip_mreqn mreq = {
1091 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1092 .imr_ifindex = ifindex,
1093 };
1094
1095 sk = vxlan->vn4_sock->sock->sk;
1096 lock_sock(sk);
1097 ret = ip_mc_join_group(sk, &mreq);
1098 release_sock(sk);
1099 #if IS_ENABLED(CONFIG_IPV6)
1100 } else {
1101 sk = vxlan->vn6_sock->sock->sk;
1102 lock_sock(sk);
1103 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1104 &ip->sin6.sin6_addr);
1105 release_sock(sk);
1106 #endif
1107 }
1108
1109 return ret;
1110 }
1111
1112 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1113 static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1114 {
1115 struct sock *sk;
1116 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1117 int ifindex = vxlan->default_dst.remote_ifindex;
1118 int ret = -EINVAL;
1119
1120 if (ip->sa.sa_family == AF_INET) {
1121 struct ip_mreqn mreq = {
1122 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1123 .imr_ifindex = ifindex,
1124 };
1125
1126 sk = vxlan->vn4_sock->sock->sk;
1127 lock_sock(sk);
1128 ret = ip_mc_leave_group(sk, &mreq);
1129 release_sock(sk);
1130 #if IS_ENABLED(CONFIG_IPV6)
1131 } else {
1132 sk = vxlan->vn6_sock->sock->sk;
1133 lock_sock(sk);
1134 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1135 &ip->sin6.sin6_addr);
1136 release_sock(sk);
1137 #endif
1138 }
1139
1140 return ret;
1141 }
1142
1143 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1144 struct sk_buff *skb, u32 vxflags)
1145 {
1146 size_t start, offset;
1147
1148 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1149 goto out;
1150
1151 start = vxlan_rco_start(unparsed->vx_vni);
1152 offset = start + vxlan_rco_offset(unparsed->vx_vni);
1153
1154 if (!pskb_may_pull(skb, offset + sizeof(u16)))
1155 return false;
1156
1157 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1158 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1159 out:
1160 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1161 unparsed->vx_vni &= VXLAN_VNI_MASK;
1162 return true;
1163 }
1164
1165 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1166 struct sk_buff *skb, u32 vxflags,
1167 struct vxlan_metadata *md)
1168 {
1169 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1170 struct metadata_dst *tun_dst;
1171
1172 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1173 goto out;
1174
1175 md->gbp = ntohs(gbp->policy_id);
1176
1177 tun_dst = (struct metadata_dst *)skb_dst(skb);
1178 if (tun_dst) {
1179 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1180 tun_dst->u.tun_info.options_len = sizeof(*md);
1181 }
1182 if (gbp->dont_learn)
1183 md->gbp |= VXLAN_GBP_DONT_LEARN;
1184
1185 if (gbp->policy_applied)
1186 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1187
1188 /* In flow-based mode, GBP is carried in dst_metadata */
1189 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1190 skb->mark = md->gbp;
1191 out:
1192 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1193 }
1194
1195 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1196 struct vxlan_sock *vs,
1197 struct sk_buff *skb)
1198 {
1199 union vxlan_addr saddr;
1200
1201 skb_reset_mac_header(skb);
1202 skb->protocol = eth_type_trans(skb, vxlan->dev);
1203 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1204
1205 /* Ignore packet loops (and multicast echo) */
1206 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1207 return false;
1208
1209 /* Get address from the outer IP header */
1210 if (vxlan_get_sk_family(vs) == AF_INET) {
1211 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1212 saddr.sa.sa_family = AF_INET;
1213 #if IS_ENABLED(CONFIG_IPV6)
1214 } else {
1215 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1216 saddr.sa.sa_family = AF_INET6;
1217 #endif
1218 }
1219
1220 if ((vxlan->flags & VXLAN_F_LEARN) &&
1221 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1222 return false;
1223
1224 return true;
1225 }
1226
1227 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1228 struct sk_buff *skb)
1229 {
1230 int err = 0;
1231
1232 if (vxlan_get_sk_family(vs) == AF_INET)
1233 err = IP_ECN_decapsulate(oiph, skb);
1234 #if IS_ENABLED(CONFIG_IPV6)
1235 else
1236 err = IP6_ECN_decapsulate(oiph, skb);
1237 #endif
1238
1239 if (unlikely(err) && log_ecn_error) {
1240 if (vxlan_get_sk_family(vs) == AF_INET)
1241 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1242 &((struct iphdr *)oiph)->saddr,
1243 ((struct iphdr *)oiph)->tos);
1244 else
1245 net_info_ratelimited("non-ECT from %pI6\n",
1246 &((struct ipv6hdr *)oiph)->saddr);
1247 }
1248 return err <= 1;
1249 }
1250
1251 /* Callback from net/ipv4/udp.c to receive packets */
1252 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1253 {
1254 struct pcpu_sw_netstats *stats;
1255 struct vxlan_dev *vxlan;
1256 struct vxlan_sock *vs;
1257 struct vxlanhdr unparsed;
1258 struct vxlan_metadata _md;
1259 struct vxlan_metadata *md = &_md;
1260 void *oiph;
1261
1262 /* Need Vxlan and inner Ethernet header to be present */
1263 if (!pskb_may_pull(skb, VXLAN_HLEN))
1264 return 1;
1265
1266 unparsed = *vxlan_hdr(skb);
1267 /* VNI flag always required to be set */
1268 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1269 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1270 ntohl(vxlan_hdr(skb)->vx_flags),
1271 ntohl(vxlan_hdr(skb)->vx_vni));
1272 /* Return non vxlan pkt */
1273 return 1;
1274 }
1275 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1276 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1277
1278 vs = rcu_dereference_sk_user_data(sk);
1279 if (!vs)
1280 goto drop;
1281
1282 vxlan = vxlan_vs_find_vni(vs, vxlan_vni(vxlan_hdr(skb)->vx_vni));
1283 if (!vxlan)
1284 goto drop;
1285
1286 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB),
1287 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1288 goto drop;
1289
1290 if (vxlan_collect_metadata(vs)) {
1291 __be32 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1292 struct metadata_dst *tun_dst;
1293
1294 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1295 vxlan_vni_to_tun_id(vni), sizeof(*md));
1296
1297 if (!tun_dst)
1298 goto drop;
1299
1300 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1301
1302 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1303 } else {
1304 memset(md, 0, sizeof(*md));
1305 }
1306
1307 /* For backwards compatibility, only allow reserved fields to be
1308 * used by VXLAN extensions if explicitly requested.
1309 */
1310 if (vs->flags & VXLAN_F_REMCSUM_RX)
1311 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1312 goto drop;
1313 if (vs->flags & VXLAN_F_GBP)
1314 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1315
1316 if (unparsed.vx_flags || unparsed.vx_vni) {
1317 /* If there are any unprocessed flags remaining treat
1318 * this as a malformed packet. This behavior diverges from
1319 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1320 * in reserved fields are to be ignored. The approach here
1321 * maintains compatibility with previous stack code, and also
1322 * is more robust and provides a little more security in
1323 * adding extensions to VXLAN.
1324 */
1325 goto drop;
1326 }
1327
1328 if (!vxlan_set_mac(vxlan, vs, skb))
1329 goto drop;
1330
1331 oiph = skb_network_header(skb);
1332 skb_reset_network_header(skb);
1333
1334 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1335 ++vxlan->dev->stats.rx_frame_errors;
1336 ++vxlan->dev->stats.rx_errors;
1337 goto drop;
1338 }
1339
1340 stats = this_cpu_ptr(vxlan->dev->tstats);
1341 u64_stats_update_begin(&stats->syncp);
1342 stats->rx_packets++;
1343 stats->rx_bytes += skb->len;
1344 u64_stats_update_end(&stats->syncp);
1345
1346 gro_cells_receive(&vxlan->gro_cells, skb);
1347 return 0;
1348
1349 drop:
1350 /* Consume bad packet */
1351 kfree_skb(skb);
1352 return 0;
1353 }
1354
1355 static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1356 {
1357 struct vxlan_dev *vxlan = netdev_priv(dev);
1358 struct arphdr *parp;
1359 u8 *arpptr, *sha;
1360 __be32 sip, tip;
1361 struct neighbour *n;
1362
1363 if (dev->flags & IFF_NOARP)
1364 goto out;
1365
1366 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1367 dev->stats.tx_dropped++;
1368 goto out;
1369 }
1370 parp = arp_hdr(skb);
1371
1372 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1373 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1374 parp->ar_pro != htons(ETH_P_IP) ||
1375 parp->ar_op != htons(ARPOP_REQUEST) ||
1376 parp->ar_hln != dev->addr_len ||
1377 parp->ar_pln != 4)
1378 goto out;
1379 arpptr = (u8 *)parp + sizeof(struct arphdr);
1380 sha = arpptr;
1381 arpptr += dev->addr_len; /* sha */
1382 memcpy(&sip, arpptr, sizeof(sip));
1383 arpptr += sizeof(sip);
1384 arpptr += dev->addr_len; /* tha */
1385 memcpy(&tip, arpptr, sizeof(tip));
1386
1387 if (ipv4_is_loopback(tip) ||
1388 ipv4_is_multicast(tip))
1389 goto out;
1390
1391 n = neigh_lookup(&arp_tbl, &tip, dev);
1392
1393 if (n) {
1394 struct vxlan_fdb *f;
1395 struct sk_buff *reply;
1396
1397 if (!(n->nud_state & NUD_CONNECTED)) {
1398 neigh_release(n);
1399 goto out;
1400 }
1401
1402 f = vxlan_find_mac(vxlan, n->ha);
1403 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1404 /* bridge-local neighbor */
1405 neigh_release(n);
1406 goto out;
1407 }
1408
1409 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1410 n->ha, sha);
1411
1412 neigh_release(n);
1413
1414 if (reply == NULL)
1415 goto out;
1416
1417 skb_reset_mac_header(reply);
1418 __skb_pull(reply, skb_network_offset(reply));
1419 reply->ip_summed = CHECKSUM_UNNECESSARY;
1420 reply->pkt_type = PACKET_HOST;
1421
1422 if (netif_rx_ni(reply) == NET_RX_DROP)
1423 dev->stats.rx_dropped++;
1424 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1425 union vxlan_addr ipa = {
1426 .sin.sin_addr.s_addr = tip,
1427 .sin.sin_family = AF_INET,
1428 };
1429
1430 vxlan_ip_miss(dev, &ipa);
1431 }
1432 out:
1433 consume_skb(skb);
1434 return NETDEV_TX_OK;
1435 }
1436
1437 #if IS_ENABLED(CONFIG_IPV6)
1438 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1439 struct neighbour *n, bool isrouter)
1440 {
1441 struct net_device *dev = request->dev;
1442 struct sk_buff *reply;
1443 struct nd_msg *ns, *na;
1444 struct ipv6hdr *pip6;
1445 u8 *daddr;
1446 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1447 int ns_olen;
1448 int i, len;
1449
1450 if (dev == NULL)
1451 return NULL;
1452
1453 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1454 sizeof(*na) + na_olen + dev->needed_tailroom;
1455 reply = alloc_skb(len, GFP_ATOMIC);
1456 if (reply == NULL)
1457 return NULL;
1458
1459 reply->protocol = htons(ETH_P_IPV6);
1460 reply->dev = dev;
1461 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1462 skb_push(reply, sizeof(struct ethhdr));
1463 skb_reset_mac_header(reply);
1464
1465 ns = (struct nd_msg *)skb_transport_header(request);
1466
1467 daddr = eth_hdr(request)->h_source;
1468 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1469 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1470 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1471 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1472 break;
1473 }
1474 }
1475
1476 /* Ethernet header */
1477 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1478 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1479 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1480 reply->protocol = htons(ETH_P_IPV6);
1481
1482 skb_pull(reply, sizeof(struct ethhdr));
1483 skb_reset_network_header(reply);
1484 skb_put(reply, sizeof(struct ipv6hdr));
1485
1486 /* IPv6 header */
1487
1488 pip6 = ipv6_hdr(reply);
1489 memset(pip6, 0, sizeof(struct ipv6hdr));
1490 pip6->version = 6;
1491 pip6->priority = ipv6_hdr(request)->priority;
1492 pip6->nexthdr = IPPROTO_ICMPV6;
1493 pip6->hop_limit = 255;
1494 pip6->daddr = ipv6_hdr(request)->saddr;
1495 pip6->saddr = *(struct in6_addr *)n->primary_key;
1496
1497 skb_pull(reply, sizeof(struct ipv6hdr));
1498 skb_reset_transport_header(reply);
1499
1500 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1501
1502 /* Neighbor Advertisement */
1503 memset(na, 0, sizeof(*na)+na_olen);
1504 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1505 na->icmph.icmp6_router = isrouter;
1506 na->icmph.icmp6_override = 1;
1507 na->icmph.icmp6_solicited = 1;
1508 na->target = ns->target;
1509 ether_addr_copy(&na->opt[2], n->ha);
1510 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1511 na->opt[1] = na_olen >> 3;
1512
1513 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1514 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1515 csum_partial(na, sizeof(*na)+na_olen, 0));
1516
1517 pip6->payload_len = htons(sizeof(*na)+na_olen);
1518
1519 skb_push(reply, sizeof(struct ipv6hdr));
1520
1521 reply->ip_summed = CHECKSUM_UNNECESSARY;
1522
1523 return reply;
1524 }
1525
1526 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1527 {
1528 struct vxlan_dev *vxlan = netdev_priv(dev);
1529 struct nd_msg *msg;
1530 const struct ipv6hdr *iphdr;
1531 const struct in6_addr *saddr, *daddr;
1532 struct neighbour *n;
1533 struct inet6_dev *in6_dev;
1534
1535 in6_dev = __in6_dev_get(dev);
1536 if (!in6_dev)
1537 goto out;
1538
1539 iphdr = ipv6_hdr(skb);
1540 saddr = &iphdr->saddr;
1541 daddr = &iphdr->daddr;
1542
1543 msg = (struct nd_msg *)skb_transport_header(skb);
1544 if (msg->icmph.icmp6_code != 0 ||
1545 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1546 goto out;
1547
1548 if (ipv6_addr_loopback(daddr) ||
1549 ipv6_addr_is_multicast(&msg->target))
1550 goto out;
1551
1552 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
1553
1554 if (n) {
1555 struct vxlan_fdb *f;
1556 struct sk_buff *reply;
1557
1558 if (!(n->nud_state & NUD_CONNECTED)) {
1559 neigh_release(n);
1560 goto out;
1561 }
1562
1563 f = vxlan_find_mac(vxlan, n->ha);
1564 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1565 /* bridge-local neighbor */
1566 neigh_release(n);
1567 goto out;
1568 }
1569
1570 reply = vxlan_na_create(skb, n,
1571 !!(f ? f->flags & NTF_ROUTER : 0));
1572
1573 neigh_release(n);
1574
1575 if (reply == NULL)
1576 goto out;
1577
1578 if (netif_rx_ni(reply) == NET_RX_DROP)
1579 dev->stats.rx_dropped++;
1580
1581 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1582 union vxlan_addr ipa = {
1583 .sin6.sin6_addr = msg->target,
1584 .sin6.sin6_family = AF_INET6,
1585 };
1586
1587 vxlan_ip_miss(dev, &ipa);
1588 }
1589
1590 out:
1591 consume_skb(skb);
1592 return NETDEV_TX_OK;
1593 }
1594 #endif
1595
1596 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1597 {
1598 struct vxlan_dev *vxlan = netdev_priv(dev);
1599 struct neighbour *n;
1600
1601 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1602 return false;
1603
1604 n = NULL;
1605 switch (ntohs(eth_hdr(skb)->h_proto)) {
1606 case ETH_P_IP:
1607 {
1608 struct iphdr *pip;
1609
1610 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1611 return false;
1612 pip = ip_hdr(skb);
1613 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1614 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1615 union vxlan_addr ipa = {
1616 .sin.sin_addr.s_addr = pip->daddr,
1617 .sin.sin_family = AF_INET,
1618 };
1619
1620 vxlan_ip_miss(dev, &ipa);
1621 return false;
1622 }
1623
1624 break;
1625 }
1626 #if IS_ENABLED(CONFIG_IPV6)
1627 case ETH_P_IPV6:
1628 {
1629 struct ipv6hdr *pip6;
1630
1631 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1632 return false;
1633 pip6 = ipv6_hdr(skb);
1634 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1635 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1636 union vxlan_addr ipa = {
1637 .sin6.sin6_addr = pip6->daddr,
1638 .sin6.sin6_family = AF_INET6,
1639 };
1640
1641 vxlan_ip_miss(dev, &ipa);
1642 return false;
1643 }
1644
1645 break;
1646 }
1647 #endif
1648 default:
1649 return false;
1650 }
1651
1652 if (n) {
1653 bool diff;
1654
1655 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1656 if (diff) {
1657 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1658 dev->addr_len);
1659 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1660 }
1661 neigh_release(n);
1662 return diff;
1663 }
1664
1665 return false;
1666 }
1667
1668 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
1669 struct vxlan_metadata *md)
1670 {
1671 struct vxlanhdr_gbp *gbp;
1672
1673 if (!md->gbp)
1674 return;
1675
1676 gbp = (struct vxlanhdr_gbp *)vxh;
1677 vxh->vx_flags |= VXLAN_HF_GBP;
1678
1679 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1680 gbp->dont_learn = 1;
1681
1682 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1683 gbp->policy_applied = 1;
1684
1685 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1686 }
1687
1688 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1689 int iphdr_len, __be32 vni,
1690 struct vxlan_metadata *md, u32 vxflags,
1691 bool udp_sum)
1692 {
1693 struct vxlanhdr *vxh;
1694 int min_headroom;
1695 int err;
1696 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1697
1698 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
1699 skb->ip_summed == CHECKSUM_PARTIAL) {
1700 int csum_start = skb_checksum_start_offset(skb);
1701
1702 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1703 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1704 (skb->csum_offset == offsetof(struct udphdr, check) ||
1705 skb->csum_offset == offsetof(struct tcphdr, check)))
1706 type |= SKB_GSO_TUNNEL_REMCSUM;
1707 }
1708
1709 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1710 + VXLAN_HLEN + iphdr_len
1711 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
1712
1713 /* Need space for new headers (invalidates iph ptr) */
1714 err = skb_cow_head(skb, min_headroom);
1715 if (unlikely(err)) {
1716 kfree_skb(skb);
1717 return err;
1718 }
1719
1720 skb = vlan_hwaccel_push_inside(skb);
1721 if (WARN_ON(!skb))
1722 return -ENOMEM;
1723
1724 skb = iptunnel_handle_offloads(skb, type);
1725 if (IS_ERR(skb))
1726 return PTR_ERR(skb);
1727
1728 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1729 vxh->vx_flags = VXLAN_HF_VNI;
1730 vxh->vx_vni = vxlan_vni_field(vni);
1731
1732 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1733 unsigned int start;
1734
1735 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1736 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1737 vxh->vx_flags |= VXLAN_HF_RCO;
1738
1739 if (!skb_is_gso(skb)) {
1740 skb->ip_summed = CHECKSUM_NONE;
1741 skb->encapsulation = 0;
1742 }
1743 }
1744
1745 if (vxflags & VXLAN_F_GBP)
1746 vxlan_build_gbp_hdr(vxh, vxflags, md);
1747
1748 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1749 return 0;
1750 }
1751
1752 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan,
1753 struct sk_buff *skb, int oif, u8 tos,
1754 __be32 daddr, __be32 *saddr,
1755 struct dst_cache *dst_cache,
1756 const struct ip_tunnel_info *info)
1757 {
1758 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1759 struct rtable *rt = NULL;
1760 struct flowi4 fl4;
1761
1762 if (tos && !info)
1763 use_cache = false;
1764 if (use_cache) {
1765 rt = dst_cache_get_ip4(dst_cache, saddr);
1766 if (rt)
1767 return rt;
1768 }
1769
1770 memset(&fl4, 0, sizeof(fl4));
1771 fl4.flowi4_oif = oif;
1772 fl4.flowi4_tos = RT_TOS(tos);
1773 fl4.flowi4_mark = skb->mark;
1774 fl4.flowi4_proto = IPPROTO_UDP;
1775 fl4.daddr = daddr;
1776 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
1777
1778 rt = ip_route_output_key(vxlan->net, &fl4);
1779 if (!IS_ERR(rt)) {
1780 *saddr = fl4.saddr;
1781 if (use_cache)
1782 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
1783 }
1784 return rt;
1785 }
1786
1787 #if IS_ENABLED(CONFIG_IPV6)
1788 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1789 struct sk_buff *skb, int oif, u8 tos,
1790 __be32 label,
1791 const struct in6_addr *daddr,
1792 struct in6_addr *saddr,
1793 struct dst_cache *dst_cache,
1794 const struct ip_tunnel_info *info)
1795 {
1796 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1797 struct dst_entry *ndst;
1798 struct flowi6 fl6;
1799 int err;
1800
1801 if (tos && !info)
1802 use_cache = false;
1803 if (use_cache) {
1804 ndst = dst_cache_get_ip6(dst_cache, saddr);
1805 if (ndst)
1806 return ndst;
1807 }
1808
1809 memset(&fl6, 0, sizeof(fl6));
1810 fl6.flowi6_oif = oif;
1811 fl6.daddr = *daddr;
1812 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
1813 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
1814 fl6.flowi6_mark = skb->mark;
1815 fl6.flowi6_proto = IPPROTO_UDP;
1816
1817 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1818 vxlan->vn6_sock->sock->sk,
1819 &ndst, &fl6);
1820 if (err < 0)
1821 return ERR_PTR(err);
1822
1823 *saddr = fl6.saddr;
1824 if (use_cache)
1825 dst_cache_set_ip6(dst_cache, ndst, saddr);
1826 return ndst;
1827 }
1828 #endif
1829
1830 /* Bypass encapsulation if the destination is local */
1831 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1832 struct vxlan_dev *dst_vxlan)
1833 {
1834 struct pcpu_sw_netstats *tx_stats, *rx_stats;
1835 union vxlan_addr loopback;
1836 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
1837 struct net_device *dev = skb->dev;
1838 int len = skb->len;
1839
1840 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1841 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1842 skb->pkt_type = PACKET_HOST;
1843 skb->encapsulation = 0;
1844 skb->dev = dst_vxlan->dev;
1845 __skb_pull(skb, skb_network_offset(skb));
1846
1847 if (remote_ip->sa.sa_family == AF_INET) {
1848 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1849 loopback.sa.sa_family = AF_INET;
1850 #if IS_ENABLED(CONFIG_IPV6)
1851 } else {
1852 loopback.sin6.sin6_addr = in6addr_loopback;
1853 loopback.sa.sa_family = AF_INET6;
1854 #endif
1855 }
1856
1857 if (dst_vxlan->flags & VXLAN_F_LEARN)
1858 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
1859
1860 u64_stats_update_begin(&tx_stats->syncp);
1861 tx_stats->tx_packets++;
1862 tx_stats->tx_bytes += len;
1863 u64_stats_update_end(&tx_stats->syncp);
1864
1865 if (netif_rx(skb) == NET_RX_SUCCESS) {
1866 u64_stats_update_begin(&rx_stats->syncp);
1867 rx_stats->rx_packets++;
1868 rx_stats->rx_bytes += len;
1869 u64_stats_update_end(&rx_stats->syncp);
1870 } else {
1871 dev->stats.rx_dropped++;
1872 }
1873 }
1874
1875 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1876 struct vxlan_rdst *rdst, bool did_rsc)
1877 {
1878 struct dst_cache *dst_cache;
1879 struct ip_tunnel_info *info;
1880 struct vxlan_dev *vxlan = netdev_priv(dev);
1881 struct sock *sk;
1882 struct rtable *rt = NULL;
1883 const struct iphdr *old_iph;
1884 union vxlan_addr *dst;
1885 union vxlan_addr remote_ip;
1886 struct vxlan_metadata _md;
1887 struct vxlan_metadata *md = &_md;
1888 __be16 src_port = 0, dst_port;
1889 __be32 vni, label;
1890 __be16 df = 0;
1891 __u8 tos, ttl;
1892 int err;
1893 u32 flags = vxlan->flags;
1894 bool udp_sum = false;
1895 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
1896
1897 info = skb_tunnel_info(skb);
1898
1899 if (rdst) {
1900 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
1901 vni = rdst->remote_vni;
1902 dst = &rdst->remote_ip;
1903 dst_cache = &rdst->dst_cache;
1904 } else {
1905 if (!info) {
1906 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1907 dev->name);
1908 goto drop;
1909 }
1910 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
1911 vni = vxlan_tun_id_to_vni(info->key.tun_id);
1912 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
1913 if (remote_ip.sa.sa_family == AF_INET)
1914 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1915 else
1916 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
1917 dst = &remote_ip;
1918 dst_cache = &info->dst_cache;
1919 }
1920
1921 if (vxlan_addr_any(dst)) {
1922 if (did_rsc) {
1923 /* short-circuited back to local bridge */
1924 vxlan_encap_bypass(skb, vxlan, vxlan);
1925 return;
1926 }
1927 goto drop;
1928 }
1929
1930 old_iph = ip_hdr(skb);
1931
1932 ttl = vxlan->cfg.ttl;
1933 if (!ttl && vxlan_addr_multicast(dst))
1934 ttl = 1;
1935
1936 tos = vxlan->cfg.tos;
1937 if (tos == 1)
1938 tos = ip_tunnel_get_dsfield(old_iph, skb);
1939
1940 label = vxlan->cfg.label;
1941 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1942 vxlan->cfg.port_max, true);
1943
1944 if (info) {
1945 ttl = info->key.ttl;
1946 tos = info->key.tos;
1947 label = info->key.label;
1948 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
1949
1950 if (info->options_len)
1951 md = ip_tunnel_info_opts(info);
1952 } else {
1953 md->gbp = skb->mark;
1954 }
1955
1956 if (dst->sa.sa_family == AF_INET) {
1957 __be32 saddr;
1958
1959 if (!vxlan->vn4_sock)
1960 goto drop;
1961 sk = vxlan->vn4_sock->sock->sk;
1962
1963 rt = vxlan_get_route(vxlan, skb,
1964 rdst ? rdst->remote_ifindex : 0, tos,
1965 dst->sin.sin_addr.s_addr, &saddr,
1966 dst_cache, info);
1967 if (IS_ERR(rt)) {
1968 netdev_dbg(dev, "no route to %pI4\n",
1969 &dst->sin.sin_addr.s_addr);
1970 dev->stats.tx_carrier_errors++;
1971 goto tx_error;
1972 }
1973
1974 if (rt->dst.dev == dev) {
1975 netdev_dbg(dev, "circular route to %pI4\n",
1976 &dst->sin.sin_addr.s_addr);
1977 dev->stats.collisions++;
1978 goto rt_tx_error;
1979 }
1980
1981 /* Bypass encapsulation if the destination is local */
1982 if (rt->rt_flags & RTCF_LOCAL &&
1983 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1984 struct vxlan_dev *dst_vxlan;
1985
1986 ip_rt_put(rt);
1987 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
1988 dst->sa.sa_family, dst_port,
1989 vxlan->flags);
1990 if (!dst_vxlan)
1991 goto tx_error;
1992 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1993 return;
1994 }
1995
1996 if (!info)
1997 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
1998 else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
1999 df = htons(IP_DF);
2000
2001 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2002 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2003 err = vxlan_build_skb(skb, &rt->dst, sizeof(struct iphdr),
2004 vni, md, flags, udp_sum);
2005 if (err < 0)
2006 goto xmit_tx_error;
2007
2008 udp_tunnel_xmit_skb(rt, sk, skb, saddr,
2009 dst->sin.sin_addr.s_addr, tos, ttl, df,
2010 src_port, dst_port, xnet, !udp_sum);
2011 #if IS_ENABLED(CONFIG_IPV6)
2012 } else {
2013 struct dst_entry *ndst;
2014 struct in6_addr saddr;
2015 u32 rt6i_flags;
2016
2017 if (!vxlan->vn6_sock)
2018 goto drop;
2019 sk = vxlan->vn6_sock->sock->sk;
2020
2021 ndst = vxlan6_get_route(vxlan, skb,
2022 rdst ? rdst->remote_ifindex : 0, tos,
2023 label, &dst->sin6.sin6_addr, &saddr,
2024 dst_cache, info);
2025 if (IS_ERR(ndst)) {
2026 netdev_dbg(dev, "no route to %pI6\n",
2027 &dst->sin6.sin6_addr);
2028 dev->stats.tx_carrier_errors++;
2029 goto tx_error;
2030 }
2031
2032 if (ndst->dev == dev) {
2033 netdev_dbg(dev, "circular route to %pI6\n",
2034 &dst->sin6.sin6_addr);
2035 dst_release(ndst);
2036 dev->stats.collisions++;
2037 goto tx_error;
2038 }
2039
2040 /* Bypass encapsulation if the destination is local */
2041 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2042 if (rt6i_flags & RTF_LOCAL &&
2043 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2044 struct vxlan_dev *dst_vxlan;
2045
2046 dst_release(ndst);
2047 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
2048 dst->sa.sa_family, dst_port,
2049 vxlan->flags);
2050 if (!dst_vxlan)
2051 goto tx_error;
2052 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2053 return;
2054 }
2055
2056 if (!info)
2057 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2058
2059 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2060 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2061 skb_scrub_packet(skb, xnet);
2062 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2063 vni, md, flags, udp_sum);
2064 if (err < 0) {
2065 dst_release(ndst);
2066 return;
2067 }
2068 udp_tunnel6_xmit_skb(ndst, sk, skb, dev,
2069 &saddr, &dst->sin6.sin6_addr, tos, ttl,
2070 label, src_port, dst_port, !udp_sum);
2071 #endif
2072 }
2073
2074 return;
2075
2076 drop:
2077 dev->stats.tx_dropped++;
2078 goto tx_free;
2079
2080 xmit_tx_error:
2081 /* skb is already freed. */
2082 skb = NULL;
2083 rt_tx_error:
2084 ip_rt_put(rt);
2085 tx_error:
2086 dev->stats.tx_errors++;
2087 tx_free:
2088 dev_kfree_skb(skb);
2089 }
2090
2091 /* Transmit local packets over Vxlan
2092 *
2093 * Outer IP header inherits ECN and DF from inner header.
2094 * Outer UDP destination is the VXLAN assigned port.
2095 * source port is based on hash of flow
2096 */
2097 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2098 {
2099 struct vxlan_dev *vxlan = netdev_priv(dev);
2100 const struct ip_tunnel_info *info;
2101 struct ethhdr *eth;
2102 bool did_rsc = false;
2103 struct vxlan_rdst *rdst, *fdst = NULL;
2104 struct vxlan_fdb *f;
2105
2106 info = skb_tunnel_info(skb);
2107
2108 skb_reset_mac_header(skb);
2109 eth = eth_hdr(skb);
2110
2111 if ((vxlan->flags & VXLAN_F_PROXY)) {
2112 if (ntohs(eth->h_proto) == ETH_P_ARP)
2113 return arp_reduce(dev, skb);
2114 #if IS_ENABLED(CONFIG_IPV6)
2115 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2116 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2117 + sizeof(struct nd_msg)) &&
2118 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2119 struct nd_msg *msg;
2120
2121 msg = (struct nd_msg *)skb_transport_header(skb);
2122 if (msg->icmph.icmp6_code == 0 &&
2123 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2124 return neigh_reduce(dev, skb);
2125 }
2126 eth = eth_hdr(skb);
2127 #endif
2128 }
2129
2130 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
2131 if (info && info->mode & IP_TUNNEL_INFO_TX)
2132 vxlan_xmit_one(skb, dev, NULL, false);
2133 else
2134 kfree_skb(skb);
2135 return NETDEV_TX_OK;
2136 }
2137
2138 f = vxlan_find_mac(vxlan, eth->h_dest);
2139 did_rsc = false;
2140
2141 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
2142 (ntohs(eth->h_proto) == ETH_P_IP ||
2143 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2144 did_rsc = route_shortcircuit(dev, skb);
2145 if (did_rsc)
2146 f = vxlan_find_mac(vxlan, eth->h_dest);
2147 }
2148
2149 if (f == NULL) {
2150 f = vxlan_find_mac(vxlan, all_zeros_mac);
2151 if (f == NULL) {
2152 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2153 !is_multicast_ether_addr(eth->h_dest))
2154 vxlan_fdb_miss(vxlan, eth->h_dest);
2155
2156 dev->stats.tx_dropped++;
2157 kfree_skb(skb);
2158 return NETDEV_TX_OK;
2159 }
2160 }
2161
2162 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2163 struct sk_buff *skb1;
2164
2165 if (!fdst) {
2166 fdst = rdst;
2167 continue;
2168 }
2169 skb1 = skb_clone(skb, GFP_ATOMIC);
2170 if (skb1)
2171 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
2172 }
2173
2174 if (fdst)
2175 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2176 else
2177 kfree_skb(skb);
2178 return NETDEV_TX_OK;
2179 }
2180
2181 /* Walk the forwarding table and purge stale entries */
2182 static void vxlan_cleanup(unsigned long arg)
2183 {
2184 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2185 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2186 unsigned int h;
2187
2188 if (!netif_running(vxlan->dev))
2189 return;
2190
2191 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2192 struct hlist_node *p, *n;
2193
2194 spin_lock_bh(&vxlan->hash_lock);
2195 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2196 struct vxlan_fdb *f
2197 = container_of(p, struct vxlan_fdb, hlist);
2198 unsigned long timeout;
2199
2200 if (f->state & NUD_PERMANENT)
2201 continue;
2202
2203 timeout = f->used + vxlan->cfg.age_interval * HZ;
2204 if (time_before_eq(timeout, jiffies)) {
2205 netdev_dbg(vxlan->dev,
2206 "garbage collect %pM\n",
2207 f->eth_addr);
2208 f->state = NUD_STALE;
2209 vxlan_fdb_destroy(vxlan, f);
2210 } else if (time_before(timeout, next_timer))
2211 next_timer = timeout;
2212 }
2213 spin_unlock_bh(&vxlan->hash_lock);
2214 }
2215
2216 mod_timer(&vxlan->age_timer, next_timer);
2217 }
2218
2219 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2220 {
2221 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2222 __be32 vni = vxlan->default_dst.remote_vni;
2223
2224 spin_lock(&vn->sock_lock);
2225 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
2226 spin_unlock(&vn->sock_lock);
2227 }
2228
2229 /* Setup stats when device is created */
2230 static int vxlan_init(struct net_device *dev)
2231 {
2232 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2233 if (!dev->tstats)
2234 return -ENOMEM;
2235
2236 return 0;
2237 }
2238
2239 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
2240 {
2241 struct vxlan_fdb *f;
2242
2243 spin_lock_bh(&vxlan->hash_lock);
2244 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2245 if (f)
2246 vxlan_fdb_destroy(vxlan, f);
2247 spin_unlock_bh(&vxlan->hash_lock);
2248 }
2249
2250 static void vxlan_uninit(struct net_device *dev)
2251 {
2252 struct vxlan_dev *vxlan = netdev_priv(dev);
2253
2254 vxlan_fdb_delete_default(vxlan);
2255
2256 free_percpu(dev->tstats);
2257 }
2258
2259 /* Start ageing timer and join group when device is brought up */
2260 static int vxlan_open(struct net_device *dev)
2261 {
2262 struct vxlan_dev *vxlan = netdev_priv(dev);
2263 int ret;
2264
2265 ret = vxlan_sock_add(vxlan);
2266 if (ret < 0)
2267 return ret;
2268
2269 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
2270 ret = vxlan_igmp_join(vxlan);
2271 if (ret == -EADDRINUSE)
2272 ret = 0;
2273 if (ret) {
2274 vxlan_sock_release(vxlan);
2275 return ret;
2276 }
2277 }
2278
2279 if (vxlan->cfg.age_interval)
2280 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2281
2282 return ret;
2283 }
2284
2285 /* Purge the forwarding table */
2286 static void vxlan_flush(struct vxlan_dev *vxlan)
2287 {
2288 unsigned int h;
2289
2290 spin_lock_bh(&vxlan->hash_lock);
2291 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2292 struct hlist_node *p, *n;
2293 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2294 struct vxlan_fdb *f
2295 = container_of(p, struct vxlan_fdb, hlist);
2296 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2297 if (!is_zero_ether_addr(f->eth_addr))
2298 vxlan_fdb_destroy(vxlan, f);
2299 }
2300 }
2301 spin_unlock_bh(&vxlan->hash_lock);
2302 }
2303
2304 /* Cleanup timer and forwarding table on shutdown */
2305 static int vxlan_stop(struct net_device *dev)
2306 {
2307 struct vxlan_dev *vxlan = netdev_priv(dev);
2308 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2309 int ret = 0;
2310
2311 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2312 !vxlan_group_used(vn, vxlan))
2313 ret = vxlan_igmp_leave(vxlan);
2314
2315 del_timer_sync(&vxlan->age_timer);
2316
2317 vxlan_flush(vxlan);
2318 vxlan_sock_release(vxlan);
2319
2320 return ret;
2321 }
2322
2323 /* Stub, nothing needs to be done. */
2324 static void vxlan_set_multicast_list(struct net_device *dev)
2325 {
2326 }
2327
2328 static int __vxlan_change_mtu(struct net_device *dev,
2329 struct net_device *lowerdev,
2330 struct vxlan_rdst *dst, int new_mtu, bool strict)
2331 {
2332 int max_mtu = IP_MAX_MTU;
2333
2334 if (lowerdev)
2335 max_mtu = lowerdev->mtu;
2336
2337 if (dst->remote_ip.sa.sa_family == AF_INET6)
2338 max_mtu -= VXLAN6_HEADROOM;
2339 else
2340 max_mtu -= VXLAN_HEADROOM;
2341
2342 if (new_mtu < 68)
2343 return -EINVAL;
2344
2345 if (new_mtu > max_mtu) {
2346 if (strict)
2347 return -EINVAL;
2348
2349 new_mtu = max_mtu;
2350 }
2351
2352 dev->mtu = new_mtu;
2353 return 0;
2354 }
2355
2356 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2357 {
2358 struct vxlan_dev *vxlan = netdev_priv(dev);
2359 struct vxlan_rdst *dst = &vxlan->default_dst;
2360 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2361 dst->remote_ifindex);
2362 return __vxlan_change_mtu(dev, lowerdev, dst, new_mtu, true);
2363 }
2364
2365 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2366 {
2367 struct vxlan_dev *vxlan = netdev_priv(dev);
2368 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2369 __be16 sport, dport;
2370
2371 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2372 vxlan->cfg.port_max, true);
2373 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2374
2375 if (ip_tunnel_info_af(info) == AF_INET) {
2376 struct rtable *rt;
2377
2378 if (!vxlan->vn4_sock)
2379 return -EINVAL;
2380 rt = vxlan_get_route(vxlan, skb, 0, info->key.tos,
2381 info->key.u.ipv4.dst,
2382 &info->key.u.ipv4.src, NULL, info);
2383 if (IS_ERR(rt))
2384 return PTR_ERR(rt);
2385 ip_rt_put(rt);
2386 } else {
2387 #if IS_ENABLED(CONFIG_IPV6)
2388 struct dst_entry *ndst;
2389
2390 if (!vxlan->vn6_sock)
2391 return -EINVAL;
2392 ndst = vxlan6_get_route(vxlan, skb, 0, info->key.tos,
2393 info->key.label, &info->key.u.ipv6.dst,
2394 &info->key.u.ipv6.src, NULL, info);
2395 if (IS_ERR(ndst))
2396 return PTR_ERR(ndst);
2397 dst_release(ndst);
2398 #else /* !CONFIG_IPV6 */
2399 return -EPFNOSUPPORT;
2400 #endif
2401 }
2402 info->key.tp_src = sport;
2403 info->key.tp_dst = dport;
2404 return 0;
2405 }
2406
2407 static const struct net_device_ops vxlan_netdev_ops = {
2408 .ndo_init = vxlan_init,
2409 .ndo_uninit = vxlan_uninit,
2410 .ndo_open = vxlan_open,
2411 .ndo_stop = vxlan_stop,
2412 .ndo_start_xmit = vxlan_xmit,
2413 .ndo_get_stats64 = ip_tunnel_get_stats64,
2414 .ndo_set_rx_mode = vxlan_set_multicast_list,
2415 .ndo_change_mtu = vxlan_change_mtu,
2416 .ndo_validate_addr = eth_validate_addr,
2417 .ndo_set_mac_address = eth_mac_addr,
2418 .ndo_fdb_add = vxlan_fdb_add,
2419 .ndo_fdb_del = vxlan_fdb_delete,
2420 .ndo_fdb_dump = vxlan_fdb_dump,
2421 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2422 };
2423
2424 /* Info for udev, that this is a virtual tunnel endpoint */
2425 static struct device_type vxlan_type = {
2426 .name = "vxlan",
2427 };
2428
2429 /* Calls the ndo_add_vxlan_port of the caller in order to
2430 * supply the listening VXLAN udp ports. Callers are expected
2431 * to implement the ndo_add_vxlan_port.
2432 */
2433 void vxlan_get_rx_port(struct net_device *dev)
2434 {
2435 struct vxlan_sock *vs;
2436 struct net *net = dev_net(dev);
2437 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2438 sa_family_t sa_family;
2439 __be16 port;
2440 unsigned int i;
2441
2442 spin_lock(&vn->sock_lock);
2443 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2444 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2445 port = inet_sk(vs->sock->sk)->inet_sport;
2446 sa_family = vxlan_get_sk_family(vs);
2447 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2448 port);
2449 }
2450 }
2451 spin_unlock(&vn->sock_lock);
2452 }
2453 EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2454
2455 /* Initialize the device structure. */
2456 static void vxlan_setup(struct net_device *dev)
2457 {
2458 struct vxlan_dev *vxlan = netdev_priv(dev);
2459 unsigned int h;
2460
2461 eth_hw_addr_random(dev);
2462 ether_setup(dev);
2463
2464 dev->netdev_ops = &vxlan_netdev_ops;
2465 dev->destructor = free_netdev;
2466 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2467
2468 dev->features |= NETIF_F_LLTX;
2469 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
2470 dev->features |= NETIF_F_RXCSUM;
2471 dev->features |= NETIF_F_GSO_SOFTWARE;
2472
2473 dev->vlan_features = dev->features;
2474 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2475 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2476 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2477 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2478 netif_keep_dst(dev);
2479 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2480 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
2481
2482 INIT_LIST_HEAD(&vxlan->next);
2483 spin_lock_init(&vxlan->hash_lock);
2484
2485 init_timer_deferrable(&vxlan->age_timer);
2486 vxlan->age_timer.function = vxlan_cleanup;
2487 vxlan->age_timer.data = (unsigned long) vxlan;
2488
2489 vxlan->cfg.dst_port = htons(vxlan_port);
2490
2491 vxlan->dev = dev;
2492
2493 gro_cells_init(&vxlan->gro_cells, dev);
2494
2495 for (h = 0; h < FDB_HASH_SIZE; ++h)
2496 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2497 }
2498
2499 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2500 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
2501 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2502 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
2503 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2504 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2505 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
2506 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2507 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2508 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
2509 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2510 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2511 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
2512 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
2513 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2514 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2515 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2516 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
2517 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
2518 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
2519 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2520 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2521 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
2522 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2523 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
2524 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
2525 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
2526 };
2527
2528 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2529 {
2530 if (tb[IFLA_ADDRESS]) {
2531 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2532 pr_debug("invalid link address (not ethernet)\n");
2533 return -EINVAL;
2534 }
2535
2536 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2537 pr_debug("invalid all zero ethernet address\n");
2538 return -EADDRNOTAVAIL;
2539 }
2540 }
2541
2542 if (!data)
2543 return -EINVAL;
2544
2545 if (data[IFLA_VXLAN_ID]) {
2546 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2547 if (id >= VXLAN_VID_MASK)
2548 return -ERANGE;
2549 }
2550
2551 if (data[IFLA_VXLAN_PORT_RANGE]) {
2552 const struct ifla_vxlan_port_range *p
2553 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2554
2555 if (ntohs(p->high) < ntohs(p->low)) {
2556 pr_debug("port range %u .. %u not valid\n",
2557 ntohs(p->low), ntohs(p->high));
2558 return -EINVAL;
2559 }
2560 }
2561
2562 return 0;
2563 }
2564
2565 static void vxlan_get_drvinfo(struct net_device *netdev,
2566 struct ethtool_drvinfo *drvinfo)
2567 {
2568 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2569 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2570 }
2571
2572 static const struct ethtool_ops vxlan_ethtool_ops = {
2573 .get_drvinfo = vxlan_get_drvinfo,
2574 .get_link = ethtool_op_get_link,
2575 };
2576
2577 static void vxlan_del_work(struct work_struct *work)
2578 {
2579 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2580 udp_tunnel_sock_release(vs->sock);
2581 kfree_rcu(vs, rcu);
2582 }
2583
2584 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2585 __be16 port, u32 flags)
2586 {
2587 struct socket *sock;
2588 struct udp_port_cfg udp_conf;
2589 int err;
2590
2591 memset(&udp_conf, 0, sizeof(udp_conf));
2592
2593 if (ipv6) {
2594 udp_conf.family = AF_INET6;
2595 udp_conf.use_udp6_rx_checksums =
2596 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
2597 udp_conf.ipv6_v6only = 1;
2598 } else {
2599 udp_conf.family = AF_INET;
2600 }
2601
2602 udp_conf.local_udp_port = port;
2603
2604 /* Open UDP socket */
2605 err = udp_sock_create(net, &udp_conf, &sock);
2606 if (err < 0)
2607 return ERR_PTR(err);
2608
2609 return sock;
2610 }
2611
2612 /* Create new listen socket if needed */
2613 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2614 __be16 port, u32 flags)
2615 {
2616 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2617 struct vxlan_sock *vs;
2618 struct socket *sock;
2619 unsigned int h;
2620 struct udp_tunnel_sock_cfg tunnel_cfg;
2621
2622 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
2623 if (!vs)
2624 return ERR_PTR(-ENOMEM);
2625
2626 for (h = 0; h < VNI_HASH_SIZE; ++h)
2627 INIT_HLIST_HEAD(&vs->vni_list[h]);
2628
2629 INIT_WORK(&vs->del_work, vxlan_del_work);
2630
2631 sock = vxlan_create_sock(net, ipv6, port, flags);
2632 if (IS_ERR(sock)) {
2633 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2634 PTR_ERR(sock));
2635 kfree(vs);
2636 return ERR_CAST(sock);
2637 }
2638
2639 vs->sock = sock;
2640 atomic_set(&vs->refcnt, 1);
2641 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
2642
2643 /* Initialize the vxlan udp offloads structure */
2644 vs->udp_offloads.port = port;
2645 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2646 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2647
2648 spin_lock(&vn->sock_lock);
2649 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2650 vxlan_notify_add_rx_port(vs);
2651 spin_unlock(&vn->sock_lock);
2652
2653 /* Mark socket as an encapsulation socket. */
2654 tunnel_cfg.sk_user_data = vs;
2655 tunnel_cfg.encap_type = 1;
2656 tunnel_cfg.encap_rcv = vxlan_rcv;
2657 tunnel_cfg.encap_destroy = NULL;
2658
2659 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
2660
2661 return vs;
2662 }
2663
2664 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
2665 {
2666 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2667 struct vxlan_sock *vs = NULL;
2668
2669 if (!vxlan->cfg.no_share) {
2670 spin_lock(&vn->sock_lock);
2671 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2672 vxlan->cfg.dst_port, vxlan->flags);
2673 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
2674 spin_unlock(&vn->sock_lock);
2675 return -EBUSY;
2676 }
2677 spin_unlock(&vn->sock_lock);
2678 }
2679 if (!vs)
2680 vs = vxlan_socket_create(vxlan->net, ipv6,
2681 vxlan->cfg.dst_port, vxlan->flags);
2682 if (IS_ERR(vs))
2683 return PTR_ERR(vs);
2684 #if IS_ENABLED(CONFIG_IPV6)
2685 if (ipv6)
2686 vxlan->vn6_sock = vs;
2687 else
2688 #endif
2689 vxlan->vn4_sock = vs;
2690 vxlan_vs_add_dev(vs, vxlan);
2691 return 0;
2692 }
2693
2694 static int vxlan_sock_add(struct vxlan_dev *vxlan)
2695 {
2696 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2697 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2698 int ret = 0;
2699
2700 vxlan->vn4_sock = NULL;
2701 #if IS_ENABLED(CONFIG_IPV6)
2702 vxlan->vn6_sock = NULL;
2703 if (ipv6 || metadata)
2704 ret = __vxlan_sock_add(vxlan, true);
2705 #endif
2706 if (!ret && (!ipv6 || metadata))
2707 ret = __vxlan_sock_add(vxlan, false);
2708 if (ret < 0)
2709 vxlan_sock_release(vxlan);
2710 return ret;
2711 }
2712
2713 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2714 struct vxlan_config *conf)
2715 {
2716 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
2717 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
2718 struct vxlan_rdst *dst = &vxlan->default_dst;
2719 unsigned short needed_headroom = ETH_HLEN;
2720 int err;
2721 bool use_ipv6 = false;
2722 __be16 default_port = vxlan->cfg.dst_port;
2723 struct net_device *lowerdev = NULL;
2724
2725 vxlan->net = src_net;
2726
2727 dst->remote_vni = conf->vni;
2728
2729 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
2730
2731 /* Unless IPv6 is explicitly requested, assume IPv4 */
2732 if (!dst->remote_ip.sa.sa_family)
2733 dst->remote_ip.sa.sa_family = AF_INET;
2734
2735 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
2736 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2737 if (!IS_ENABLED(CONFIG_IPV6))
2738 return -EPFNOSUPPORT;
2739 use_ipv6 = true;
2740 vxlan->flags |= VXLAN_F_IPV6;
2741 }
2742
2743 if (conf->label && !use_ipv6) {
2744 pr_info("label only supported in use with IPv6\n");
2745 return -EINVAL;
2746 }
2747
2748 if (conf->remote_ifindex) {
2749 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
2750 dst->remote_ifindex = conf->remote_ifindex;
2751
2752 if (!lowerdev) {
2753 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
2754 return -ENODEV;
2755 }
2756
2757 #if IS_ENABLED(CONFIG_IPV6)
2758 if (use_ipv6) {
2759 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2760 if (idev && idev->cnf.disable_ipv6) {
2761 pr_info("IPv6 is disabled via sysctl\n");
2762 return -EPERM;
2763 }
2764 }
2765 #endif
2766
2767 if (!conf->mtu)
2768 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2769
2770 needed_headroom = lowerdev->hard_header_len;
2771 }
2772
2773 if (conf->mtu) {
2774 err = __vxlan_change_mtu(dev, lowerdev, dst, conf->mtu, false);
2775 if (err)
2776 return err;
2777 }
2778
2779 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2780 needed_headroom += VXLAN6_HEADROOM;
2781 else
2782 needed_headroom += VXLAN_HEADROOM;
2783 dev->needed_headroom = needed_headroom;
2784
2785 memcpy(&vxlan->cfg, conf, sizeof(*conf));
2786 if (!vxlan->cfg.dst_port)
2787 vxlan->cfg.dst_port = default_port;
2788 vxlan->flags |= conf->flags;
2789
2790 if (!vxlan->cfg.age_interval)
2791 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2792
2793 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2794 if (tmp->cfg.vni == conf->vni &&
2795 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2796 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2797 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2798 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2799 (vxlan->flags & VXLAN_F_RCV_FLAGS))
2800 return -EEXIST;
2801 }
2802
2803 dev->ethtool_ops = &vxlan_ethtool_ops;
2804
2805 /* create an fdb entry for a valid default destination */
2806 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2807 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2808 &vxlan->default_dst.remote_ip,
2809 NUD_REACHABLE|NUD_PERMANENT,
2810 NLM_F_EXCL|NLM_F_CREATE,
2811 vxlan->cfg.dst_port,
2812 vxlan->default_dst.remote_vni,
2813 vxlan->default_dst.remote_ifindex,
2814 NTF_SELF);
2815 if (err)
2816 return err;
2817 }
2818
2819 err = register_netdevice(dev);
2820 if (err) {
2821 vxlan_fdb_delete_default(vxlan);
2822 return err;
2823 }
2824
2825 list_add(&vxlan->next, &vn->vxlan_list);
2826
2827 return 0;
2828 }
2829
2830 struct net_device *vxlan_dev_create(struct net *net, const char *name,
2831 u8 name_assign_type, struct vxlan_config *conf)
2832 {
2833 struct nlattr *tb[IFLA_MAX+1];
2834 struct net_device *dev;
2835 int err;
2836
2837 memset(&tb, 0, sizeof(tb));
2838
2839 dev = rtnl_create_link(net, name, name_assign_type,
2840 &vxlan_link_ops, tb);
2841 if (IS_ERR(dev))
2842 return dev;
2843
2844 err = vxlan_dev_configure(net, dev, conf);
2845 if (err < 0) {
2846 free_netdev(dev);
2847 return ERR_PTR(err);
2848 }
2849
2850 return dev;
2851 }
2852 EXPORT_SYMBOL_GPL(vxlan_dev_create);
2853
2854 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2855 struct nlattr *tb[], struct nlattr *data[])
2856 {
2857 struct vxlan_config conf;
2858 int err;
2859
2860 memset(&conf, 0, sizeof(conf));
2861
2862 if (data[IFLA_VXLAN_ID])
2863 conf.vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
2864
2865 if (data[IFLA_VXLAN_GROUP]) {
2866 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2867 } else if (data[IFLA_VXLAN_GROUP6]) {
2868 if (!IS_ENABLED(CONFIG_IPV6))
2869 return -EPFNOSUPPORT;
2870
2871 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2872 conf.remote_ip.sa.sa_family = AF_INET6;
2873 }
2874
2875 if (data[IFLA_VXLAN_LOCAL]) {
2876 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2877 conf.saddr.sa.sa_family = AF_INET;
2878 } else if (data[IFLA_VXLAN_LOCAL6]) {
2879 if (!IS_ENABLED(CONFIG_IPV6))
2880 return -EPFNOSUPPORT;
2881
2882 /* TODO: respect scope id */
2883 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2884 conf.saddr.sa.sa_family = AF_INET6;
2885 }
2886
2887 if (data[IFLA_VXLAN_LINK])
2888 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2889
2890 if (data[IFLA_VXLAN_TOS])
2891 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2892
2893 if (data[IFLA_VXLAN_TTL])
2894 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2895
2896 if (data[IFLA_VXLAN_LABEL])
2897 conf.label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
2898 IPV6_FLOWLABEL_MASK;
2899
2900 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
2901 conf.flags |= VXLAN_F_LEARN;
2902
2903 if (data[IFLA_VXLAN_AGEING])
2904 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2905
2906 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2907 conf.flags |= VXLAN_F_PROXY;
2908
2909 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2910 conf.flags |= VXLAN_F_RSC;
2911
2912 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2913 conf.flags |= VXLAN_F_L2MISS;
2914
2915 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2916 conf.flags |= VXLAN_F_L3MISS;
2917
2918 if (data[IFLA_VXLAN_LIMIT])
2919 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2920
2921 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
2922 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
2923 conf.flags |= VXLAN_F_COLLECT_METADATA;
2924
2925 if (data[IFLA_VXLAN_PORT_RANGE]) {
2926 const struct ifla_vxlan_port_range *p
2927 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2928 conf.port_min = ntohs(p->low);
2929 conf.port_max = ntohs(p->high);
2930 }
2931
2932 if (data[IFLA_VXLAN_PORT])
2933 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2934
2935 if (data[IFLA_VXLAN_UDP_CSUM] &&
2936 !nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2937 conf.flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
2938
2939 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2940 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2941 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2942
2943 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2944 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2945 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2946
2947 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2948 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
2949 conf.flags |= VXLAN_F_REMCSUM_TX;
2950
2951 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2952 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
2953 conf.flags |= VXLAN_F_REMCSUM_RX;
2954
2955 if (data[IFLA_VXLAN_GBP])
2956 conf.flags |= VXLAN_F_GBP;
2957
2958 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
2959 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
2960
2961 err = vxlan_dev_configure(src_net, dev, &conf);
2962 switch (err) {
2963 case -ENODEV:
2964 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
2965 break;
2966
2967 case -EPERM:
2968 pr_info("IPv6 is disabled via sysctl\n");
2969 break;
2970
2971 case -EEXIST:
2972 pr_info("duplicate VNI %u\n", be32_to_cpu(conf.vni));
2973 break;
2974 }
2975
2976 return err;
2977 }
2978
2979 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2980 {
2981 struct vxlan_dev *vxlan = netdev_priv(dev);
2982 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2983
2984 spin_lock(&vn->sock_lock);
2985 if (!hlist_unhashed(&vxlan->hlist))
2986 hlist_del_rcu(&vxlan->hlist);
2987 spin_unlock(&vn->sock_lock);
2988
2989 gro_cells_destroy(&vxlan->gro_cells);
2990 list_del(&vxlan->next);
2991 unregister_netdevice_queue(dev, head);
2992 }
2993
2994 static size_t vxlan_get_size(const struct net_device *dev)
2995 {
2996
2997 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
2998 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
2999 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
3000 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
3001 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3002 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
3003 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
3004 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
3005 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3006 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3007 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3008 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
3009 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
3010 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3011 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
3012 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
3013 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3014 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3015 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3016 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3017 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3018 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
3019 0;
3020 }
3021
3022 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3023 {
3024 const struct vxlan_dev *vxlan = netdev_priv(dev);
3025 const struct vxlan_rdst *dst = &vxlan->default_dst;
3026 struct ifla_vxlan_port_range ports = {
3027 .low = htons(vxlan->cfg.port_min),
3028 .high = htons(vxlan->cfg.port_max),
3029 };
3030
3031 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
3032 goto nla_put_failure;
3033
3034 if (!vxlan_addr_any(&dst->remote_ip)) {
3035 if (dst->remote_ip.sa.sa_family == AF_INET) {
3036 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3037 dst->remote_ip.sin.sin_addr.s_addr))
3038 goto nla_put_failure;
3039 #if IS_ENABLED(CONFIG_IPV6)
3040 } else {
3041 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3042 &dst->remote_ip.sin6.sin6_addr))
3043 goto nla_put_failure;
3044 #endif
3045 }
3046 }
3047
3048 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
3049 goto nla_put_failure;
3050
3051 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3052 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
3053 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
3054 vxlan->cfg.saddr.sin.sin_addr.s_addr))
3055 goto nla_put_failure;
3056 #if IS_ENABLED(CONFIG_IPV6)
3057 } else {
3058 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
3059 &vxlan->cfg.saddr.sin6.sin6_addr))
3060 goto nla_put_failure;
3061 #endif
3062 }
3063 }
3064
3065 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3066 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
3067 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
3068 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3069 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3070 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3071 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3072 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3073 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3074 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3075 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3076 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
3077 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3078 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
3079 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3080 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3081 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
3082 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
3083 !(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
3084 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3085 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3086 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
3087 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3088 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3089 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3090 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3091 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
3092 goto nla_put_failure;
3093
3094 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3095 goto nla_put_failure;
3096
3097 if (vxlan->flags & VXLAN_F_GBP &&
3098 nla_put_flag(skb, IFLA_VXLAN_GBP))
3099 goto nla_put_failure;
3100
3101 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3102 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3103 goto nla_put_failure;
3104
3105 return 0;
3106
3107 nla_put_failure:
3108 return -EMSGSIZE;
3109 }
3110
3111 static struct net *vxlan_get_link_net(const struct net_device *dev)
3112 {
3113 struct vxlan_dev *vxlan = netdev_priv(dev);
3114
3115 return vxlan->net;
3116 }
3117
3118 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3119 .kind = "vxlan",
3120 .maxtype = IFLA_VXLAN_MAX,
3121 .policy = vxlan_policy,
3122 .priv_size = sizeof(struct vxlan_dev),
3123 .setup = vxlan_setup,
3124 .validate = vxlan_validate,
3125 .newlink = vxlan_newlink,
3126 .dellink = vxlan_dellink,
3127 .get_size = vxlan_get_size,
3128 .fill_info = vxlan_fill_info,
3129 .get_link_net = vxlan_get_link_net,
3130 };
3131
3132 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3133 struct net_device *dev)
3134 {
3135 struct vxlan_dev *vxlan, *next;
3136 LIST_HEAD(list_kill);
3137
3138 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3139 struct vxlan_rdst *dst = &vxlan->default_dst;
3140
3141 /* In case we created vxlan device with carrier
3142 * and we loose the carrier due to module unload
3143 * we also need to remove vxlan device. In other
3144 * cases, it's not necessary and remote_ifindex
3145 * is 0 here, so no matches.
3146 */
3147 if (dst->remote_ifindex == dev->ifindex)
3148 vxlan_dellink(vxlan->dev, &list_kill);
3149 }
3150
3151 unregister_netdevice_many(&list_kill);
3152 }
3153
3154 static int vxlan_lowerdev_event(struct notifier_block *unused,
3155 unsigned long event, void *ptr)
3156 {
3157 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3158 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
3159
3160 if (event == NETDEV_UNREGISTER)
3161 vxlan_handle_lowerdev_unregister(vn, dev);
3162
3163 return NOTIFY_DONE;
3164 }
3165
3166 static struct notifier_block vxlan_notifier_block __read_mostly = {
3167 .notifier_call = vxlan_lowerdev_event,
3168 };
3169
3170 static __net_init int vxlan_init_net(struct net *net)
3171 {
3172 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3173 unsigned int h;
3174
3175 INIT_LIST_HEAD(&vn->vxlan_list);
3176 spin_lock_init(&vn->sock_lock);
3177
3178 for (h = 0; h < PORT_HASH_SIZE; ++h)
3179 INIT_HLIST_HEAD(&vn->sock_list[h]);
3180
3181 return 0;
3182 }
3183
3184 static void __net_exit vxlan_exit_net(struct net *net)
3185 {
3186 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3187 struct vxlan_dev *vxlan, *next;
3188 struct net_device *dev, *aux;
3189 LIST_HEAD(list);
3190
3191 rtnl_lock();
3192 for_each_netdev_safe(net, dev, aux)
3193 if (dev->rtnl_link_ops == &vxlan_link_ops)
3194 unregister_netdevice_queue(dev, &list);
3195
3196 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3197 /* If vxlan->dev is in the same netns, it has already been added
3198 * to the list by the previous loop.
3199 */
3200 if (!net_eq(dev_net(vxlan->dev), net)) {
3201 gro_cells_destroy(&vxlan->gro_cells);
3202 unregister_netdevice_queue(vxlan->dev, &list);
3203 }
3204 }
3205
3206 unregister_netdevice_many(&list);
3207 rtnl_unlock();
3208 }
3209
3210 static struct pernet_operations vxlan_net_ops = {
3211 .init = vxlan_init_net,
3212 .exit = vxlan_exit_net,
3213 .id = &vxlan_net_id,
3214 .size = sizeof(struct vxlan_net),
3215 };
3216
3217 static int __init vxlan_init_module(void)
3218 {
3219 int rc;
3220
3221 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3222 if (!vxlan_wq)
3223 return -ENOMEM;
3224
3225 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3226
3227 rc = register_pernet_subsys(&vxlan_net_ops);
3228 if (rc)
3229 goto out1;
3230
3231 rc = register_netdevice_notifier(&vxlan_notifier_block);
3232 if (rc)
3233 goto out2;
3234
3235 rc = rtnl_link_register(&vxlan_link_ops);
3236 if (rc)
3237 goto out3;
3238
3239 return 0;
3240 out3:
3241 unregister_netdevice_notifier(&vxlan_notifier_block);
3242 out2:
3243 unregister_pernet_subsys(&vxlan_net_ops);
3244 out1:
3245 destroy_workqueue(vxlan_wq);
3246 return rc;
3247 }
3248 late_initcall(vxlan_init_module);
3249
3250 static void __exit vxlan_cleanup_module(void)
3251 {
3252 rtnl_link_unregister(&vxlan_link_ops);
3253 unregister_netdevice_notifier(&vxlan_notifier_block);
3254 destroy_workqueue(vxlan_wq);
3255 unregister_pernet_subsys(&vxlan_net_ops);
3256 /* rcu_barrier() is called by netns */
3257 }
3258 module_exit(vxlan_cleanup_module);
3259
3260 MODULE_LICENSE("GPL");
3261 MODULE_VERSION(VXLAN_VERSION);
3262 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3263 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3264 MODULE_ALIAS_RTNL_LINK("vxlan");
Linux with added FPC-III support