]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - drivers/net/vxlan.c
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next...
[mirror_ubuntu-jammy-kernel.git] / drivers / net / vxlan.c
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * VXLAN: Virtual eXtensible Local Area Network
4 *
5 * Copyright (c) 2012-2013 Vyatta Inc.
6 */
7
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
10 #include <linux/kernel.h>
11 #include <linux/module.h>
12 #include <linux/errno.h>
13 #include <linux/slab.h>
14 #include <linux/udp.h>
15 #include <linux/igmp.h>
16 #include <linux/if_ether.h>
17 #include <linux/ethtool.h>
18 #include <net/arp.h>
19 #include <net/ndisc.h>
20 #include <net/ipv6_stubs.h>
21 #include <net/ip.h>
22 #include <net/icmp.h>
23 #include <net/rtnetlink.h>
24 #include <net/inet_ecn.h>
25 #include <net/net_namespace.h>
26 #include <net/netns/generic.h>
27 #include <net/tun_proto.h>
28 #include <net/vxlan.h>
29
30 #if IS_ENABLED(CONFIG_IPV6)
31 #include <net/ip6_tunnel.h>
32 #include <net/ip6_checksum.h>
33 #endif
34
35 #define VXLAN_VERSION "0.1"
36
37 #define PORT_HASH_BITS 8
38 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
39 #define FDB_AGE_DEFAULT 300 /* 5 min */
40 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
41
42 /* UDP port for VXLAN traffic.
43 * The IANA assigned port is 4789, but the Linux default is 8472
44 * for compatibility with early adopters.
45 */
46 static unsigned short vxlan_port __read_mostly = 8472;
47 module_param_named(udp_port, vxlan_port, ushort, 0444);
48 MODULE_PARM_DESC(udp_port, "Destination UDP port");
49
50 static bool log_ecn_error = true;
51 module_param(log_ecn_error, bool, 0644);
52 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
53
54 static unsigned int vxlan_net_id;
55 static struct rtnl_link_ops vxlan_link_ops;
56
57 static const u8 all_zeros_mac[ETH_ALEN + 2];
58
59 static int vxlan_sock_add(struct vxlan_dev *vxlan);
60
61 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
62
63 /* per-network namespace private data for this module */
64 struct vxlan_net {
65 struct list_head vxlan_list;
66 struct hlist_head sock_list[PORT_HASH_SIZE];
67 spinlock_t sock_lock;
68 };
69
70 /* Forwarding table entry */
71 struct vxlan_fdb {
72 struct hlist_node hlist; /* linked list of entries */
73 struct rcu_head rcu;
74 unsigned long updated; /* jiffies */
75 unsigned long used;
76 struct list_head remotes;
77 u8 eth_addr[ETH_ALEN];
78 u16 state; /* see ndm_state */
79 __be32 vni;
80 u16 flags; /* see ndm_flags and below */
81 };
82
83 #define NTF_VXLAN_ADDED_BY_USER 0x100
84
85 /* salt for hash table */
86 static u32 vxlan_salt __read_mostly;
87
88 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
89 {
90 return vs->flags & VXLAN_F_COLLECT_METADATA ||
91 ip_tunnel_collect_metadata();
92 }
93
94 #if IS_ENABLED(CONFIG_IPV6)
95 static inline
96 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
97 {
98 if (a->sa.sa_family != b->sa.sa_family)
99 return false;
100 if (a->sa.sa_family == AF_INET6)
101 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
102 else
103 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
104 }
105
106 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
107 {
108 if (nla_len(nla) >= sizeof(struct in6_addr)) {
109 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
110 ip->sa.sa_family = AF_INET6;
111 return 0;
112 } else if (nla_len(nla) >= sizeof(__be32)) {
113 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
114 ip->sa.sa_family = AF_INET;
115 return 0;
116 } else {
117 return -EAFNOSUPPORT;
118 }
119 }
120
121 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
122 const union vxlan_addr *ip)
123 {
124 if (ip->sa.sa_family == AF_INET6)
125 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
126 else
127 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
128 }
129
130 #else /* !CONFIG_IPV6 */
131
132 static inline
133 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
134 {
135 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
136 }
137
138 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
139 {
140 if (nla_len(nla) >= sizeof(struct in6_addr)) {
141 return -EAFNOSUPPORT;
142 } else if (nla_len(nla) >= sizeof(__be32)) {
143 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
144 ip->sa.sa_family = AF_INET;
145 return 0;
146 } else {
147 return -EAFNOSUPPORT;
148 }
149 }
150
151 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
152 const union vxlan_addr *ip)
153 {
154 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
155 }
156 #endif
157
158 /* Virtual Network hash table head */
159 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
160 {
161 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
162 }
163
164 /* Socket hash table head */
165 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
166 {
167 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
168
169 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
170 }
171
172 /* First remote destination for a forwarding entry.
173 * Guaranteed to be non-NULL because remotes are never deleted.
174 */
175 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
176 {
177 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
178 }
179
180 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
181 {
182 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
183 }
184
185 /* Find VXLAN socket based on network namespace, address family and UDP port
186 * and enabled unshareable flags.
187 */
188 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
189 __be16 port, u32 flags, int ifindex)
190 {
191 struct vxlan_sock *vs;
192
193 flags &= VXLAN_F_RCV_FLAGS;
194
195 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
196 if (inet_sk(vs->sock->sk)->inet_sport == port &&
197 vxlan_get_sk_family(vs) == family &&
198 vs->flags == flags &&
199 vs->sock->sk->sk_bound_dev_if == ifindex)
200 return vs;
201 }
202 return NULL;
203 }
204
205 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
206 __be32 vni)
207 {
208 struct vxlan_dev_node *node;
209
210 /* For flow based devices, map all packets to VNI 0 */
211 if (vs->flags & VXLAN_F_COLLECT_METADATA)
212 vni = 0;
213
214 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
215 if (node->vxlan->default_dst.remote_vni != vni)
216 continue;
217
218 if (IS_ENABLED(CONFIG_IPV6)) {
219 const struct vxlan_config *cfg = &node->vxlan->cfg;
220
221 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
222 cfg->remote_ifindex != ifindex)
223 continue;
224 }
225
226 return node->vxlan;
227 }
228
229 return NULL;
230 }
231
232 /* Look up VNI in a per net namespace table */
233 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
234 __be32 vni, sa_family_t family,
235 __be16 port, u32 flags)
236 {
237 struct vxlan_sock *vs;
238
239 vs = vxlan_find_sock(net, family, port, flags, ifindex);
240 if (!vs)
241 return NULL;
242
243 return vxlan_vs_find_vni(vs, ifindex, vni);
244 }
245
246 /* Fill in neighbour message in skbuff. */
247 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
248 const struct vxlan_fdb *fdb,
249 u32 portid, u32 seq, int type, unsigned int flags,
250 const struct vxlan_rdst *rdst)
251 {
252 unsigned long now = jiffies;
253 struct nda_cacheinfo ci;
254 struct nlmsghdr *nlh;
255 struct ndmsg *ndm;
256 bool send_ip, send_eth;
257
258 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
259 if (nlh == NULL)
260 return -EMSGSIZE;
261
262 ndm = nlmsg_data(nlh);
263 memset(ndm, 0, sizeof(*ndm));
264
265 send_eth = send_ip = true;
266
267 if (type == RTM_GETNEIGH) {
268 send_ip = !vxlan_addr_any(&rdst->remote_ip);
269 send_eth = !is_zero_ether_addr(fdb->eth_addr);
270 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
271 } else
272 ndm->ndm_family = AF_BRIDGE;
273 ndm->ndm_state = fdb->state;
274 ndm->ndm_ifindex = vxlan->dev->ifindex;
275 ndm->ndm_flags = fdb->flags;
276 if (rdst->offloaded)
277 ndm->ndm_flags |= NTF_OFFLOADED;
278 ndm->ndm_type = RTN_UNICAST;
279
280 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
281 nla_put_s32(skb, NDA_LINK_NETNSID,
282 peernet2id(dev_net(vxlan->dev), vxlan->net)))
283 goto nla_put_failure;
284
285 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
286 goto nla_put_failure;
287
288 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
289 goto nla_put_failure;
290
291 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
292 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
293 goto nla_put_failure;
294 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
295 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
296 goto nla_put_failure;
297 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
298 nla_put_u32(skb, NDA_SRC_VNI,
299 be32_to_cpu(fdb->vni)))
300 goto nla_put_failure;
301 if (rdst->remote_ifindex &&
302 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
303 goto nla_put_failure;
304
305 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
306 ci.ndm_confirmed = 0;
307 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
308 ci.ndm_refcnt = 0;
309
310 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
311 goto nla_put_failure;
312
313 nlmsg_end(skb, nlh);
314 return 0;
315
316 nla_put_failure:
317 nlmsg_cancel(skb, nlh);
318 return -EMSGSIZE;
319 }
320
321 static inline size_t vxlan_nlmsg_size(void)
322 {
323 return NLMSG_ALIGN(sizeof(struct ndmsg))
324 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
325 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
326 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
327 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
328 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
329 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
330 + nla_total_size(sizeof(struct nda_cacheinfo));
331 }
332
333 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
334 struct vxlan_rdst *rd, int type)
335 {
336 struct net *net = dev_net(vxlan->dev);
337 struct sk_buff *skb;
338 int err = -ENOBUFS;
339
340 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
341 if (skb == NULL)
342 goto errout;
343
344 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
345 if (err < 0) {
346 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
347 WARN_ON(err == -EMSGSIZE);
348 kfree_skb(skb);
349 goto errout;
350 }
351
352 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
353 return;
354 errout:
355 if (err < 0)
356 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
357 }
358
359 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
360 const struct vxlan_fdb *fdb,
361 const struct vxlan_rdst *rd,
362 struct netlink_ext_ack *extack,
363 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
364 {
365 fdb_info->info.dev = vxlan->dev;
366 fdb_info->info.extack = extack;
367 fdb_info->remote_ip = rd->remote_ip;
368 fdb_info->remote_port = rd->remote_port;
369 fdb_info->remote_vni = rd->remote_vni;
370 fdb_info->remote_ifindex = rd->remote_ifindex;
371 memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
372 fdb_info->vni = fdb->vni;
373 fdb_info->offloaded = rd->offloaded;
374 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
375 }
376
377 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
378 struct vxlan_fdb *fdb,
379 struct vxlan_rdst *rd,
380 bool adding,
381 struct netlink_ext_ack *extack)
382 {
383 struct switchdev_notifier_vxlan_fdb_info info;
384 enum switchdev_notifier_type notifier_type;
385 int ret;
386
387 if (WARN_ON(!rd))
388 return 0;
389
390 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
391 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
392 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
393 ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
394 &info.info, extack);
395 return notifier_to_errno(ret);
396 }
397
398 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
399 struct vxlan_rdst *rd, int type, bool swdev_notify,
400 struct netlink_ext_ack *extack)
401 {
402 int err;
403
404 if (swdev_notify) {
405 switch (type) {
406 case RTM_NEWNEIGH:
407 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
408 true, extack);
409 if (err)
410 return err;
411 break;
412 case RTM_DELNEIGH:
413 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
414 false, extack);
415 break;
416 }
417 }
418
419 __vxlan_fdb_notify(vxlan, fdb, rd, type);
420 return 0;
421 }
422
423 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
424 {
425 struct vxlan_dev *vxlan = netdev_priv(dev);
426 struct vxlan_fdb f = {
427 .state = NUD_STALE,
428 };
429 struct vxlan_rdst remote = {
430 .remote_ip = *ipa, /* goes to NDA_DST */
431 .remote_vni = cpu_to_be32(VXLAN_N_VID),
432 };
433
434 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
435 }
436
437 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
438 {
439 struct vxlan_fdb f = {
440 .state = NUD_STALE,
441 };
442 struct vxlan_rdst remote = { };
443
444 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
445
446 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
447 }
448
449 /* Hash Ethernet address */
450 static u32 eth_hash(const unsigned char *addr)
451 {
452 u64 value = get_unaligned((u64 *)addr);
453
454 /* only want 6 bytes */
455 #ifdef __BIG_ENDIAN
456 value >>= 16;
457 #else
458 value <<= 16;
459 #endif
460 return hash_64(value, FDB_HASH_BITS);
461 }
462
463 static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
464 {
465 /* use 1 byte of OUI and 3 bytes of NIC */
466 u32 key = get_unaligned((u32 *)(addr + 2));
467
468 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
469 }
470
471 static u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
472 {
473 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
474 return eth_vni_hash(mac, vni);
475 else
476 return eth_hash(mac);
477 }
478
479 /* Hash chain to use given mac address */
480 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
481 const u8 *mac, __be32 vni)
482 {
483 return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
484 }
485
486 /* Look up Ethernet address in forwarding table */
487 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
488 const u8 *mac, __be32 vni)
489 {
490 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
491 struct vxlan_fdb *f;
492
493 hlist_for_each_entry_rcu(f, head, hlist) {
494 if (ether_addr_equal(mac, f->eth_addr)) {
495 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
496 if (vni == f->vni)
497 return f;
498 } else {
499 return f;
500 }
501 }
502 }
503
504 return NULL;
505 }
506
507 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
508 const u8 *mac, __be32 vni)
509 {
510 struct vxlan_fdb *f;
511
512 f = __vxlan_find_mac(vxlan, mac, vni);
513 if (f && f->used != jiffies)
514 f->used = jiffies;
515
516 return f;
517 }
518
519 /* caller should hold vxlan->hash_lock */
520 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
521 union vxlan_addr *ip, __be16 port,
522 __be32 vni, __u32 ifindex)
523 {
524 struct vxlan_rdst *rd;
525
526 list_for_each_entry(rd, &f->remotes, list) {
527 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
528 rd->remote_port == port &&
529 rd->remote_vni == vni &&
530 rd->remote_ifindex == ifindex)
531 return rd;
532 }
533
534 return NULL;
535 }
536
537 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
538 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
539 {
540 struct vxlan_dev *vxlan = netdev_priv(dev);
541 u8 eth_addr[ETH_ALEN + 2] = { 0 };
542 struct vxlan_rdst *rdst;
543 struct vxlan_fdb *f;
544 int rc = 0;
545
546 if (is_multicast_ether_addr(mac) ||
547 is_zero_ether_addr(mac))
548 return -EINVAL;
549
550 ether_addr_copy(eth_addr, mac);
551
552 rcu_read_lock();
553
554 f = __vxlan_find_mac(vxlan, eth_addr, vni);
555 if (!f) {
556 rc = -ENOENT;
557 goto out;
558 }
559
560 rdst = first_remote_rcu(f);
561 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
562
563 out:
564 rcu_read_unlock();
565 return rc;
566 }
567 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
568
569 static int vxlan_fdb_notify_one(struct notifier_block *nb,
570 const struct vxlan_dev *vxlan,
571 const struct vxlan_fdb *f,
572 const struct vxlan_rdst *rdst,
573 struct netlink_ext_ack *extack)
574 {
575 struct switchdev_notifier_vxlan_fdb_info fdb_info;
576 int rc;
577
578 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
579 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
580 &fdb_info);
581 return notifier_to_errno(rc);
582 }
583
584 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
585 struct notifier_block *nb,
586 struct netlink_ext_ack *extack)
587 {
588 struct vxlan_dev *vxlan;
589 struct vxlan_rdst *rdst;
590 struct vxlan_fdb *f;
591 unsigned int h;
592 int rc = 0;
593
594 if (!netif_is_vxlan(dev))
595 return -EINVAL;
596 vxlan = netdev_priv(dev);
597
598 for (h = 0; h < FDB_HASH_SIZE; ++h) {
599 spin_lock_bh(&vxlan->hash_lock[h]);
600 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
601 if (f->vni == vni) {
602 list_for_each_entry(rdst, &f->remotes, list) {
603 rc = vxlan_fdb_notify_one(nb, vxlan,
604 f, rdst,
605 extack);
606 if (rc)
607 goto unlock;
608 }
609 }
610 }
611 spin_unlock_bh(&vxlan->hash_lock[h]);
612 }
613 return 0;
614
615 unlock:
616 spin_unlock_bh(&vxlan->hash_lock[h]);
617 return rc;
618 }
619 EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
620
621 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
622 {
623 struct vxlan_dev *vxlan;
624 struct vxlan_rdst *rdst;
625 struct vxlan_fdb *f;
626 unsigned int h;
627
628 if (!netif_is_vxlan(dev))
629 return;
630 vxlan = netdev_priv(dev);
631
632 for (h = 0; h < FDB_HASH_SIZE; ++h) {
633 spin_lock_bh(&vxlan->hash_lock[h]);
634 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
635 if (f->vni == vni)
636 list_for_each_entry(rdst, &f->remotes, list)
637 rdst->offloaded = false;
638 spin_unlock_bh(&vxlan->hash_lock[h]);
639 }
640
641 }
642 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
643
644 /* Replace destination of unicast mac */
645 static int vxlan_fdb_replace(struct vxlan_fdb *f,
646 union vxlan_addr *ip, __be16 port, __be32 vni,
647 __u32 ifindex, struct vxlan_rdst *oldrd)
648 {
649 struct vxlan_rdst *rd;
650
651 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
652 if (rd)
653 return 0;
654
655 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
656 if (!rd)
657 return 0;
658
659 *oldrd = *rd;
660 dst_cache_reset(&rd->dst_cache);
661 rd->remote_ip = *ip;
662 rd->remote_port = port;
663 rd->remote_vni = vni;
664 rd->remote_ifindex = ifindex;
665 rd->offloaded = false;
666 return 1;
667 }
668
669 /* Add/update destinations for multicast */
670 static int vxlan_fdb_append(struct vxlan_fdb *f,
671 union vxlan_addr *ip, __be16 port, __be32 vni,
672 __u32 ifindex, struct vxlan_rdst **rdp)
673 {
674 struct vxlan_rdst *rd;
675
676 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
677 if (rd)
678 return 0;
679
680 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
681 if (rd == NULL)
682 return -ENOBUFS;
683
684 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
685 kfree(rd);
686 return -ENOBUFS;
687 }
688
689 rd->remote_ip = *ip;
690 rd->remote_port = port;
691 rd->offloaded = false;
692 rd->remote_vni = vni;
693 rd->remote_ifindex = ifindex;
694
695 list_add_tail_rcu(&rd->list, &f->remotes);
696
697 *rdp = rd;
698 return 1;
699 }
700
701 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
702 unsigned int off,
703 struct vxlanhdr *vh, size_t hdrlen,
704 __be32 vni_field,
705 struct gro_remcsum *grc,
706 bool nopartial)
707 {
708 size_t start, offset;
709
710 if (skb->remcsum_offload)
711 return vh;
712
713 if (!NAPI_GRO_CB(skb)->csum_valid)
714 return NULL;
715
716 start = vxlan_rco_start(vni_field);
717 offset = start + vxlan_rco_offset(vni_field);
718
719 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
720 start, offset, grc, nopartial);
721
722 skb->remcsum_offload = 1;
723
724 return vh;
725 }
726
727 static struct sk_buff *vxlan_gro_receive(struct sock *sk,
728 struct list_head *head,
729 struct sk_buff *skb)
730 {
731 struct sk_buff *pp = NULL;
732 struct sk_buff *p;
733 struct vxlanhdr *vh, *vh2;
734 unsigned int hlen, off_vx;
735 int flush = 1;
736 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
737 __be32 flags;
738 struct gro_remcsum grc;
739
740 skb_gro_remcsum_init(&grc);
741
742 off_vx = skb_gro_offset(skb);
743 hlen = off_vx + sizeof(*vh);
744 vh = skb_gro_header_fast(skb, off_vx);
745 if (skb_gro_header_hard(skb, hlen)) {
746 vh = skb_gro_header_slow(skb, hlen, off_vx);
747 if (unlikely(!vh))
748 goto out;
749 }
750
751 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
752
753 flags = vh->vx_flags;
754
755 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
756 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
757 vh->vx_vni, &grc,
758 !!(vs->flags &
759 VXLAN_F_REMCSUM_NOPARTIAL));
760
761 if (!vh)
762 goto out;
763 }
764
765 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
766
767 list_for_each_entry(p, head, list) {
768 if (!NAPI_GRO_CB(p)->same_flow)
769 continue;
770
771 vh2 = (struct vxlanhdr *)(p->data + off_vx);
772 if (vh->vx_flags != vh2->vx_flags ||
773 vh->vx_vni != vh2->vx_vni) {
774 NAPI_GRO_CB(p)->same_flow = 0;
775 continue;
776 }
777 }
778
779 pp = call_gro_receive(eth_gro_receive, head, skb);
780 flush = 0;
781
782 out:
783 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
784
785 return pp;
786 }
787
788 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
789 {
790 /* Sets 'skb->inner_mac_header' since we are always called with
791 * 'skb->encapsulation' set.
792 */
793 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
794 }
795
796 static struct vxlan_fdb *vxlan_fdb_alloc(const u8 *mac, __u16 state,
797 __be32 src_vni, __u16 ndm_flags)
798 {
799 struct vxlan_fdb *f;
800
801 f = kmalloc(sizeof(*f), GFP_ATOMIC);
802 if (!f)
803 return NULL;
804 f->state = state;
805 f->flags = ndm_flags;
806 f->updated = f->used = jiffies;
807 f->vni = src_vni;
808 INIT_LIST_HEAD(&f->remotes);
809 memcpy(f->eth_addr, mac, ETH_ALEN);
810
811 return f;
812 }
813
814 static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
815 __be32 src_vni, struct vxlan_fdb *f)
816 {
817 ++vxlan->addrcnt;
818 hlist_add_head_rcu(&f->hlist,
819 vxlan_fdb_head(vxlan, mac, src_vni));
820 }
821
822 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
823 const u8 *mac, union vxlan_addr *ip,
824 __u16 state, __be16 port, __be32 src_vni,
825 __be32 vni, __u32 ifindex, __u16 ndm_flags,
826 struct vxlan_fdb **fdb)
827 {
828 struct vxlan_rdst *rd = NULL;
829 struct vxlan_fdb *f;
830 int rc;
831
832 if (vxlan->cfg.addrmax &&
833 vxlan->addrcnt >= vxlan->cfg.addrmax)
834 return -ENOSPC;
835
836 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
837 f = vxlan_fdb_alloc(mac, state, src_vni, ndm_flags);
838 if (!f)
839 return -ENOMEM;
840
841 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
842 if (rc < 0) {
843 kfree(f);
844 return rc;
845 }
846
847 *fdb = f;
848
849 return 0;
850 }
851
852 static void __vxlan_fdb_free(struct vxlan_fdb *f)
853 {
854 struct vxlan_rdst *rd, *nd;
855
856 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
857 dst_cache_destroy(&rd->dst_cache);
858 kfree(rd);
859 }
860 kfree(f);
861 }
862
863 static void vxlan_fdb_free(struct rcu_head *head)
864 {
865 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
866
867 __vxlan_fdb_free(f);
868 }
869
870 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
871 bool do_notify, bool swdev_notify)
872 {
873 struct vxlan_rdst *rd;
874
875 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
876
877 --vxlan->addrcnt;
878 if (do_notify)
879 list_for_each_entry(rd, &f->remotes, list)
880 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
881 swdev_notify, NULL);
882
883 hlist_del_rcu(&f->hlist);
884 call_rcu(&f->rcu, vxlan_fdb_free);
885 }
886
887 static void vxlan_dst_free(struct rcu_head *head)
888 {
889 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
890
891 dst_cache_destroy(&rd->dst_cache);
892 kfree(rd);
893 }
894
895 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
896 union vxlan_addr *ip,
897 __u16 state, __u16 flags,
898 __be16 port, __be32 vni,
899 __u32 ifindex, __u16 ndm_flags,
900 struct vxlan_fdb *f,
901 bool swdev_notify,
902 struct netlink_ext_ack *extack)
903 {
904 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
905 struct vxlan_rdst *rd = NULL;
906 struct vxlan_rdst oldrd;
907 int notify = 0;
908 int rc = 0;
909 int err;
910
911 /* Do not allow an externally learned entry to take over an entry added
912 * by the user.
913 */
914 if (!(fdb_flags & NTF_EXT_LEARNED) ||
915 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
916 if (f->state != state) {
917 f->state = state;
918 f->updated = jiffies;
919 notify = 1;
920 }
921 if (f->flags != fdb_flags) {
922 f->flags = fdb_flags;
923 f->updated = jiffies;
924 notify = 1;
925 }
926 }
927
928 if ((flags & NLM_F_REPLACE)) {
929 /* Only change unicasts */
930 if (!(is_multicast_ether_addr(f->eth_addr) ||
931 is_zero_ether_addr(f->eth_addr))) {
932 rc = vxlan_fdb_replace(f, ip, port, vni,
933 ifindex, &oldrd);
934 notify |= rc;
935 } else {
936 return -EOPNOTSUPP;
937 }
938 }
939 if ((flags & NLM_F_APPEND) &&
940 (is_multicast_ether_addr(f->eth_addr) ||
941 is_zero_ether_addr(f->eth_addr))) {
942 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
943
944 if (rc < 0)
945 return rc;
946 notify |= rc;
947 }
948
949 if (ndm_flags & NTF_USE)
950 f->used = jiffies;
951
952 if (notify) {
953 if (rd == NULL)
954 rd = first_remote_rtnl(f);
955
956 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
957 swdev_notify, extack);
958 if (err)
959 goto err_notify;
960 }
961
962 return 0;
963
964 err_notify:
965 if ((flags & NLM_F_REPLACE) && rc)
966 *rd = oldrd;
967 else if ((flags & NLM_F_APPEND) && rc) {
968 list_del_rcu(&rd->list);
969 call_rcu(&rd->rcu, vxlan_dst_free);
970 }
971 return err;
972 }
973
974 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
975 const u8 *mac, union vxlan_addr *ip,
976 __u16 state, __u16 flags,
977 __be16 port, __be32 src_vni, __be32 vni,
978 __u32 ifindex, __u16 ndm_flags,
979 bool swdev_notify,
980 struct netlink_ext_ack *extack)
981 {
982 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
983 struct vxlan_fdb *f;
984 int rc;
985
986 /* Disallow replace to add a multicast entry */
987 if ((flags & NLM_F_REPLACE) &&
988 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
989 return -EOPNOTSUPP;
990
991 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
992 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
993 vni, ifindex, fdb_flags, &f);
994 if (rc < 0)
995 return rc;
996
997 vxlan_fdb_insert(vxlan, mac, src_vni, f);
998 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
999 swdev_notify, extack);
1000 if (rc)
1001 goto err_notify;
1002
1003 return 0;
1004
1005 err_notify:
1006 vxlan_fdb_destroy(vxlan, f, false, false);
1007 return rc;
1008 }
1009
1010 /* Add new entry to forwarding table -- assumes lock held */
1011 static int vxlan_fdb_update(struct vxlan_dev *vxlan,
1012 const u8 *mac, union vxlan_addr *ip,
1013 __u16 state, __u16 flags,
1014 __be16 port, __be32 src_vni, __be32 vni,
1015 __u32 ifindex, __u16 ndm_flags,
1016 bool swdev_notify,
1017 struct netlink_ext_ack *extack)
1018 {
1019 struct vxlan_fdb *f;
1020
1021 f = __vxlan_find_mac(vxlan, mac, src_vni);
1022 if (f) {
1023 if (flags & NLM_F_EXCL) {
1024 netdev_dbg(vxlan->dev,
1025 "lost race to create %pM\n", mac);
1026 return -EEXIST;
1027 }
1028
1029 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1030 vni, ifindex, ndm_flags, f,
1031 swdev_notify, extack);
1032 } else {
1033 if (!(flags & NLM_F_CREATE))
1034 return -ENOENT;
1035
1036 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1037 port, src_vni, vni, ifindex,
1038 ndm_flags, swdev_notify, extack);
1039 }
1040 }
1041
1042 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1043 struct vxlan_rdst *rd, bool swdev_notify)
1044 {
1045 list_del_rcu(&rd->list);
1046 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1047 call_rcu(&rd->rcu, vxlan_dst_free);
1048 }
1049
1050 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1051 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1052 __be32 *vni, u32 *ifindex)
1053 {
1054 struct net *net = dev_net(vxlan->dev);
1055 int err;
1056
1057 if (tb[NDA_DST]) {
1058 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1059 if (err)
1060 return err;
1061 } else {
1062 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1063 if (remote->sa.sa_family == AF_INET) {
1064 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1065 ip->sa.sa_family = AF_INET;
1066 #if IS_ENABLED(CONFIG_IPV6)
1067 } else {
1068 ip->sin6.sin6_addr = in6addr_any;
1069 ip->sa.sa_family = AF_INET6;
1070 #endif
1071 }
1072 }
1073
1074 if (tb[NDA_PORT]) {
1075 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
1076 return -EINVAL;
1077 *port = nla_get_be16(tb[NDA_PORT]);
1078 } else {
1079 *port = vxlan->cfg.dst_port;
1080 }
1081
1082 if (tb[NDA_VNI]) {
1083 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
1084 return -EINVAL;
1085 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1086 } else {
1087 *vni = vxlan->default_dst.remote_vni;
1088 }
1089
1090 if (tb[NDA_SRC_VNI]) {
1091 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
1092 return -EINVAL;
1093 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1094 } else {
1095 *src_vni = vxlan->default_dst.remote_vni;
1096 }
1097
1098 if (tb[NDA_IFINDEX]) {
1099 struct net_device *tdev;
1100
1101 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
1102 return -EINVAL;
1103 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1104 tdev = __dev_get_by_index(net, *ifindex);
1105 if (!tdev)
1106 return -EADDRNOTAVAIL;
1107 } else {
1108 *ifindex = 0;
1109 }
1110
1111 return 0;
1112 }
1113
1114 /* Add static entry (via netlink) */
1115 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1116 struct net_device *dev,
1117 const unsigned char *addr, u16 vid, u16 flags,
1118 struct netlink_ext_ack *extack)
1119 {
1120 struct vxlan_dev *vxlan = netdev_priv(dev);
1121 /* struct net *net = dev_net(vxlan->dev); */
1122 union vxlan_addr ip;
1123 __be16 port;
1124 __be32 src_vni, vni;
1125 u32 ifindex;
1126 u32 hash_index;
1127 int err;
1128
1129 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1130 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1131 ndm->ndm_state);
1132 return -EINVAL;
1133 }
1134
1135 if (tb[NDA_DST] == NULL)
1136 return -EINVAL;
1137
1138 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
1139 if (err)
1140 return err;
1141
1142 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1143 return -EAFNOSUPPORT;
1144
1145 hash_index = fdb_head_index(vxlan, addr, src_vni);
1146 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1147 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1148 port, src_vni, vni, ifindex,
1149 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1150 true, extack);
1151 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1152
1153 return err;
1154 }
1155
1156 static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1157 const unsigned char *addr, union vxlan_addr ip,
1158 __be16 port, __be32 src_vni, __be32 vni,
1159 u32 ifindex, bool swdev_notify)
1160 {
1161 struct vxlan_fdb *f;
1162 struct vxlan_rdst *rd = NULL;
1163 int err = -ENOENT;
1164
1165 f = vxlan_find_mac(vxlan, addr, src_vni);
1166 if (!f)
1167 return err;
1168
1169 if (!vxlan_addr_any(&ip)) {
1170 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1171 if (!rd)
1172 goto out;
1173 }
1174
1175 /* remove a destination if it's not the only one on the list,
1176 * otherwise destroy the fdb entry
1177 */
1178 if (rd && !list_is_singular(&f->remotes)) {
1179 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1180 goto out;
1181 }
1182
1183 vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1184
1185 out:
1186 return 0;
1187 }
1188
1189 /* Delete entry (via netlink) */
1190 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1191 struct net_device *dev,
1192 const unsigned char *addr, u16 vid)
1193 {
1194 struct vxlan_dev *vxlan = netdev_priv(dev);
1195 union vxlan_addr ip;
1196 __be32 src_vni, vni;
1197 __be16 port;
1198 u32 ifindex;
1199 u32 hash_index;
1200 int err;
1201
1202 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
1203 if (err)
1204 return err;
1205
1206 hash_index = fdb_head_index(vxlan, addr, src_vni);
1207 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1208 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1209 true);
1210 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1211
1212 return err;
1213 }
1214
1215 /* Dump forwarding table */
1216 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1217 struct net_device *dev,
1218 struct net_device *filter_dev, int *idx)
1219 {
1220 struct vxlan_dev *vxlan = netdev_priv(dev);
1221 unsigned int h;
1222 int err = 0;
1223
1224 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1225 struct vxlan_fdb *f;
1226
1227 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1228 struct vxlan_rdst *rd;
1229
1230 list_for_each_entry_rcu(rd, &f->remotes, list) {
1231 if (*idx < cb->args[2])
1232 goto skip;
1233
1234 err = vxlan_fdb_info(skb, vxlan, f,
1235 NETLINK_CB(cb->skb).portid,
1236 cb->nlh->nlmsg_seq,
1237 RTM_NEWNEIGH,
1238 NLM_F_MULTI, rd);
1239 if (err < 0)
1240 goto out;
1241 skip:
1242 *idx += 1;
1243 }
1244 }
1245 }
1246 out:
1247 return err;
1248 }
1249
1250 static int vxlan_fdb_get(struct sk_buff *skb,
1251 struct nlattr *tb[],
1252 struct net_device *dev,
1253 const unsigned char *addr,
1254 u16 vid, u32 portid, u32 seq,
1255 struct netlink_ext_ack *extack)
1256 {
1257 struct vxlan_dev *vxlan = netdev_priv(dev);
1258 struct vxlan_fdb *f;
1259 __be32 vni;
1260 int err;
1261
1262 if (tb[NDA_VNI])
1263 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1264 else
1265 vni = vxlan->default_dst.remote_vni;
1266
1267 rcu_read_lock();
1268
1269 f = __vxlan_find_mac(vxlan, addr, vni);
1270 if (!f) {
1271 NL_SET_ERR_MSG(extack, "Fdb entry not found");
1272 err = -ENOENT;
1273 goto errout;
1274 }
1275
1276 err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1277 RTM_NEWNEIGH, 0, first_remote_rcu(f));
1278 errout:
1279 rcu_read_unlock();
1280 return err;
1281 }
1282
1283 /* Watch incoming packets to learn mapping between Ethernet address
1284 * and Tunnel endpoint.
1285 * Return true if packet is bogus and should be dropped.
1286 */
1287 static bool vxlan_snoop(struct net_device *dev,
1288 union vxlan_addr *src_ip, const u8 *src_mac,
1289 u32 src_ifindex, __be32 vni)
1290 {
1291 struct vxlan_dev *vxlan = netdev_priv(dev);
1292 struct vxlan_fdb *f;
1293 u32 ifindex = 0;
1294
1295 #if IS_ENABLED(CONFIG_IPV6)
1296 if (src_ip->sa.sa_family == AF_INET6 &&
1297 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1298 ifindex = src_ifindex;
1299 #endif
1300
1301 f = vxlan_find_mac(vxlan, src_mac, vni);
1302 if (likely(f)) {
1303 struct vxlan_rdst *rdst = first_remote_rcu(f);
1304
1305 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1306 rdst->remote_ifindex == ifindex))
1307 return false;
1308
1309 /* Don't migrate static entries, drop packets */
1310 if (f->state & (NUD_PERMANENT | NUD_NOARP))
1311 return true;
1312
1313 if (net_ratelimit())
1314 netdev_info(dev,
1315 "%pM migrated from %pIS to %pIS\n",
1316 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1317
1318 rdst->remote_ip = *src_ip;
1319 f->updated = jiffies;
1320 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1321 } else {
1322 u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1323
1324 /* learned new entry */
1325 spin_lock(&vxlan->hash_lock[hash_index]);
1326
1327 /* close off race between vxlan_flush and incoming packets */
1328 if (netif_running(dev))
1329 vxlan_fdb_update(vxlan, src_mac, src_ip,
1330 NUD_REACHABLE,
1331 NLM_F_EXCL|NLM_F_CREATE,
1332 vxlan->cfg.dst_port,
1333 vni,
1334 vxlan->default_dst.remote_vni,
1335 ifindex, NTF_SELF, true, NULL);
1336 spin_unlock(&vxlan->hash_lock[hash_index]);
1337 }
1338
1339 return false;
1340 }
1341
1342 /* See if multicast group is already in use by other ID */
1343 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1344 {
1345 struct vxlan_dev *vxlan;
1346 struct vxlan_sock *sock4;
1347 #if IS_ENABLED(CONFIG_IPV6)
1348 struct vxlan_sock *sock6;
1349 #endif
1350 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1351
1352 sock4 = rtnl_dereference(dev->vn4_sock);
1353
1354 /* The vxlan_sock is only used by dev, leaving group has
1355 * no effect on other vxlan devices.
1356 */
1357 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
1358 return false;
1359 #if IS_ENABLED(CONFIG_IPV6)
1360 sock6 = rtnl_dereference(dev->vn6_sock);
1361 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
1362 return false;
1363 #endif
1364
1365 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1366 if (!netif_running(vxlan->dev) || vxlan == dev)
1367 continue;
1368
1369 if (family == AF_INET &&
1370 rtnl_dereference(vxlan->vn4_sock) != sock4)
1371 continue;
1372 #if IS_ENABLED(CONFIG_IPV6)
1373 if (family == AF_INET6 &&
1374 rtnl_dereference(vxlan->vn6_sock) != sock6)
1375 continue;
1376 #endif
1377
1378 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1379 &dev->default_dst.remote_ip))
1380 continue;
1381
1382 if (vxlan->default_dst.remote_ifindex !=
1383 dev->default_dst.remote_ifindex)
1384 continue;
1385
1386 return true;
1387 }
1388
1389 return false;
1390 }
1391
1392 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1393 {
1394 struct vxlan_net *vn;
1395
1396 if (!vs)
1397 return false;
1398 if (!refcount_dec_and_test(&vs->refcnt))
1399 return false;
1400
1401 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1402 spin_lock(&vn->sock_lock);
1403 hlist_del_rcu(&vs->hlist);
1404 udp_tunnel_notify_del_rx_port(vs->sock,
1405 (vs->flags & VXLAN_F_GPE) ?
1406 UDP_TUNNEL_TYPE_VXLAN_GPE :
1407 UDP_TUNNEL_TYPE_VXLAN);
1408 spin_unlock(&vn->sock_lock);
1409
1410 return true;
1411 }
1412
1413 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1414 {
1415 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1416 #if IS_ENABLED(CONFIG_IPV6)
1417 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1418
1419 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1420 #endif
1421
1422 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1423 synchronize_net();
1424
1425 vxlan_vs_del_dev(vxlan);
1426
1427 if (__vxlan_sock_release_prep(sock4)) {
1428 udp_tunnel_sock_release(sock4->sock);
1429 kfree(sock4);
1430 }
1431
1432 #if IS_ENABLED(CONFIG_IPV6)
1433 if (__vxlan_sock_release_prep(sock6)) {
1434 udp_tunnel_sock_release(sock6->sock);
1435 kfree(sock6);
1436 }
1437 #endif
1438 }
1439
1440 /* Update multicast group membership when first VNI on
1441 * multicast address is brought up
1442 */
1443 static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1444 {
1445 struct sock *sk;
1446 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1447 int ifindex = vxlan->default_dst.remote_ifindex;
1448 int ret = -EINVAL;
1449
1450 if (ip->sa.sa_family == AF_INET) {
1451 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1452 struct ip_mreqn mreq = {
1453 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1454 .imr_ifindex = ifindex,
1455 };
1456
1457 sk = sock4->sock->sk;
1458 lock_sock(sk);
1459 ret = ip_mc_join_group(sk, &mreq);
1460 release_sock(sk);
1461 #if IS_ENABLED(CONFIG_IPV6)
1462 } else {
1463 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1464
1465 sk = sock6->sock->sk;
1466 lock_sock(sk);
1467 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1468 &ip->sin6.sin6_addr);
1469 release_sock(sk);
1470 #endif
1471 }
1472
1473 return ret;
1474 }
1475
1476 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1477 static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1478 {
1479 struct sock *sk;
1480 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1481 int ifindex = vxlan->default_dst.remote_ifindex;
1482 int ret = -EINVAL;
1483
1484 if (ip->sa.sa_family == AF_INET) {
1485 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1486 struct ip_mreqn mreq = {
1487 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1488 .imr_ifindex = ifindex,
1489 };
1490
1491 sk = sock4->sock->sk;
1492 lock_sock(sk);
1493 ret = ip_mc_leave_group(sk, &mreq);
1494 release_sock(sk);
1495 #if IS_ENABLED(CONFIG_IPV6)
1496 } else {
1497 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1498
1499 sk = sock6->sock->sk;
1500 lock_sock(sk);
1501 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1502 &ip->sin6.sin6_addr);
1503 release_sock(sk);
1504 #endif
1505 }
1506
1507 return ret;
1508 }
1509
1510 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1511 struct sk_buff *skb, u32 vxflags)
1512 {
1513 size_t start, offset;
1514
1515 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1516 goto out;
1517
1518 start = vxlan_rco_start(unparsed->vx_vni);
1519 offset = start + vxlan_rco_offset(unparsed->vx_vni);
1520
1521 if (!pskb_may_pull(skb, offset + sizeof(u16)))
1522 return false;
1523
1524 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1525 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1526 out:
1527 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1528 unparsed->vx_vni &= VXLAN_VNI_MASK;
1529 return true;
1530 }
1531
1532 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1533 struct sk_buff *skb, u32 vxflags,
1534 struct vxlan_metadata *md)
1535 {
1536 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1537 struct metadata_dst *tun_dst;
1538
1539 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1540 goto out;
1541
1542 md->gbp = ntohs(gbp->policy_id);
1543
1544 tun_dst = (struct metadata_dst *)skb_dst(skb);
1545 if (tun_dst) {
1546 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1547 tun_dst->u.tun_info.options_len = sizeof(*md);
1548 }
1549 if (gbp->dont_learn)
1550 md->gbp |= VXLAN_GBP_DONT_LEARN;
1551
1552 if (gbp->policy_applied)
1553 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1554
1555 /* In flow-based mode, GBP is carried in dst_metadata */
1556 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1557 skb->mark = md->gbp;
1558 out:
1559 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1560 }
1561
1562 static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1563 __be16 *protocol,
1564 struct sk_buff *skb, u32 vxflags)
1565 {
1566 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1567
1568 /* Need to have Next Protocol set for interfaces in GPE mode. */
1569 if (!gpe->np_applied)
1570 return false;
1571 /* "The initial version is 0. If a receiver does not support the
1572 * version indicated it MUST drop the packet.
1573 */
1574 if (gpe->version != 0)
1575 return false;
1576 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1577 * processing MUST occur." However, we don't implement OAM
1578 * processing, thus drop the packet.
1579 */
1580 if (gpe->oam_flag)
1581 return false;
1582
1583 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1584 if (!*protocol)
1585 return false;
1586
1587 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1588 return true;
1589 }
1590
1591 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1592 struct vxlan_sock *vs,
1593 struct sk_buff *skb, __be32 vni)
1594 {
1595 union vxlan_addr saddr;
1596 u32 ifindex = skb->dev->ifindex;
1597
1598 skb_reset_mac_header(skb);
1599 skb->protocol = eth_type_trans(skb, vxlan->dev);
1600 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1601
1602 /* Ignore packet loops (and multicast echo) */
1603 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1604 return false;
1605
1606 /* Get address from the outer IP header */
1607 if (vxlan_get_sk_family(vs) == AF_INET) {
1608 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1609 saddr.sa.sa_family = AF_INET;
1610 #if IS_ENABLED(CONFIG_IPV6)
1611 } else {
1612 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1613 saddr.sa.sa_family = AF_INET6;
1614 #endif
1615 }
1616
1617 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1618 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1619 return false;
1620
1621 return true;
1622 }
1623
1624 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1625 struct sk_buff *skb)
1626 {
1627 int err = 0;
1628
1629 if (vxlan_get_sk_family(vs) == AF_INET)
1630 err = IP_ECN_decapsulate(oiph, skb);
1631 #if IS_ENABLED(CONFIG_IPV6)
1632 else
1633 err = IP6_ECN_decapsulate(oiph, skb);
1634 #endif
1635
1636 if (unlikely(err) && log_ecn_error) {
1637 if (vxlan_get_sk_family(vs) == AF_INET)
1638 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1639 &((struct iphdr *)oiph)->saddr,
1640 ((struct iphdr *)oiph)->tos);
1641 else
1642 net_info_ratelimited("non-ECT from %pI6\n",
1643 &((struct ipv6hdr *)oiph)->saddr);
1644 }
1645 return err <= 1;
1646 }
1647
1648 /* Callback from net/ipv4/udp.c to receive packets */
1649 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1650 {
1651 struct pcpu_sw_netstats *stats;
1652 struct vxlan_dev *vxlan;
1653 struct vxlan_sock *vs;
1654 struct vxlanhdr unparsed;
1655 struct vxlan_metadata _md;
1656 struct vxlan_metadata *md = &_md;
1657 __be16 protocol = htons(ETH_P_TEB);
1658 bool raw_proto = false;
1659 void *oiph;
1660 __be32 vni = 0;
1661
1662 /* Need UDP and VXLAN header to be present */
1663 if (!pskb_may_pull(skb, VXLAN_HLEN))
1664 goto drop;
1665
1666 unparsed = *vxlan_hdr(skb);
1667 /* VNI flag always required to be set */
1668 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1669 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1670 ntohl(vxlan_hdr(skb)->vx_flags),
1671 ntohl(vxlan_hdr(skb)->vx_vni));
1672 /* Return non vxlan pkt */
1673 goto drop;
1674 }
1675 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1676 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1677
1678 vs = rcu_dereference_sk_user_data(sk);
1679 if (!vs)
1680 goto drop;
1681
1682 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1683
1684 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1685 if (!vxlan)
1686 goto drop;
1687
1688 /* For backwards compatibility, only allow reserved fields to be
1689 * used by VXLAN extensions if explicitly requested.
1690 */
1691 if (vs->flags & VXLAN_F_GPE) {
1692 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1693 goto drop;
1694 raw_proto = true;
1695 }
1696
1697 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1698 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1699 goto drop;
1700
1701 if (vxlan_collect_metadata(vs)) {
1702 struct metadata_dst *tun_dst;
1703
1704 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1705 key32_to_tunnel_id(vni), sizeof(*md));
1706
1707 if (!tun_dst)
1708 goto drop;
1709
1710 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1711
1712 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1713 } else {
1714 memset(md, 0, sizeof(*md));
1715 }
1716
1717 if (vs->flags & VXLAN_F_REMCSUM_RX)
1718 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1719 goto drop;
1720 if (vs->flags & VXLAN_F_GBP)
1721 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1722 /* Note that GBP and GPE can never be active together. This is
1723 * ensured in vxlan_dev_configure.
1724 */
1725
1726 if (unparsed.vx_flags || unparsed.vx_vni) {
1727 /* If there are any unprocessed flags remaining treat
1728 * this as a malformed packet. This behavior diverges from
1729 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1730 * in reserved fields are to be ignored. The approach here
1731 * maintains compatibility with previous stack code, and also
1732 * is more robust and provides a little more security in
1733 * adding extensions to VXLAN.
1734 */
1735 goto drop;
1736 }
1737
1738 if (!raw_proto) {
1739 if (!vxlan_set_mac(vxlan, vs, skb, vni))
1740 goto drop;
1741 } else {
1742 skb_reset_mac_header(skb);
1743 skb->dev = vxlan->dev;
1744 skb->pkt_type = PACKET_HOST;
1745 }
1746
1747 oiph = skb_network_header(skb);
1748 skb_reset_network_header(skb);
1749
1750 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1751 ++vxlan->dev->stats.rx_frame_errors;
1752 ++vxlan->dev->stats.rx_errors;
1753 goto drop;
1754 }
1755
1756 rcu_read_lock();
1757
1758 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1759 rcu_read_unlock();
1760 atomic_long_inc(&vxlan->dev->rx_dropped);
1761 goto drop;
1762 }
1763
1764 stats = this_cpu_ptr(vxlan->dev->tstats);
1765 u64_stats_update_begin(&stats->syncp);
1766 stats->rx_packets++;
1767 stats->rx_bytes += skb->len;
1768 u64_stats_update_end(&stats->syncp);
1769
1770 gro_cells_receive(&vxlan->gro_cells, skb);
1771
1772 rcu_read_unlock();
1773
1774 return 0;
1775
1776 drop:
1777 /* Consume bad packet */
1778 kfree_skb(skb);
1779 return 0;
1780 }
1781
1782 /* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1783 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1784 {
1785 struct vxlan_dev *vxlan;
1786 struct vxlan_sock *vs;
1787 struct vxlanhdr *hdr;
1788 __be32 vni;
1789
1790 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1791 return -EINVAL;
1792
1793 hdr = vxlan_hdr(skb);
1794
1795 if (!(hdr->vx_flags & VXLAN_HF_VNI))
1796 return -EINVAL;
1797
1798 vs = rcu_dereference_sk_user_data(sk);
1799 if (!vs)
1800 return -ENOENT;
1801
1802 vni = vxlan_vni(hdr->vx_vni);
1803 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1804 if (!vxlan)
1805 return -ENOENT;
1806
1807 return 0;
1808 }
1809
1810 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1811 {
1812 struct vxlan_dev *vxlan = netdev_priv(dev);
1813 struct arphdr *parp;
1814 u8 *arpptr, *sha;
1815 __be32 sip, tip;
1816 struct neighbour *n;
1817
1818 if (dev->flags & IFF_NOARP)
1819 goto out;
1820
1821 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1822 dev->stats.tx_dropped++;
1823 goto out;
1824 }
1825 parp = arp_hdr(skb);
1826
1827 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1828 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1829 parp->ar_pro != htons(ETH_P_IP) ||
1830 parp->ar_op != htons(ARPOP_REQUEST) ||
1831 parp->ar_hln != dev->addr_len ||
1832 parp->ar_pln != 4)
1833 goto out;
1834 arpptr = (u8 *)parp + sizeof(struct arphdr);
1835 sha = arpptr;
1836 arpptr += dev->addr_len; /* sha */
1837 memcpy(&sip, arpptr, sizeof(sip));
1838 arpptr += sizeof(sip);
1839 arpptr += dev->addr_len; /* tha */
1840 memcpy(&tip, arpptr, sizeof(tip));
1841
1842 if (ipv4_is_loopback(tip) ||
1843 ipv4_is_multicast(tip))
1844 goto out;
1845
1846 n = neigh_lookup(&arp_tbl, &tip, dev);
1847
1848 if (n) {
1849 struct vxlan_fdb *f;
1850 struct sk_buff *reply;
1851
1852 if (!(n->nud_state & NUD_CONNECTED)) {
1853 neigh_release(n);
1854 goto out;
1855 }
1856
1857 f = vxlan_find_mac(vxlan, n->ha, vni);
1858 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1859 /* bridge-local neighbor */
1860 neigh_release(n);
1861 goto out;
1862 }
1863
1864 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1865 n->ha, sha);
1866
1867 neigh_release(n);
1868
1869 if (reply == NULL)
1870 goto out;
1871
1872 skb_reset_mac_header(reply);
1873 __skb_pull(reply, skb_network_offset(reply));
1874 reply->ip_summed = CHECKSUM_UNNECESSARY;
1875 reply->pkt_type = PACKET_HOST;
1876
1877 if (netif_rx_ni(reply) == NET_RX_DROP)
1878 dev->stats.rx_dropped++;
1879 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1880 union vxlan_addr ipa = {
1881 .sin.sin_addr.s_addr = tip,
1882 .sin.sin_family = AF_INET,
1883 };
1884
1885 vxlan_ip_miss(dev, &ipa);
1886 }
1887 out:
1888 consume_skb(skb);
1889 return NETDEV_TX_OK;
1890 }
1891
1892 #if IS_ENABLED(CONFIG_IPV6)
1893 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1894 struct neighbour *n, bool isrouter)
1895 {
1896 struct net_device *dev = request->dev;
1897 struct sk_buff *reply;
1898 struct nd_msg *ns, *na;
1899 struct ipv6hdr *pip6;
1900 u8 *daddr;
1901 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1902 int ns_olen;
1903 int i, len;
1904
1905 if (dev == NULL || !pskb_may_pull(request, request->len))
1906 return NULL;
1907
1908 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1909 sizeof(*na) + na_olen + dev->needed_tailroom;
1910 reply = alloc_skb(len, GFP_ATOMIC);
1911 if (reply == NULL)
1912 return NULL;
1913
1914 reply->protocol = htons(ETH_P_IPV6);
1915 reply->dev = dev;
1916 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1917 skb_push(reply, sizeof(struct ethhdr));
1918 skb_reset_mac_header(reply);
1919
1920 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
1921
1922 daddr = eth_hdr(request)->h_source;
1923 ns_olen = request->len - skb_network_offset(request) -
1924 sizeof(struct ipv6hdr) - sizeof(*ns);
1925 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1926 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1927 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1928 break;
1929 }
1930 }
1931
1932 /* Ethernet header */
1933 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1934 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1935 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1936 reply->protocol = htons(ETH_P_IPV6);
1937
1938 skb_pull(reply, sizeof(struct ethhdr));
1939 skb_reset_network_header(reply);
1940 skb_put(reply, sizeof(struct ipv6hdr));
1941
1942 /* IPv6 header */
1943
1944 pip6 = ipv6_hdr(reply);
1945 memset(pip6, 0, sizeof(struct ipv6hdr));
1946 pip6->version = 6;
1947 pip6->priority = ipv6_hdr(request)->priority;
1948 pip6->nexthdr = IPPROTO_ICMPV6;
1949 pip6->hop_limit = 255;
1950 pip6->daddr = ipv6_hdr(request)->saddr;
1951 pip6->saddr = *(struct in6_addr *)n->primary_key;
1952
1953 skb_pull(reply, sizeof(struct ipv6hdr));
1954 skb_reset_transport_header(reply);
1955
1956 /* Neighbor Advertisement */
1957 na = skb_put_zero(reply, sizeof(*na) + na_olen);
1958 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1959 na->icmph.icmp6_router = isrouter;
1960 na->icmph.icmp6_override = 1;
1961 na->icmph.icmp6_solicited = 1;
1962 na->target = ns->target;
1963 ether_addr_copy(&na->opt[2], n->ha);
1964 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1965 na->opt[1] = na_olen >> 3;
1966
1967 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1968 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1969 csum_partial(na, sizeof(*na)+na_olen, 0));
1970
1971 pip6->payload_len = htons(sizeof(*na)+na_olen);
1972
1973 skb_push(reply, sizeof(struct ipv6hdr));
1974
1975 reply->ip_summed = CHECKSUM_UNNECESSARY;
1976
1977 return reply;
1978 }
1979
1980 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1981 {
1982 struct vxlan_dev *vxlan = netdev_priv(dev);
1983 const struct in6_addr *daddr;
1984 const struct ipv6hdr *iphdr;
1985 struct inet6_dev *in6_dev;
1986 struct neighbour *n;
1987 struct nd_msg *msg;
1988
1989 in6_dev = __in6_dev_get(dev);
1990 if (!in6_dev)
1991 goto out;
1992
1993 iphdr = ipv6_hdr(skb);
1994 daddr = &iphdr->daddr;
1995 msg = (struct nd_msg *)(iphdr + 1);
1996
1997 if (ipv6_addr_loopback(daddr) ||
1998 ipv6_addr_is_multicast(&msg->target))
1999 goto out;
2000
2001 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2002
2003 if (n) {
2004 struct vxlan_fdb *f;
2005 struct sk_buff *reply;
2006
2007 if (!(n->nud_state & NUD_CONNECTED)) {
2008 neigh_release(n);
2009 goto out;
2010 }
2011
2012 f = vxlan_find_mac(vxlan, n->ha, vni);
2013 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2014 /* bridge-local neighbor */
2015 neigh_release(n);
2016 goto out;
2017 }
2018
2019 reply = vxlan_na_create(skb, n,
2020 !!(f ? f->flags & NTF_ROUTER : 0));
2021
2022 neigh_release(n);
2023
2024 if (reply == NULL)
2025 goto out;
2026
2027 if (netif_rx_ni(reply) == NET_RX_DROP)
2028 dev->stats.rx_dropped++;
2029
2030 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2031 union vxlan_addr ipa = {
2032 .sin6.sin6_addr = msg->target,
2033 .sin6.sin6_family = AF_INET6,
2034 };
2035
2036 vxlan_ip_miss(dev, &ipa);
2037 }
2038
2039 out:
2040 consume_skb(skb);
2041 return NETDEV_TX_OK;
2042 }
2043 #endif
2044
2045 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2046 {
2047 struct vxlan_dev *vxlan = netdev_priv(dev);
2048 struct neighbour *n;
2049
2050 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2051 return false;
2052
2053 n = NULL;
2054 switch (ntohs(eth_hdr(skb)->h_proto)) {
2055 case ETH_P_IP:
2056 {
2057 struct iphdr *pip;
2058
2059 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2060 return false;
2061 pip = ip_hdr(skb);
2062 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2063 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2064 union vxlan_addr ipa = {
2065 .sin.sin_addr.s_addr = pip->daddr,
2066 .sin.sin_family = AF_INET,
2067 };
2068
2069 vxlan_ip_miss(dev, &ipa);
2070 return false;
2071 }
2072
2073 break;
2074 }
2075 #if IS_ENABLED(CONFIG_IPV6)
2076 case ETH_P_IPV6:
2077 {
2078 struct ipv6hdr *pip6;
2079
2080 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2081 return false;
2082 pip6 = ipv6_hdr(skb);
2083 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2084 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2085 union vxlan_addr ipa = {
2086 .sin6.sin6_addr = pip6->daddr,
2087 .sin6.sin6_family = AF_INET6,
2088 };
2089
2090 vxlan_ip_miss(dev, &ipa);
2091 return false;
2092 }
2093
2094 break;
2095 }
2096 #endif
2097 default:
2098 return false;
2099 }
2100
2101 if (n) {
2102 bool diff;
2103
2104 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2105 if (diff) {
2106 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2107 dev->addr_len);
2108 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2109 }
2110 neigh_release(n);
2111 return diff;
2112 }
2113
2114 return false;
2115 }
2116
2117 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
2118 struct vxlan_metadata *md)
2119 {
2120 struct vxlanhdr_gbp *gbp;
2121
2122 if (!md->gbp)
2123 return;
2124
2125 gbp = (struct vxlanhdr_gbp *)vxh;
2126 vxh->vx_flags |= VXLAN_HF_GBP;
2127
2128 if (md->gbp & VXLAN_GBP_DONT_LEARN)
2129 gbp->dont_learn = 1;
2130
2131 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
2132 gbp->policy_applied = 1;
2133
2134 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
2135 }
2136
2137 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
2138 __be16 protocol)
2139 {
2140 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2141
2142 gpe->np_applied = 1;
2143 gpe->next_protocol = tun_p_from_eth_p(protocol);
2144 if (!gpe->next_protocol)
2145 return -EPFNOSUPPORT;
2146 return 0;
2147 }
2148
2149 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2150 int iphdr_len, __be32 vni,
2151 struct vxlan_metadata *md, u32 vxflags,
2152 bool udp_sum)
2153 {
2154 struct vxlanhdr *vxh;
2155 int min_headroom;
2156 int err;
2157 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2158 __be16 inner_protocol = htons(ETH_P_TEB);
2159
2160 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2161 skb->ip_summed == CHECKSUM_PARTIAL) {
2162 int csum_start = skb_checksum_start_offset(skb);
2163
2164 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2165 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2166 (skb->csum_offset == offsetof(struct udphdr, check) ||
2167 skb->csum_offset == offsetof(struct tcphdr, check)))
2168 type |= SKB_GSO_TUNNEL_REMCSUM;
2169 }
2170
2171 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2172 + VXLAN_HLEN + iphdr_len;
2173
2174 /* Need space for new headers (invalidates iph ptr) */
2175 err = skb_cow_head(skb, min_headroom);
2176 if (unlikely(err))
2177 return err;
2178
2179 err = iptunnel_handle_offloads(skb, type);
2180 if (err)
2181 return err;
2182
2183 vxh = __skb_push(skb, sizeof(*vxh));
2184 vxh->vx_flags = VXLAN_HF_VNI;
2185 vxh->vx_vni = vxlan_vni_field(vni);
2186
2187 if (type & SKB_GSO_TUNNEL_REMCSUM) {
2188 unsigned int start;
2189
2190 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2191 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2192 vxh->vx_flags |= VXLAN_HF_RCO;
2193
2194 if (!skb_is_gso(skb)) {
2195 skb->ip_summed = CHECKSUM_NONE;
2196 skb->encapsulation = 0;
2197 }
2198 }
2199
2200 if (vxflags & VXLAN_F_GBP)
2201 vxlan_build_gbp_hdr(vxh, vxflags, md);
2202 if (vxflags & VXLAN_F_GPE) {
2203 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2204 if (err < 0)
2205 return err;
2206 inner_protocol = skb->protocol;
2207 }
2208
2209 skb_set_inner_protocol(skb, inner_protocol);
2210 return 0;
2211 }
2212
2213 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2214 struct vxlan_sock *sock4,
2215 struct sk_buff *skb, int oif, u8 tos,
2216 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
2217 struct dst_cache *dst_cache,
2218 const struct ip_tunnel_info *info)
2219 {
2220 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2221 struct rtable *rt = NULL;
2222 struct flowi4 fl4;
2223
2224 if (!sock4)
2225 return ERR_PTR(-EIO);
2226
2227 if (tos && !info)
2228 use_cache = false;
2229 if (use_cache) {
2230 rt = dst_cache_get_ip4(dst_cache, saddr);
2231 if (rt)
2232 return rt;
2233 }
2234
2235 memset(&fl4, 0, sizeof(fl4));
2236 fl4.flowi4_oif = oif;
2237 fl4.flowi4_tos = RT_TOS(tos);
2238 fl4.flowi4_mark = skb->mark;
2239 fl4.flowi4_proto = IPPROTO_UDP;
2240 fl4.daddr = daddr;
2241 fl4.saddr = *saddr;
2242 fl4.fl4_dport = dport;
2243 fl4.fl4_sport = sport;
2244
2245 rt = ip_route_output_key(vxlan->net, &fl4);
2246 if (!IS_ERR(rt)) {
2247 if (rt->dst.dev == dev) {
2248 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2249 ip_rt_put(rt);
2250 return ERR_PTR(-ELOOP);
2251 }
2252
2253 *saddr = fl4.saddr;
2254 if (use_cache)
2255 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2256 } else {
2257 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2258 return ERR_PTR(-ENETUNREACH);
2259 }
2260 return rt;
2261 }
2262
2263 #if IS_ENABLED(CONFIG_IPV6)
2264 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2265 struct net_device *dev,
2266 struct vxlan_sock *sock6,
2267 struct sk_buff *skb, int oif, u8 tos,
2268 __be32 label,
2269 const struct in6_addr *daddr,
2270 struct in6_addr *saddr,
2271 __be16 dport, __be16 sport,
2272 struct dst_cache *dst_cache,
2273 const struct ip_tunnel_info *info)
2274 {
2275 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2276 struct dst_entry *ndst;
2277 struct flowi6 fl6;
2278 int err;
2279
2280 if (!sock6)
2281 return ERR_PTR(-EIO);
2282
2283 if (tos && !info)
2284 use_cache = false;
2285 if (use_cache) {
2286 ndst = dst_cache_get_ip6(dst_cache, saddr);
2287 if (ndst)
2288 return ndst;
2289 }
2290
2291 memset(&fl6, 0, sizeof(fl6));
2292 fl6.flowi6_oif = oif;
2293 fl6.daddr = *daddr;
2294 fl6.saddr = *saddr;
2295 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
2296 fl6.flowi6_mark = skb->mark;
2297 fl6.flowi6_proto = IPPROTO_UDP;
2298 fl6.fl6_dport = dport;
2299 fl6.fl6_sport = sport;
2300
2301 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
2302 sock6->sock->sk,
2303 &ndst, &fl6);
2304 if (unlikely(err < 0)) {
2305 netdev_dbg(dev, "no route to %pI6\n", daddr);
2306 return ERR_PTR(-ENETUNREACH);
2307 }
2308
2309 if (unlikely(ndst->dev == dev)) {
2310 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2311 dst_release(ndst);
2312 return ERR_PTR(-ELOOP);
2313 }
2314
2315 *saddr = fl6.saddr;
2316 if (use_cache)
2317 dst_cache_set_ip6(dst_cache, ndst, saddr);
2318 return ndst;
2319 }
2320 #endif
2321
2322 /* Bypass encapsulation if the destination is local */
2323 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2324 struct vxlan_dev *dst_vxlan, __be32 vni)
2325 {
2326 struct pcpu_sw_netstats *tx_stats, *rx_stats;
2327 union vxlan_addr loopback;
2328 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2329 struct net_device *dev;
2330 int len = skb->len;
2331
2332 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2333 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2334 skb->pkt_type = PACKET_HOST;
2335 skb->encapsulation = 0;
2336 skb->dev = dst_vxlan->dev;
2337 __skb_pull(skb, skb_network_offset(skb));
2338
2339 if (remote_ip->sa.sa_family == AF_INET) {
2340 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2341 loopback.sa.sa_family = AF_INET;
2342 #if IS_ENABLED(CONFIG_IPV6)
2343 } else {
2344 loopback.sin6.sin6_addr = in6addr_loopback;
2345 loopback.sa.sa_family = AF_INET6;
2346 #endif
2347 }
2348
2349 rcu_read_lock();
2350 dev = skb->dev;
2351 if (unlikely(!(dev->flags & IFF_UP))) {
2352 kfree_skb(skb);
2353 goto drop;
2354 }
2355
2356 if (dst_vxlan->cfg.flags & VXLAN_F_LEARN)
2357 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2358
2359 u64_stats_update_begin(&tx_stats->syncp);
2360 tx_stats->tx_packets++;
2361 tx_stats->tx_bytes += len;
2362 u64_stats_update_end(&tx_stats->syncp);
2363
2364 if (netif_rx(skb) == NET_RX_SUCCESS) {
2365 u64_stats_update_begin(&rx_stats->syncp);
2366 rx_stats->rx_packets++;
2367 rx_stats->rx_bytes += len;
2368 u64_stats_update_end(&rx_stats->syncp);
2369 } else {
2370 drop:
2371 dev->stats.rx_dropped++;
2372 }
2373 rcu_read_unlock();
2374 }
2375
2376 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2377 struct vxlan_dev *vxlan,
2378 union vxlan_addr *daddr,
2379 __be16 dst_port, int dst_ifindex, __be32 vni,
2380 struct dst_entry *dst,
2381 u32 rt_flags)
2382 {
2383 #if IS_ENABLED(CONFIG_IPV6)
2384 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2385 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2386 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2387 */
2388 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2389 #endif
2390 /* Bypass encapsulation if the destination is local */
2391 if (rt_flags & RTCF_LOCAL &&
2392 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2393 struct vxlan_dev *dst_vxlan;
2394
2395 dst_release(dst);
2396 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2397 daddr->sa.sa_family, dst_port,
2398 vxlan->cfg.flags);
2399 if (!dst_vxlan) {
2400 dev->stats.tx_errors++;
2401 kfree_skb(skb);
2402
2403 return -ENOENT;
2404 }
2405 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
2406 return 1;
2407 }
2408
2409 return 0;
2410 }
2411
2412 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2413 __be32 default_vni, struct vxlan_rdst *rdst,
2414 bool did_rsc)
2415 {
2416 struct dst_cache *dst_cache;
2417 struct ip_tunnel_info *info;
2418 struct vxlan_dev *vxlan = netdev_priv(dev);
2419 const struct iphdr *old_iph = ip_hdr(skb);
2420 union vxlan_addr *dst;
2421 union vxlan_addr remote_ip, local_ip;
2422 struct vxlan_metadata _md;
2423 struct vxlan_metadata *md = &_md;
2424 __be16 src_port = 0, dst_port;
2425 struct dst_entry *ndst = NULL;
2426 __be32 vni, label;
2427 __u8 tos, ttl;
2428 int ifindex;
2429 int err;
2430 u32 flags = vxlan->cfg.flags;
2431 bool udp_sum = false;
2432 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2433
2434 info = skb_tunnel_info(skb);
2435
2436 if (rdst) {
2437 dst = &rdst->remote_ip;
2438 if (vxlan_addr_any(dst)) {
2439 if (did_rsc) {
2440 /* short-circuited back to local bridge */
2441 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
2442 return;
2443 }
2444 goto drop;
2445 }
2446
2447 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2448 vni = (rdst->remote_vni) ? : default_vni;
2449 ifindex = rdst->remote_ifindex;
2450 local_ip = vxlan->cfg.saddr;
2451 dst_cache = &rdst->dst_cache;
2452 md->gbp = skb->mark;
2453 if (flags & VXLAN_F_TTL_INHERIT) {
2454 ttl = ip_tunnel_get_ttl(old_iph, skb);
2455 } else {
2456 ttl = vxlan->cfg.ttl;
2457 if (!ttl && vxlan_addr_multicast(dst))
2458 ttl = 1;
2459 }
2460
2461 tos = vxlan->cfg.tos;
2462 if (tos == 1)
2463 tos = ip_tunnel_get_dsfield(old_iph, skb);
2464
2465 if (dst->sa.sa_family == AF_INET)
2466 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2467 else
2468 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2469 label = vxlan->cfg.label;
2470 } else {
2471 if (!info) {
2472 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2473 dev->name);
2474 goto drop;
2475 }
2476 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2477 if (remote_ip.sa.sa_family == AF_INET) {
2478 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2479 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2480 } else {
2481 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2482 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2483 }
2484 dst = &remote_ip;
2485 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2486 vni = tunnel_id_to_key32(info->key.tun_id);
2487 ifindex = 0;
2488 dst_cache = &info->dst_cache;
2489 if (info->options_len &&
2490 info->key.tun_flags & TUNNEL_VXLAN_OPT)
2491 md = ip_tunnel_info_opts(info);
2492 ttl = info->key.ttl;
2493 tos = info->key.tos;
2494 label = info->key.label;
2495 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2496 }
2497 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2498 vxlan->cfg.port_max, true);
2499
2500 rcu_read_lock();
2501 if (dst->sa.sa_family == AF_INET) {
2502 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2503 struct rtable *rt;
2504 __be16 df = 0;
2505
2506 if (!ifindex)
2507 ifindex = sock4->sock->sk->sk_bound_dev_if;
2508
2509 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2510 dst->sin.sin_addr.s_addr,
2511 &local_ip.sin.sin_addr.s_addr,
2512 dst_port, src_port,
2513 dst_cache, info);
2514 if (IS_ERR(rt)) {
2515 err = PTR_ERR(rt);
2516 goto tx_error;
2517 }
2518
2519 if (!info) {
2520 /* Bypass encapsulation if the destination is local */
2521 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2522 dst_port, ifindex, vni,
2523 &rt->dst, rt->rt_flags);
2524 if (err)
2525 goto out_unlock;
2526
2527 if (vxlan->cfg.df == VXLAN_DF_SET) {
2528 df = htons(IP_DF);
2529 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2530 struct ethhdr *eth = eth_hdr(skb);
2531
2532 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2533 (ntohs(eth->h_proto) == ETH_P_IP &&
2534 old_iph->frag_off & htons(IP_DF)))
2535 df = htons(IP_DF);
2536 }
2537 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2538 df = htons(IP_DF);
2539 }
2540
2541 ndst = &rt->dst;
2542 skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM);
2543
2544 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2545 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2546 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2547 vni, md, flags, udp_sum);
2548 if (err < 0)
2549 goto tx_error;
2550
2551 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2552 dst->sin.sin_addr.s_addr, tos, ttl, df,
2553 src_port, dst_port, xnet, !udp_sum);
2554 #if IS_ENABLED(CONFIG_IPV6)
2555 } else {
2556 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2557
2558 if (!ifindex)
2559 ifindex = sock6->sock->sk->sk_bound_dev_if;
2560
2561 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2562 label, &dst->sin6.sin6_addr,
2563 &local_ip.sin6.sin6_addr,
2564 dst_port, src_port,
2565 dst_cache, info);
2566 if (IS_ERR(ndst)) {
2567 err = PTR_ERR(ndst);
2568 ndst = NULL;
2569 goto tx_error;
2570 }
2571
2572 if (!info) {
2573 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2574
2575 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2576 dst_port, ifindex, vni,
2577 ndst, rt6i_flags);
2578 if (err)
2579 goto out_unlock;
2580 }
2581
2582 skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM);
2583
2584 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2585 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2586 skb_scrub_packet(skb, xnet);
2587 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2588 vni, md, flags, udp_sum);
2589 if (err < 0)
2590 goto tx_error;
2591
2592 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2593 &local_ip.sin6.sin6_addr,
2594 &dst->sin6.sin6_addr, tos, ttl,
2595 label, src_port, dst_port, !udp_sum);
2596 #endif
2597 }
2598 out_unlock:
2599 rcu_read_unlock();
2600 return;
2601
2602 drop:
2603 dev->stats.tx_dropped++;
2604 dev_kfree_skb(skb);
2605 return;
2606
2607 tx_error:
2608 rcu_read_unlock();
2609 if (err == -ELOOP)
2610 dev->stats.collisions++;
2611 else if (err == -ENETUNREACH)
2612 dev->stats.tx_carrier_errors++;
2613 dst_release(ndst);
2614 dev->stats.tx_errors++;
2615 kfree_skb(skb);
2616 }
2617
2618 /* Transmit local packets over Vxlan
2619 *
2620 * Outer IP header inherits ECN and DF from inner header.
2621 * Outer UDP destination is the VXLAN assigned port.
2622 * source port is based on hash of flow
2623 */
2624 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2625 {
2626 struct vxlan_dev *vxlan = netdev_priv(dev);
2627 struct vxlan_rdst *rdst, *fdst = NULL;
2628 const struct ip_tunnel_info *info;
2629 bool did_rsc = false;
2630 struct vxlan_fdb *f;
2631 struct ethhdr *eth;
2632 __be32 vni = 0;
2633
2634 info = skb_tunnel_info(skb);
2635
2636 skb_reset_mac_header(skb);
2637
2638 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2639 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2640 info->mode & IP_TUNNEL_INFO_TX) {
2641 vni = tunnel_id_to_key32(info->key.tun_id);
2642 } else {
2643 if (info && info->mode & IP_TUNNEL_INFO_TX)
2644 vxlan_xmit_one(skb, dev, vni, NULL, false);
2645 else
2646 kfree_skb(skb);
2647 return NETDEV_TX_OK;
2648 }
2649 }
2650
2651 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2652 eth = eth_hdr(skb);
2653 if (ntohs(eth->h_proto) == ETH_P_ARP)
2654 return arp_reduce(dev, skb, vni);
2655 #if IS_ENABLED(CONFIG_IPV6)
2656 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2657 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2658 sizeof(struct nd_msg)) &&
2659 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2660 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2661
2662 if (m->icmph.icmp6_code == 0 &&
2663 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2664 return neigh_reduce(dev, skb, vni);
2665 }
2666 #endif
2667 }
2668
2669 eth = eth_hdr(skb);
2670 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2671 did_rsc = false;
2672
2673 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2674 (ntohs(eth->h_proto) == ETH_P_IP ||
2675 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2676 did_rsc = route_shortcircuit(dev, skb);
2677 if (did_rsc)
2678 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2679 }
2680
2681 if (f == NULL) {
2682 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2683 if (f == NULL) {
2684 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2685 !is_multicast_ether_addr(eth->h_dest))
2686 vxlan_fdb_miss(vxlan, eth->h_dest);
2687
2688 dev->stats.tx_dropped++;
2689 kfree_skb(skb);
2690 return NETDEV_TX_OK;
2691 }
2692 }
2693
2694 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2695 struct sk_buff *skb1;
2696
2697 if (!fdst) {
2698 fdst = rdst;
2699 continue;
2700 }
2701 skb1 = skb_clone(skb, GFP_ATOMIC);
2702 if (skb1)
2703 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2704 }
2705
2706 if (fdst)
2707 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2708 else
2709 kfree_skb(skb);
2710 return NETDEV_TX_OK;
2711 }
2712
2713 /* Walk the forwarding table and purge stale entries */
2714 static void vxlan_cleanup(struct timer_list *t)
2715 {
2716 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2717 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2718 unsigned int h;
2719
2720 if (!netif_running(vxlan->dev))
2721 return;
2722
2723 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2724 struct hlist_node *p, *n;
2725
2726 spin_lock(&vxlan->hash_lock[h]);
2727 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2728 struct vxlan_fdb *f
2729 = container_of(p, struct vxlan_fdb, hlist);
2730 unsigned long timeout;
2731
2732 if (f->state & (NUD_PERMANENT | NUD_NOARP))
2733 continue;
2734
2735 if (f->flags & NTF_EXT_LEARNED)
2736 continue;
2737
2738 timeout = f->used + vxlan->cfg.age_interval * HZ;
2739 if (time_before_eq(timeout, jiffies)) {
2740 netdev_dbg(vxlan->dev,
2741 "garbage collect %pM\n",
2742 f->eth_addr);
2743 f->state = NUD_STALE;
2744 vxlan_fdb_destroy(vxlan, f, true, true);
2745 } else if (time_before(timeout, next_timer))
2746 next_timer = timeout;
2747 }
2748 spin_unlock(&vxlan->hash_lock[h]);
2749 }
2750
2751 mod_timer(&vxlan->age_timer, next_timer);
2752 }
2753
2754 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2755 {
2756 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2757
2758 spin_lock(&vn->sock_lock);
2759 hlist_del_init_rcu(&vxlan->hlist4.hlist);
2760 #if IS_ENABLED(CONFIG_IPV6)
2761 hlist_del_init_rcu(&vxlan->hlist6.hlist);
2762 #endif
2763 spin_unlock(&vn->sock_lock);
2764 }
2765
2766 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2767 struct vxlan_dev_node *node)
2768 {
2769 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2770 __be32 vni = vxlan->default_dst.remote_vni;
2771
2772 node->vxlan = vxlan;
2773 spin_lock(&vn->sock_lock);
2774 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2775 spin_unlock(&vn->sock_lock);
2776 }
2777
2778 /* Setup stats when device is created */
2779 static int vxlan_init(struct net_device *dev)
2780 {
2781 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2782 if (!dev->tstats)
2783 return -ENOMEM;
2784
2785 return 0;
2786 }
2787
2788 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
2789 {
2790 struct vxlan_fdb *f;
2791 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
2792
2793 spin_lock_bh(&vxlan->hash_lock[hash_index]);
2794 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
2795 if (f)
2796 vxlan_fdb_destroy(vxlan, f, true, true);
2797 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
2798 }
2799
2800 static void vxlan_uninit(struct net_device *dev)
2801 {
2802 struct vxlan_dev *vxlan = netdev_priv(dev);
2803
2804 gro_cells_destroy(&vxlan->gro_cells);
2805
2806 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
2807
2808 free_percpu(dev->tstats);
2809 }
2810
2811 /* Start ageing timer and join group when device is brought up */
2812 static int vxlan_open(struct net_device *dev)
2813 {
2814 struct vxlan_dev *vxlan = netdev_priv(dev);
2815 int ret;
2816
2817 ret = vxlan_sock_add(vxlan);
2818 if (ret < 0)
2819 return ret;
2820
2821 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
2822 ret = vxlan_igmp_join(vxlan);
2823 if (ret == -EADDRINUSE)
2824 ret = 0;
2825 if (ret) {
2826 vxlan_sock_release(vxlan);
2827 return ret;
2828 }
2829 }
2830
2831 if (vxlan->cfg.age_interval)
2832 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2833
2834 return ret;
2835 }
2836
2837 /* Purge the forwarding table */
2838 static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
2839 {
2840 unsigned int h;
2841
2842 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2843 struct hlist_node *p, *n;
2844
2845 spin_lock_bh(&vxlan->hash_lock[h]);
2846 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2847 struct vxlan_fdb *f
2848 = container_of(p, struct vxlan_fdb, hlist);
2849 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2850 continue;
2851 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2852 if (!is_zero_ether_addr(f->eth_addr))
2853 vxlan_fdb_destroy(vxlan, f, true, true);
2854 }
2855 spin_unlock_bh(&vxlan->hash_lock[h]);
2856 }
2857 }
2858
2859 /* Cleanup timer and forwarding table on shutdown */
2860 static int vxlan_stop(struct net_device *dev)
2861 {
2862 struct vxlan_dev *vxlan = netdev_priv(dev);
2863 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2864 int ret = 0;
2865
2866 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2867 !vxlan_group_used(vn, vxlan))
2868 ret = vxlan_igmp_leave(vxlan);
2869
2870 del_timer_sync(&vxlan->age_timer);
2871
2872 vxlan_flush(vxlan, false);
2873 vxlan_sock_release(vxlan);
2874
2875 return ret;
2876 }
2877
2878 /* Stub, nothing needs to be done. */
2879 static void vxlan_set_multicast_list(struct net_device *dev)
2880 {
2881 }
2882
2883 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2884 {
2885 struct vxlan_dev *vxlan = netdev_priv(dev);
2886 struct vxlan_rdst *dst = &vxlan->default_dst;
2887 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2888 dst->remote_ifindex);
2889 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
2890
2891 /* This check is different than dev->max_mtu, because it looks at
2892 * the lowerdev->mtu, rather than the static dev->max_mtu
2893 */
2894 if (lowerdev) {
2895 int max_mtu = lowerdev->mtu -
2896 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2897 if (new_mtu > max_mtu)
2898 return -EINVAL;
2899 }
2900
2901 dev->mtu = new_mtu;
2902 return 0;
2903 }
2904
2905 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2906 {
2907 struct vxlan_dev *vxlan = netdev_priv(dev);
2908 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2909 __be16 sport, dport;
2910
2911 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2912 vxlan->cfg.port_max, true);
2913 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2914
2915 if (ip_tunnel_info_af(info) == AF_INET) {
2916 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2917 struct rtable *rt;
2918
2919 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
2920 info->key.u.ipv4.dst,
2921 &info->key.u.ipv4.src, dport, sport,
2922 &info->dst_cache, info);
2923 if (IS_ERR(rt))
2924 return PTR_ERR(rt);
2925 ip_rt_put(rt);
2926 } else {
2927 #if IS_ENABLED(CONFIG_IPV6)
2928 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2929 struct dst_entry *ndst;
2930
2931 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
2932 info->key.label, &info->key.u.ipv6.dst,
2933 &info->key.u.ipv6.src, dport, sport,
2934 &info->dst_cache, info);
2935 if (IS_ERR(ndst))
2936 return PTR_ERR(ndst);
2937 dst_release(ndst);
2938 #else /* !CONFIG_IPV6 */
2939 return -EPFNOSUPPORT;
2940 #endif
2941 }
2942 info->key.tp_src = sport;
2943 info->key.tp_dst = dport;
2944 return 0;
2945 }
2946
2947 static const struct net_device_ops vxlan_netdev_ether_ops = {
2948 .ndo_init = vxlan_init,
2949 .ndo_uninit = vxlan_uninit,
2950 .ndo_open = vxlan_open,
2951 .ndo_stop = vxlan_stop,
2952 .ndo_start_xmit = vxlan_xmit,
2953 .ndo_get_stats64 = ip_tunnel_get_stats64,
2954 .ndo_set_rx_mode = vxlan_set_multicast_list,
2955 .ndo_change_mtu = vxlan_change_mtu,
2956 .ndo_validate_addr = eth_validate_addr,
2957 .ndo_set_mac_address = eth_mac_addr,
2958 .ndo_fdb_add = vxlan_fdb_add,
2959 .ndo_fdb_del = vxlan_fdb_delete,
2960 .ndo_fdb_dump = vxlan_fdb_dump,
2961 .ndo_fdb_get = vxlan_fdb_get,
2962 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2963 .ndo_change_proto_down = dev_change_proto_down_generic,
2964 };
2965
2966 static const struct net_device_ops vxlan_netdev_raw_ops = {
2967 .ndo_init = vxlan_init,
2968 .ndo_uninit = vxlan_uninit,
2969 .ndo_open = vxlan_open,
2970 .ndo_stop = vxlan_stop,
2971 .ndo_start_xmit = vxlan_xmit,
2972 .ndo_get_stats64 = ip_tunnel_get_stats64,
2973 .ndo_change_mtu = vxlan_change_mtu,
2974 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2975 };
2976
2977 /* Info for udev, that this is a virtual tunnel endpoint */
2978 static struct device_type vxlan_type = {
2979 .name = "vxlan",
2980 };
2981
2982 /* Calls the ndo_udp_tunnel_add of the caller in order to
2983 * supply the listening VXLAN udp ports. Callers are expected
2984 * to implement the ndo_udp_tunnel_add.
2985 */
2986 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
2987 {
2988 struct vxlan_sock *vs;
2989 struct net *net = dev_net(dev);
2990 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2991 unsigned int i;
2992
2993 spin_lock(&vn->sock_lock);
2994 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2995 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2996 unsigned short type;
2997
2998 if (vs->flags & VXLAN_F_GPE)
2999 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3000 else
3001 type = UDP_TUNNEL_TYPE_VXLAN;
3002
3003 if (push)
3004 udp_tunnel_push_rx_port(dev, vs->sock, type);
3005 else
3006 udp_tunnel_drop_rx_port(dev, vs->sock, type);
3007 }
3008 }
3009 spin_unlock(&vn->sock_lock);
3010 }
3011
3012 /* Initialize the device structure. */
3013 static void vxlan_setup(struct net_device *dev)
3014 {
3015 struct vxlan_dev *vxlan = netdev_priv(dev);
3016 unsigned int h;
3017
3018 eth_hw_addr_random(dev);
3019 ether_setup(dev);
3020
3021 dev->needs_free_netdev = true;
3022 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3023
3024 dev->features |= NETIF_F_LLTX;
3025 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
3026 dev->features |= NETIF_F_RXCSUM;
3027 dev->features |= NETIF_F_GSO_SOFTWARE;
3028
3029 dev->vlan_features = dev->features;
3030 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
3031 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3032 netif_keep_dst(dev);
3033 dev->priv_flags |= IFF_NO_QUEUE;
3034
3035 /* MTU range: 68 - 65535 */
3036 dev->min_mtu = ETH_MIN_MTU;
3037 dev->max_mtu = ETH_MAX_MTU;
3038
3039 INIT_LIST_HEAD(&vxlan->next);
3040
3041 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3042
3043 vxlan->dev = dev;
3044
3045 gro_cells_init(&vxlan->gro_cells, dev);
3046
3047 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3048 spin_lock_init(&vxlan->hash_lock[h]);
3049 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3050 }
3051 }
3052
3053 static void vxlan_ether_setup(struct net_device *dev)
3054 {
3055 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3056 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3057 dev->netdev_ops = &vxlan_netdev_ether_ops;
3058 }
3059
3060 static void vxlan_raw_setup(struct net_device *dev)
3061 {
3062 dev->header_ops = NULL;
3063 dev->type = ARPHRD_NONE;
3064 dev->hard_header_len = 0;
3065 dev->addr_len = 0;
3066 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3067 dev->netdev_ops = &vxlan_netdev_raw_ops;
3068 }
3069
3070 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3071 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
3072 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
3073 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
3074 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
3075 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
3076 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
3077 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
3078 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
3079 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
3080 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
3081 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
3082 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
3083 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
3084 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
3085 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
3086 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
3087 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
3088 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
3089 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
3090 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
3091 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
3092 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
3093 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3094 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3095 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
3096 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
3097 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
3098 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
3099 [IFLA_VXLAN_DF] = { .type = NLA_U8 },
3100 };
3101
3102 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3103 struct netlink_ext_ack *extack)
3104 {
3105 if (tb[IFLA_ADDRESS]) {
3106 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3107 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3108 "Provided link layer address is not Ethernet");
3109 return -EINVAL;
3110 }
3111
3112 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3113 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3114 "Provided Ethernet address is not unicast");
3115 return -EADDRNOTAVAIL;
3116 }
3117 }
3118
3119 if (tb[IFLA_MTU]) {
3120 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3121
3122 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3123 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3124 "MTU must be between 68 and 65535");
3125 return -EINVAL;
3126 }
3127 }
3128
3129 if (!data) {
3130 NL_SET_ERR_MSG(extack,
3131 "Required attributes not provided to perform the operation");
3132 return -EINVAL;
3133 }
3134
3135 if (data[IFLA_VXLAN_ID]) {
3136 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3137
3138 if (id >= VXLAN_N_VID) {
3139 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID],
3140 "VXLAN ID must be lower than 16777216");
3141 return -ERANGE;
3142 }
3143 }
3144
3145 if (data[IFLA_VXLAN_PORT_RANGE]) {
3146 const struct ifla_vxlan_port_range *p
3147 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3148
3149 if (ntohs(p->high) < ntohs(p->low)) {
3150 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
3151 "Invalid source port range");
3152 return -EINVAL;
3153 }
3154 }
3155
3156 if (data[IFLA_VXLAN_DF]) {
3157 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3158
3159 if (df < 0 || df > VXLAN_DF_MAX) {
3160 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_DF],
3161 "Invalid DF attribute");
3162 return -EINVAL;
3163 }
3164 }
3165
3166 return 0;
3167 }
3168
3169 static void vxlan_get_drvinfo(struct net_device *netdev,
3170 struct ethtool_drvinfo *drvinfo)
3171 {
3172 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3173 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3174 }
3175
3176 static const struct ethtool_ops vxlan_ethtool_ops = {
3177 .get_drvinfo = vxlan_get_drvinfo,
3178 .get_link = ethtool_op_get_link,
3179 };
3180
3181 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3182 __be16 port, u32 flags, int ifindex)
3183 {
3184 struct socket *sock;
3185 struct udp_port_cfg udp_conf;
3186 int err;
3187
3188 memset(&udp_conf, 0, sizeof(udp_conf));
3189
3190 if (ipv6) {
3191 udp_conf.family = AF_INET6;
3192 udp_conf.use_udp6_rx_checksums =
3193 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3194 udp_conf.ipv6_v6only = 1;
3195 } else {
3196 udp_conf.family = AF_INET;
3197 }
3198
3199 udp_conf.local_udp_port = port;
3200 udp_conf.bind_ifindex = ifindex;
3201
3202 /* Open UDP socket */
3203 err = udp_sock_create(net, &udp_conf, &sock);
3204 if (err < 0)
3205 return ERR_PTR(err);
3206
3207 return sock;
3208 }
3209
3210 /* Create new listen socket if needed */
3211 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3212 __be16 port, u32 flags,
3213 int ifindex)
3214 {
3215 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3216 struct vxlan_sock *vs;
3217 struct socket *sock;
3218 unsigned int h;
3219 struct udp_tunnel_sock_cfg tunnel_cfg;
3220
3221 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3222 if (!vs)
3223 return ERR_PTR(-ENOMEM);
3224
3225 for (h = 0; h < VNI_HASH_SIZE; ++h)
3226 INIT_HLIST_HEAD(&vs->vni_list[h]);
3227
3228 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3229 if (IS_ERR(sock)) {
3230 kfree(vs);
3231 return ERR_CAST(sock);
3232 }
3233
3234 vs->sock = sock;
3235 refcount_set(&vs->refcnt, 1);
3236 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3237
3238 spin_lock(&vn->sock_lock);
3239 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3240 udp_tunnel_notify_add_rx_port(sock,
3241 (vs->flags & VXLAN_F_GPE) ?
3242 UDP_TUNNEL_TYPE_VXLAN_GPE :
3243 UDP_TUNNEL_TYPE_VXLAN);
3244 spin_unlock(&vn->sock_lock);
3245
3246 /* Mark socket as an encapsulation socket. */
3247 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3248 tunnel_cfg.sk_user_data = vs;
3249 tunnel_cfg.encap_type = 1;
3250 tunnel_cfg.encap_rcv = vxlan_rcv;
3251 tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3252 tunnel_cfg.encap_destroy = NULL;
3253 tunnel_cfg.gro_receive = vxlan_gro_receive;
3254 tunnel_cfg.gro_complete = vxlan_gro_complete;
3255
3256 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3257
3258 return vs;
3259 }
3260
3261 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3262 {
3263 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3264 struct vxlan_sock *vs = NULL;
3265 struct vxlan_dev_node *node;
3266 int l3mdev_index = 0;
3267
3268 if (vxlan->cfg.remote_ifindex)
3269 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3270 vxlan->net, vxlan->cfg.remote_ifindex);
3271
3272 if (!vxlan->cfg.no_share) {
3273 spin_lock(&vn->sock_lock);
3274 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3275 vxlan->cfg.dst_port, vxlan->cfg.flags,
3276 l3mdev_index);
3277 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3278 spin_unlock(&vn->sock_lock);
3279 return -EBUSY;
3280 }
3281 spin_unlock(&vn->sock_lock);
3282 }
3283 if (!vs)
3284 vs = vxlan_socket_create(vxlan->net, ipv6,
3285 vxlan->cfg.dst_port, vxlan->cfg.flags,
3286 l3mdev_index);
3287 if (IS_ERR(vs))
3288 return PTR_ERR(vs);
3289 #if IS_ENABLED(CONFIG_IPV6)
3290 if (ipv6) {
3291 rcu_assign_pointer(vxlan->vn6_sock, vs);
3292 node = &vxlan->hlist6;
3293 } else
3294 #endif
3295 {
3296 rcu_assign_pointer(vxlan->vn4_sock, vs);
3297 node = &vxlan->hlist4;
3298 }
3299 vxlan_vs_add_dev(vs, vxlan, node);
3300 return 0;
3301 }
3302
3303 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3304 {
3305 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3306 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3307 bool ipv4 = !ipv6 || metadata;
3308 int ret = 0;
3309
3310 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3311 #if IS_ENABLED(CONFIG_IPV6)
3312 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3313 if (ipv6) {
3314 ret = __vxlan_sock_add(vxlan, true);
3315 if (ret < 0 && ret != -EAFNOSUPPORT)
3316 ipv4 = false;
3317 }
3318 #endif
3319 if (ipv4)
3320 ret = __vxlan_sock_add(vxlan, false);
3321 if (ret < 0)
3322 vxlan_sock_release(vxlan);
3323 return ret;
3324 }
3325
3326 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3327 struct net_device **lower,
3328 struct vxlan_dev *old,
3329 struct netlink_ext_ack *extack)
3330 {
3331 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3332 struct vxlan_dev *tmp;
3333 bool use_ipv6 = false;
3334
3335 if (conf->flags & VXLAN_F_GPE) {
3336 /* For now, allow GPE only together with
3337 * COLLECT_METADATA. This can be relaxed later; in such
3338 * case, the other side of the PtP link will have to be
3339 * provided.
3340 */
3341 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3342 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3343 NL_SET_ERR_MSG(extack,
3344 "VXLAN GPE does not support this combination of attributes");
3345 return -EINVAL;
3346 }
3347 }
3348
3349 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3350 /* Unless IPv6 is explicitly requested, assume IPv4 */
3351 conf->remote_ip.sa.sa_family = AF_INET;
3352 conf->saddr.sa.sa_family = AF_INET;
3353 } else if (!conf->remote_ip.sa.sa_family) {
3354 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3355 } else if (!conf->saddr.sa.sa_family) {
3356 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3357 }
3358
3359 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3360 NL_SET_ERR_MSG(extack,
3361 "Local and remote address must be from the same family");
3362 return -EINVAL;
3363 }
3364
3365 if (vxlan_addr_multicast(&conf->saddr)) {
3366 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3367 return -EINVAL;
3368 }
3369
3370 if (conf->saddr.sa.sa_family == AF_INET6) {
3371 if (!IS_ENABLED(CONFIG_IPV6)) {
3372 NL_SET_ERR_MSG(extack,
3373 "IPv6 support not enabled in the kernel");
3374 return -EPFNOSUPPORT;
3375 }
3376 use_ipv6 = true;
3377 conf->flags |= VXLAN_F_IPV6;
3378
3379 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3380 int local_type =
3381 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3382 int remote_type =
3383 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3384
3385 if (local_type & IPV6_ADDR_LINKLOCAL) {
3386 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3387 (remote_type != IPV6_ADDR_ANY)) {
3388 NL_SET_ERR_MSG(extack,
3389 "Invalid combination of local and remote address scopes");
3390 return -EINVAL;
3391 }
3392
3393 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3394 } else {
3395 if (remote_type ==
3396 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3397 NL_SET_ERR_MSG(extack,
3398 "Invalid combination of local and remote address scopes");
3399 return -EINVAL;
3400 }
3401
3402 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3403 }
3404 }
3405 }
3406
3407 if (conf->label && !use_ipv6) {
3408 NL_SET_ERR_MSG(extack,
3409 "Label attribute only applies to IPv6 VXLAN devices");
3410 return -EINVAL;
3411 }
3412
3413 if (conf->remote_ifindex) {
3414 struct net_device *lowerdev;
3415
3416 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3417 if (!lowerdev) {
3418 NL_SET_ERR_MSG(extack,
3419 "Invalid local interface, device not found");
3420 return -ENODEV;
3421 }
3422
3423 #if IS_ENABLED(CONFIG_IPV6)
3424 if (use_ipv6) {
3425 struct inet6_dev *idev = __in6_dev_get(lowerdev);
3426 if (idev && idev->cnf.disable_ipv6) {
3427 NL_SET_ERR_MSG(extack,
3428 "IPv6 support disabled by administrator");
3429 return -EPERM;
3430 }
3431 }
3432 #endif
3433
3434 *lower = lowerdev;
3435 } else {
3436 if (vxlan_addr_multicast(&conf->remote_ip)) {
3437 NL_SET_ERR_MSG(extack,
3438 "Local interface required for multicast remote destination");
3439
3440 return -EINVAL;
3441 }
3442
3443 #if IS_ENABLED(CONFIG_IPV6)
3444 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3445 NL_SET_ERR_MSG(extack,
3446 "Local interface required for link-local local/remote addresses");
3447 return -EINVAL;
3448 }
3449 #endif
3450
3451 *lower = NULL;
3452 }
3453
3454 if (!conf->dst_port) {
3455 if (conf->flags & VXLAN_F_GPE)
3456 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3457 else
3458 conf->dst_port = htons(vxlan_port);
3459 }
3460
3461 if (!conf->age_interval)
3462 conf->age_interval = FDB_AGE_DEFAULT;
3463
3464 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3465 if (tmp == old)
3466 continue;
3467
3468 if (tmp->cfg.vni != conf->vni)
3469 continue;
3470 if (tmp->cfg.dst_port != conf->dst_port)
3471 continue;
3472 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3473 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3474 continue;
3475
3476 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3477 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3478 continue;
3479
3480 NL_SET_ERR_MSG(extack,
3481 "A VXLAN device with the specified VNI already exists");
3482 return -EEXIST;
3483 }
3484
3485 return 0;
3486 }
3487
3488 static void vxlan_config_apply(struct net_device *dev,
3489 struct vxlan_config *conf,
3490 struct net_device *lowerdev,
3491 struct net *src_net,
3492 bool changelink)
3493 {
3494 struct vxlan_dev *vxlan = netdev_priv(dev);
3495 struct vxlan_rdst *dst = &vxlan->default_dst;
3496 unsigned short needed_headroom = ETH_HLEN;
3497 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3498 int max_mtu = ETH_MAX_MTU;
3499
3500 if (!changelink) {
3501 if (conf->flags & VXLAN_F_GPE)
3502 vxlan_raw_setup(dev);
3503 else
3504 vxlan_ether_setup(dev);
3505
3506 if (conf->mtu)
3507 dev->mtu = conf->mtu;
3508
3509 vxlan->net = src_net;
3510 }
3511
3512 dst->remote_vni = conf->vni;
3513
3514 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3515
3516 if (lowerdev) {
3517 dst->remote_ifindex = conf->remote_ifindex;
3518
3519 dev->gso_max_size = lowerdev->gso_max_size;
3520 dev->gso_max_segs = lowerdev->gso_max_segs;
3521
3522 needed_headroom = lowerdev->hard_header_len;
3523
3524 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3525 VXLAN_HEADROOM);
3526 if (max_mtu < ETH_MIN_MTU)
3527 max_mtu = ETH_MIN_MTU;
3528
3529 if (!changelink && !conf->mtu)
3530 dev->mtu = max_mtu;
3531 }
3532
3533 if (dev->mtu > max_mtu)
3534 dev->mtu = max_mtu;
3535
3536 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3537 needed_headroom += VXLAN6_HEADROOM;
3538 else
3539 needed_headroom += VXLAN_HEADROOM;
3540 dev->needed_headroom = needed_headroom;
3541
3542 memcpy(&vxlan->cfg, conf, sizeof(*conf));
3543 }
3544
3545 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3546 struct vxlan_config *conf, bool changelink,
3547 struct netlink_ext_ack *extack)
3548 {
3549 struct vxlan_dev *vxlan = netdev_priv(dev);
3550 struct net_device *lowerdev;
3551 int ret;
3552
3553 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3554 if (ret)
3555 return ret;
3556
3557 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3558
3559 return 0;
3560 }
3561
3562 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3563 struct vxlan_config *conf,
3564 struct netlink_ext_ack *extack)
3565 {
3566 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3567 struct vxlan_dev *vxlan = netdev_priv(dev);
3568 struct vxlan_fdb *f = NULL;
3569 bool unregister = false;
3570 int err;
3571
3572 err = vxlan_dev_configure(net, dev, conf, false, extack);
3573 if (err)
3574 return err;
3575
3576 dev->ethtool_ops = &vxlan_ethtool_ops;
3577
3578 /* create an fdb entry for a valid default destination */
3579 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
3580 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3581 &vxlan->default_dst.remote_ip,
3582 NUD_REACHABLE | NUD_PERMANENT,
3583 vxlan->cfg.dst_port,
3584 vxlan->default_dst.remote_vni,
3585 vxlan->default_dst.remote_vni,
3586 vxlan->default_dst.remote_ifindex,
3587 NTF_SELF, &f);
3588 if (err)
3589 return err;
3590 }
3591
3592 err = register_netdevice(dev);
3593 if (err)
3594 goto errout;
3595 unregister = true;
3596
3597 err = rtnl_configure_link(dev, NULL);
3598 if (err)
3599 goto errout;
3600
3601 if (f) {
3602 vxlan_fdb_insert(vxlan, all_zeros_mac,
3603 vxlan->default_dst.remote_vni, f);
3604
3605 /* notify default fdb entry */
3606 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
3607 RTM_NEWNEIGH, true, extack);
3608 if (err) {
3609 vxlan_fdb_destroy(vxlan, f, false, false);
3610 goto unregister;
3611 }
3612 }
3613
3614 list_add(&vxlan->next, &vn->vxlan_list);
3615 return 0;
3616
3617 errout:
3618 /* unregister_netdevice() destroys the default FDB entry with deletion
3619 * notification. But the addition notification was not sent yet, so
3620 * destroy the entry by hand here.
3621 */
3622 if (f)
3623 __vxlan_fdb_free(f);
3624 unregister:
3625 if (unregister)
3626 unregister_netdevice(dev);
3627 return err;
3628 }
3629
3630 /* Set/clear flags based on attribute */
3631 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3632 int attrtype, unsigned long mask, bool changelink,
3633 bool changelink_supported,
3634 struct netlink_ext_ack *extack)
3635 {
3636 unsigned long flags;
3637
3638 if (!tb[attrtype])
3639 return 0;
3640
3641 if (changelink && !changelink_supported) {
3642 vxlan_flag_attr_error(attrtype, extack);
3643 return -EOPNOTSUPP;
3644 }
3645
3646 if (vxlan_policy[attrtype].type == NLA_FLAG)
3647 flags = conf->flags | mask;
3648 else if (nla_get_u8(tb[attrtype]))
3649 flags = conf->flags | mask;
3650 else
3651 flags = conf->flags & ~mask;
3652
3653 conf->flags = flags;
3654
3655 return 0;
3656 }
3657
3658 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3659 struct net_device *dev, struct vxlan_config *conf,
3660 bool changelink, struct netlink_ext_ack *extack)
3661 {
3662 struct vxlan_dev *vxlan = netdev_priv(dev);
3663 int err = 0;
3664
3665 memset(conf, 0, sizeof(*conf));
3666
3667 /* if changelink operation, start with old existing cfg */
3668 if (changelink)
3669 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3670
3671 if (data[IFLA_VXLAN_ID]) {
3672 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3673
3674 if (changelink && (vni != conf->vni)) {
3675 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
3676 return -EOPNOTSUPP;
3677 }
3678 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3679 }
3680
3681 if (data[IFLA_VXLAN_GROUP]) {
3682 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
3683 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
3684 return -EOPNOTSUPP;
3685 }
3686
3687 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
3688 conf->remote_ip.sa.sa_family = AF_INET;
3689 } else if (data[IFLA_VXLAN_GROUP6]) {
3690 if (!IS_ENABLED(CONFIG_IPV6)) {
3691 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
3692 return -EPFNOSUPPORT;
3693 }
3694
3695 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
3696 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
3697 return -EOPNOTSUPP;
3698 }
3699
3700 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3701 conf->remote_ip.sa.sa_family = AF_INET6;
3702 }
3703
3704 if (data[IFLA_VXLAN_LOCAL]) {
3705 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
3706 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
3707 return -EOPNOTSUPP;
3708 }
3709
3710 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3711 conf->saddr.sa.sa_family = AF_INET;
3712 } else if (data[IFLA_VXLAN_LOCAL6]) {
3713 if (!IS_ENABLED(CONFIG_IPV6)) {
3714 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
3715 return -EPFNOSUPPORT;
3716 }
3717
3718 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
3719 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
3720 return -EOPNOTSUPP;
3721 }
3722
3723 /* TODO: respect scope id */
3724 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3725 conf->saddr.sa.sa_family = AF_INET6;
3726 }
3727
3728 if (data[IFLA_VXLAN_LINK])
3729 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
3730
3731 if (data[IFLA_VXLAN_TOS])
3732 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
3733
3734 if (data[IFLA_VXLAN_TTL])
3735 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
3736
3737 if (data[IFLA_VXLAN_TTL_INHERIT]) {
3738 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
3739 VXLAN_F_TTL_INHERIT, changelink, false,
3740 extack);
3741 if (err)
3742 return err;
3743
3744 }
3745
3746 if (data[IFLA_VXLAN_LABEL])
3747 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
3748 IPV6_FLOWLABEL_MASK;
3749
3750 if (data[IFLA_VXLAN_LEARNING]) {
3751 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
3752 VXLAN_F_LEARN, changelink, true,
3753 extack);
3754 if (err)
3755 return err;
3756 } else if (!changelink) {
3757 /* default to learn on a new device */
3758 conf->flags |= VXLAN_F_LEARN;
3759 }
3760
3761 if (data[IFLA_VXLAN_AGEING])
3762 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3763
3764 if (data[IFLA_VXLAN_PROXY]) {
3765 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
3766 VXLAN_F_PROXY, changelink, false,
3767 extack);
3768 if (err)
3769 return err;
3770 }
3771
3772 if (data[IFLA_VXLAN_RSC]) {
3773 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
3774 VXLAN_F_RSC, changelink, false,
3775 extack);
3776 if (err)
3777 return err;
3778 }
3779
3780 if (data[IFLA_VXLAN_L2MISS]) {
3781 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
3782 VXLAN_F_L2MISS, changelink, false,
3783 extack);
3784 if (err)
3785 return err;
3786 }
3787
3788 if (data[IFLA_VXLAN_L3MISS]) {
3789 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
3790 VXLAN_F_L3MISS, changelink, false,
3791 extack);
3792 if (err)
3793 return err;
3794 }
3795
3796 if (data[IFLA_VXLAN_LIMIT]) {
3797 if (changelink) {
3798 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
3799 "Cannot change limit");
3800 return -EOPNOTSUPP;
3801 }
3802 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3803 }
3804
3805 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3806 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
3807 VXLAN_F_COLLECT_METADATA, changelink, false,
3808 extack);
3809 if (err)
3810 return err;
3811 }
3812
3813 if (data[IFLA_VXLAN_PORT_RANGE]) {
3814 if (!changelink) {
3815 const struct ifla_vxlan_port_range *p
3816 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3817 conf->port_min = ntohs(p->low);
3818 conf->port_max = ntohs(p->high);
3819 } else {
3820 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
3821 "Cannot change port range");
3822 return -EOPNOTSUPP;
3823 }
3824 }
3825
3826 if (data[IFLA_VXLAN_PORT]) {
3827 if (changelink) {
3828 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
3829 "Cannot change port");
3830 return -EOPNOTSUPP;
3831 }
3832 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3833 }
3834
3835 if (data[IFLA_VXLAN_UDP_CSUM]) {
3836 if (changelink) {
3837 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
3838 "Cannot change UDP_CSUM flag");
3839 return -EOPNOTSUPP;
3840 }
3841 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3842 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3843 }
3844
3845 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3846 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3847 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
3848 false, extack);
3849 if (err)
3850 return err;
3851 }
3852
3853 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3854 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
3855 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
3856 false, extack);
3857 if (err)
3858 return err;
3859 }
3860
3861 if (data[IFLA_VXLAN_REMCSUM_TX]) {
3862 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
3863 VXLAN_F_REMCSUM_TX, changelink, false,
3864 extack);
3865 if (err)
3866 return err;
3867 }
3868
3869 if (data[IFLA_VXLAN_REMCSUM_RX]) {
3870 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
3871 VXLAN_F_REMCSUM_RX, changelink, false,
3872 extack);
3873 if (err)
3874 return err;
3875 }
3876
3877 if (data[IFLA_VXLAN_GBP]) {
3878 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
3879 VXLAN_F_GBP, changelink, false, extack);
3880 if (err)
3881 return err;
3882 }
3883
3884 if (data[IFLA_VXLAN_GPE]) {
3885 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
3886 VXLAN_F_GPE, changelink, false,
3887 extack);
3888 if (err)
3889 return err;
3890 }
3891
3892 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3893 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
3894 VXLAN_F_REMCSUM_NOPARTIAL, changelink,
3895 false, extack);
3896 if (err)
3897 return err;
3898 }
3899
3900 if (tb[IFLA_MTU]) {
3901 if (changelink) {
3902 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3903 "Cannot change mtu");
3904 return -EOPNOTSUPP;
3905 }
3906 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3907 }
3908
3909 if (data[IFLA_VXLAN_DF])
3910 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
3911
3912 return 0;
3913 }
3914
3915 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
3916 struct nlattr *tb[], struct nlattr *data[],
3917 struct netlink_ext_ack *extack)
3918 {
3919 struct vxlan_config conf;
3920 int err;
3921
3922 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
3923 if (err)
3924 return err;
3925
3926 return __vxlan_dev_create(src_net, dev, &conf, extack);
3927 }
3928
3929 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
3930 struct nlattr *data[],
3931 struct netlink_ext_ack *extack)
3932 {
3933 struct vxlan_dev *vxlan = netdev_priv(dev);
3934 struct vxlan_rdst *dst = &vxlan->default_dst;
3935 struct net_device *lowerdev;
3936 struct vxlan_config conf;
3937 int err;
3938
3939 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
3940 if (err)
3941 return err;
3942
3943 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
3944 vxlan, extack);
3945 if (err)
3946 return err;
3947
3948 /* handle default dst entry */
3949 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
3950 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
3951
3952 spin_lock_bh(&vxlan->hash_lock[hash_index]);
3953 if (!vxlan_addr_any(&conf.remote_ip)) {
3954 err = vxlan_fdb_update(vxlan, all_zeros_mac,
3955 &conf.remote_ip,
3956 NUD_REACHABLE | NUD_PERMANENT,
3957 NLM_F_APPEND | NLM_F_CREATE,
3958 vxlan->cfg.dst_port,
3959 conf.vni, conf.vni,
3960 conf.remote_ifindex,
3961 NTF_SELF, true, extack);
3962 if (err) {
3963 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
3964 return err;
3965 }
3966 }
3967 if (!vxlan_addr_any(&dst->remote_ip))
3968 __vxlan_fdb_delete(vxlan, all_zeros_mac,
3969 dst->remote_ip,
3970 vxlan->cfg.dst_port,
3971 dst->remote_vni,
3972 dst->remote_vni,
3973 dst->remote_ifindex,
3974 true);
3975 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
3976 }
3977
3978 if (conf.age_interval != vxlan->cfg.age_interval)
3979 mod_timer(&vxlan->age_timer, jiffies);
3980
3981 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
3982 return 0;
3983 }
3984
3985 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3986 {
3987 struct vxlan_dev *vxlan = netdev_priv(dev);
3988
3989 vxlan_flush(vxlan, true);
3990
3991 list_del(&vxlan->next);
3992 unregister_netdevice_queue(dev, head);
3993 }
3994
3995 static size_t vxlan_get_size(const struct net_device *dev)
3996 {
3997
3998 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
3999 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4000 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
4001 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4002 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
4003 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
4004 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
4005 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */
4006 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4007 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
4008 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
4009 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
4010 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
4011 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
4012 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
4013 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
4014 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
4015 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4016 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4017 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4018 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4019 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4020 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4021 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4022 0;
4023 }
4024
4025 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4026 {
4027 const struct vxlan_dev *vxlan = netdev_priv(dev);
4028 const struct vxlan_rdst *dst = &vxlan->default_dst;
4029 struct ifla_vxlan_port_range ports = {
4030 .low = htons(vxlan->cfg.port_min),
4031 .high = htons(vxlan->cfg.port_max),
4032 };
4033
4034 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4035 goto nla_put_failure;
4036
4037 if (!vxlan_addr_any(&dst->remote_ip)) {
4038 if (dst->remote_ip.sa.sa_family == AF_INET) {
4039 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4040 dst->remote_ip.sin.sin_addr.s_addr))
4041 goto nla_put_failure;
4042 #if IS_ENABLED(CONFIG_IPV6)
4043 } else {
4044 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4045 &dst->remote_ip.sin6.sin6_addr))
4046 goto nla_put_failure;
4047 #endif
4048 }
4049 }
4050
4051 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4052 goto nla_put_failure;
4053
4054 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4055 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4056 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4057 vxlan->cfg.saddr.sin.sin_addr.s_addr))
4058 goto nla_put_failure;
4059 #if IS_ENABLED(CONFIG_IPV6)
4060 } else {
4061 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4062 &vxlan->cfg.saddr.sin6.sin6_addr))
4063 goto nla_put_failure;
4064 #endif
4065 }
4066 }
4067
4068 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4069 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4070 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4071 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4072 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4073 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4074 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4075 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4076 nla_put_u8(skb, IFLA_VXLAN_PROXY,
4077 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4078 nla_put_u8(skb, IFLA_VXLAN_RSC,
4079 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4080 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4081 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4082 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4083 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4084 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4085 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4086 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4087 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4088 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4089 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4090 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4091 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4092 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4093 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4094 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4095 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4096 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4097 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4098 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
4099 goto nla_put_failure;
4100
4101 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4102 goto nla_put_failure;
4103
4104 if (vxlan->cfg.flags & VXLAN_F_GBP &&
4105 nla_put_flag(skb, IFLA_VXLAN_GBP))
4106 goto nla_put_failure;
4107
4108 if (vxlan->cfg.flags & VXLAN_F_GPE &&
4109 nla_put_flag(skb, IFLA_VXLAN_GPE))
4110 goto nla_put_failure;
4111
4112 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4113 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4114 goto nla_put_failure;
4115
4116 return 0;
4117
4118 nla_put_failure:
4119 return -EMSGSIZE;
4120 }
4121
4122 static struct net *vxlan_get_link_net(const struct net_device *dev)
4123 {
4124 struct vxlan_dev *vxlan = netdev_priv(dev);
4125
4126 return vxlan->net;
4127 }
4128
4129 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4130 .kind = "vxlan",
4131 .maxtype = IFLA_VXLAN_MAX,
4132 .policy = vxlan_policy,
4133 .priv_size = sizeof(struct vxlan_dev),
4134 .setup = vxlan_setup,
4135 .validate = vxlan_validate,
4136 .newlink = vxlan_newlink,
4137 .changelink = vxlan_changelink,
4138 .dellink = vxlan_dellink,
4139 .get_size = vxlan_get_size,
4140 .fill_info = vxlan_fill_info,
4141 .get_link_net = vxlan_get_link_net,
4142 };
4143
4144 struct net_device *vxlan_dev_create(struct net *net, const char *name,
4145 u8 name_assign_type,
4146 struct vxlan_config *conf)
4147 {
4148 struct nlattr *tb[IFLA_MAX + 1];
4149 struct net_device *dev;
4150 int err;
4151
4152 memset(&tb, 0, sizeof(tb));
4153
4154 dev = rtnl_create_link(net, name, name_assign_type,
4155 &vxlan_link_ops, tb, NULL);
4156 if (IS_ERR(dev))
4157 return dev;
4158
4159 err = __vxlan_dev_create(net, dev, conf, NULL);
4160 if (err < 0) {
4161 free_netdev(dev);
4162 return ERR_PTR(err);
4163 }
4164
4165 err = rtnl_configure_link(dev, NULL);
4166 if (err < 0) {
4167 LIST_HEAD(list_kill);
4168
4169 vxlan_dellink(dev, &list_kill);
4170 unregister_netdevice_many(&list_kill);
4171 return ERR_PTR(err);
4172 }
4173
4174 return dev;
4175 }
4176 EXPORT_SYMBOL_GPL(vxlan_dev_create);
4177
4178 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4179 struct net_device *dev)
4180 {
4181 struct vxlan_dev *vxlan, *next;
4182 LIST_HEAD(list_kill);
4183
4184 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4185 struct vxlan_rdst *dst = &vxlan->default_dst;
4186
4187 /* In case we created vxlan device with carrier
4188 * and we loose the carrier due to module unload
4189 * we also need to remove vxlan device. In other
4190 * cases, it's not necessary and remote_ifindex
4191 * is 0 here, so no matches.
4192 */
4193 if (dst->remote_ifindex == dev->ifindex)
4194 vxlan_dellink(vxlan->dev, &list_kill);
4195 }
4196
4197 unregister_netdevice_many(&list_kill);
4198 }
4199
4200 static int vxlan_netdevice_event(struct notifier_block *unused,
4201 unsigned long event, void *ptr)
4202 {
4203 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4204 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4205
4206 if (event == NETDEV_UNREGISTER) {
4207 vxlan_offload_rx_ports(dev, false);
4208 vxlan_handle_lowerdev_unregister(vn, dev);
4209 } else if (event == NETDEV_REGISTER) {
4210 vxlan_offload_rx_ports(dev, true);
4211 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
4212 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
4213 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
4214 }
4215
4216 return NOTIFY_DONE;
4217 }
4218
4219 static struct notifier_block vxlan_notifier_block __read_mostly = {
4220 .notifier_call = vxlan_netdevice_event,
4221 };
4222
4223 static void
4224 vxlan_fdb_offloaded_set(struct net_device *dev,
4225 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4226 {
4227 struct vxlan_dev *vxlan = netdev_priv(dev);
4228 struct vxlan_rdst *rdst;
4229 struct vxlan_fdb *f;
4230 u32 hash_index;
4231
4232 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4233
4234 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4235
4236 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4237 if (!f)
4238 goto out;
4239
4240 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4241 fdb_info->remote_port,
4242 fdb_info->remote_vni,
4243 fdb_info->remote_ifindex);
4244 if (!rdst)
4245 goto out;
4246
4247 rdst->offloaded = fdb_info->offloaded;
4248
4249 out:
4250 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4251 }
4252
4253 static int
4254 vxlan_fdb_external_learn_add(struct net_device *dev,
4255 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4256 {
4257 struct vxlan_dev *vxlan = netdev_priv(dev);
4258 struct netlink_ext_ack *extack;
4259 u32 hash_index;
4260 int err;
4261
4262 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4263 extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4264
4265 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4266 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4267 NUD_REACHABLE,
4268 NLM_F_CREATE | NLM_F_REPLACE,
4269 fdb_info->remote_port,
4270 fdb_info->vni,
4271 fdb_info->remote_vni,
4272 fdb_info->remote_ifindex,
4273 NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4274 false, extack);
4275 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4276
4277 return err;
4278 }
4279
4280 static int
4281 vxlan_fdb_external_learn_del(struct net_device *dev,
4282 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4283 {
4284 struct vxlan_dev *vxlan = netdev_priv(dev);
4285 struct vxlan_fdb *f;
4286 u32 hash_index;
4287 int err = 0;
4288
4289 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4290 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4291
4292 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4293 if (!f)
4294 err = -ENOENT;
4295 else if (f->flags & NTF_EXT_LEARNED)
4296 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4297 fdb_info->remote_ip,
4298 fdb_info->remote_port,
4299 fdb_info->vni,
4300 fdb_info->remote_vni,
4301 fdb_info->remote_ifindex,
4302 false);
4303
4304 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4305
4306 return err;
4307 }
4308
4309 static int vxlan_switchdev_event(struct notifier_block *unused,
4310 unsigned long event, void *ptr)
4311 {
4312 struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4313 struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4314 int err = 0;
4315
4316 switch (event) {
4317 case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4318 vxlan_fdb_offloaded_set(dev, ptr);
4319 break;
4320 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4321 fdb_info = ptr;
4322 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4323 if (err) {
4324 err = notifier_from_errno(err);
4325 break;
4326 }
4327 fdb_info->offloaded = true;
4328 vxlan_fdb_offloaded_set(dev, fdb_info);
4329 break;
4330 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4331 fdb_info = ptr;
4332 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4333 if (err) {
4334 err = notifier_from_errno(err);
4335 break;
4336 }
4337 fdb_info->offloaded = false;
4338 vxlan_fdb_offloaded_set(dev, fdb_info);
4339 break;
4340 }
4341
4342 return err;
4343 }
4344
4345 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4346 .notifier_call = vxlan_switchdev_event,
4347 };
4348
4349 static __net_init int vxlan_init_net(struct net *net)
4350 {
4351 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4352 unsigned int h;
4353
4354 INIT_LIST_HEAD(&vn->vxlan_list);
4355 spin_lock_init(&vn->sock_lock);
4356
4357 for (h = 0; h < PORT_HASH_SIZE; ++h)
4358 INIT_HLIST_HEAD(&vn->sock_list[h]);
4359
4360 return 0;
4361 }
4362
4363 static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
4364 {
4365 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4366 struct vxlan_dev *vxlan, *next;
4367 struct net_device *dev, *aux;
4368 unsigned int h;
4369
4370 for_each_netdev_safe(net, dev, aux)
4371 if (dev->rtnl_link_ops == &vxlan_link_ops)
4372 unregister_netdevice_queue(dev, head);
4373
4374 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4375 /* If vxlan->dev is in the same netns, it has already been added
4376 * to the list by the previous loop.
4377 */
4378 if (!net_eq(dev_net(vxlan->dev), net))
4379 unregister_netdevice_queue(vxlan->dev, head);
4380 }
4381
4382 for (h = 0; h < PORT_HASH_SIZE; ++h)
4383 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4384 }
4385
4386 static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4387 {
4388 struct net *net;
4389 LIST_HEAD(list);
4390
4391 rtnl_lock();
4392 list_for_each_entry(net, net_list, exit_list)
4393 vxlan_destroy_tunnels(net, &list);
4394
4395 unregister_netdevice_many(&list);
4396 rtnl_unlock();
4397 }
4398
4399 static struct pernet_operations vxlan_net_ops = {
4400 .init = vxlan_init_net,
4401 .exit_batch = vxlan_exit_batch_net,
4402 .id = &vxlan_net_id,
4403 .size = sizeof(struct vxlan_net),
4404 };
4405
4406 static int __init vxlan_init_module(void)
4407 {
4408 int rc;
4409
4410 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4411
4412 rc = register_pernet_subsys(&vxlan_net_ops);
4413 if (rc)
4414 goto out1;
4415
4416 rc = register_netdevice_notifier(&vxlan_notifier_block);
4417 if (rc)
4418 goto out2;
4419
4420 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4421 if (rc)
4422 goto out3;
4423
4424 rc = rtnl_link_register(&vxlan_link_ops);
4425 if (rc)
4426 goto out4;
4427
4428 return 0;
4429 out4:
4430 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4431 out3:
4432 unregister_netdevice_notifier(&vxlan_notifier_block);
4433 out2:
4434 unregister_pernet_subsys(&vxlan_net_ops);
4435 out1:
4436 return rc;
4437 }
4438 late_initcall(vxlan_init_module);
4439
4440 static void __exit vxlan_cleanup_module(void)
4441 {
4442 rtnl_link_unregister(&vxlan_link_ops);
4443 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4444 unregister_netdevice_notifier(&vxlan_notifier_block);
4445 unregister_pernet_subsys(&vxlan_net_ops);
4446 /* rcu_barrier() is called by netns */
4447 }
4448 module_exit(vxlan_cleanup_module);
4449
4450 MODULE_LICENSE("GPL");
4451 MODULE_VERSION(VXLAN_VERSION);
4452 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4453 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4454 MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Jammy Linux kernel mirror