]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-hirsute-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
UBUNTU: Ubuntu-5.11.0-22.23
[mirror_ubuntu-hirsute-kernel.git] / drivers / net / vxlan.c
CommitLineData
d2912cb1 1// SPDX-License-Identifier: GPL-2.0-only
d342894c 2/*
eb5ce439 3 * VXLAN: Virtual eXtensible Local Area Network
d342894c 4 *
3b8df3c6 5 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 6 */
7
8#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
10#include <linux/kernel.h>
d342894c 11#include <linux/module.h>
12#include <linux/errno.h>
13#include <linux/slab.h>
d342894c 14#include <linux/udp.h>
15#include <linux/igmp.h>
4f94c353 16#include <linux/inetdevice.h>
d342894c 17#include <linux/if_ether.h>
1b13c97f 18#include <linux/ethtool.h>
e4f67add
DS		 19#include <net/arp.h>
20#include <net/ndisc.h>
3616d08b 21#include <net/ipv6_stubs.h>
d342894c 22#include <net/ip.h>
23#include <net/icmp.h>
d342894c 24#include <net/rtnetlink.h>
d342894c 25#include <net/inet_ecn.h>
26#include <net/net_namespace.h>
27#include <net/netns/generic.h>
fa20e0e3 28#include <net/tun_proto.h>
012a5729 29#include <net/vxlan.h>
1274e1cc 30#include <net/nexthop.h>
40d29af0 31
e4c7ed41 32#if IS_ENABLED(CONFIG_IPV6)
e4c7ed41 33#include <net/ip6_tunnel.h>
660d98ca 34#include <net/ip6_checksum.h>
e4c7ed41 35#endif
d342894c 36
37#define VXLAN_VERSION "0.1"
38
553675fb 39#define PORT_HASH_BITS 8
40#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 41#define FDB_AGE_DEFAULT 300 /* 5 min */
42#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
43
23c578bf 44/* UDP port for VXLAN traffic.
45 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 46 * for compatibility with early adopters.
23c578bf 47 */
9daaa397
SH		 48static unsigned short vxlan_port __read_mostly = 8472;
49module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 50MODULE_PARM_DESC(udp_port, "Destination UDP port");
51
52static bool log_ecn_error = true;
53module_param(log_ecn_error, bool, 0644);
54MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
55
c7d03a00 56static unsigned int vxlan_net_id;
0dfbdf41 57static struct rtnl_link_ops vxlan_link_ops;
553675fb 58
7256eac1 59static const u8 all_zeros_mac[ETH_ALEN + 2];
afbd8bae 60
205f356d 61static int vxlan_sock_add(struct vxlan_dev *vxlan);
614732ea 62
a53cb29b
MB		 63static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
64
553675fb 65/* per-network namespace private data for this module */
66struct vxlan_net {
67 struct list_head vxlan_list;
68 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 69 spinlock_t sock_lock;
626d667b 70 struct notifier_block nexthop_notifier_block;
553675fb 71};
72
d342894c 73/* Forwarding table entry */
74struct vxlan_fdb {
75 struct hlist_node hlist; /* linked list of entries */
76 struct rcu_head rcu;
77 unsigned long updated; /* jiffies */
78 unsigned long used;
3e61aa8f 79 struct list_head remotes;
7177a3b0 80 u8 eth_addr[ETH_ALEN];
d342894c 81 u16 state; /* see ndm_state */
3ad7a4b1 82 __be32 vni;
45598c1c 83 u16 flags; /* see ndm_flags and below */
1274e1cc
RP		 84 struct list_head nh_list;
85 struct nexthop __rcu *nh;
79472fe8 86 struct vxlan_dev __rcu *vdev;
d342894c 87};
88
45598c1c
PM		 89#define NTF_VXLAN_ADDED_BY_USER 0x100
90
d342894c 91/* salt for hash table */
92static u32 vxlan_salt __read_mostly;
93
ee122c79
TG		 94static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
95{
e7030878
TG		 96 return vs->flags & VXLAN_F_COLLECT_METADATA ||
97 ip_tunnel_collect_metadata();
ee122c79
TG		 98}
99
4f94c353
JV		 100static struct ip_fan_map *vxlan_fan_find_map(struct vxlan_dev *vxlan, __be32 daddr)
101{
102 struct ip_fan_map *fan_map;
103
104 rcu_read_lock();
105 list_for_each_entry_rcu(fan_map, &vxlan->fan.fan_maps, list) {
106 if (fan_map->overlay ==
107 (daddr & inet_make_mask(fan_map->overlay_prefix))) {
108 rcu_read_unlock();
109 return fan_map;
110 }
111 }
112 rcu_read_unlock();
113
114 return NULL;
115}
116
117static void vxlan_fan_flush_map(struct vxlan_dev *vxlan)
118{
119 struct ip_fan_map *fan_map;
120
121 list_for_each_entry_rcu(fan_map, &vxlan->fan.fan_maps, list) {
122 list_del_rcu(&fan_map->list);
123 kfree_rcu(fan_map, rcu);
124 }
125}
126
127static int vxlan_fan_del_map(struct vxlan_dev *vxlan, __be32 overlay)
128{
129 struct ip_fan_map *fan_map;
130
131 fan_map = vxlan_fan_find_map(vxlan, overlay);
132 if (!fan_map)
133 return -ENOENT;
134
135 list_del_rcu(&fan_map->list);
136 kfree_rcu(fan_map, rcu);
137
138 return 0;
139}
140
141static int vxlan_fan_add_map(struct vxlan_dev *vxlan, struct ifla_fan_map *map)
142{
143 __be32 overlay_mask, underlay_mask;
144 struct ip_fan_map *fan_map;
145
146 overlay_mask = inet_make_mask(map->overlay_prefix);
147 underlay_mask = inet_make_mask(map->underlay_prefix);
148
149 netdev_dbg(vxlan->dev, "vfam: map: o %x/%d u %x/%d om %x um %x\n",
150 map->overlay, map->overlay_prefix,
151 map->underlay, map->underlay_prefix,
152 overlay_mask, underlay_mask);
153
154 if ((map->overlay & ~overlay_mask) || (map->underlay & ~underlay_mask))
155 return -EINVAL;
156
157 if (!(map->overlay & overlay_mask) && (map->underlay & underlay_mask))
158 return -EINVAL;
159
160 /* Special case: overlay 0 and underlay 0: flush all mappings */
161 if (!map->overlay && !map->underlay) {
162 vxlan_fan_flush_map(vxlan);
163 return 0;
164 }
165
166 /* Special case: overlay set and underlay 0: clear map for overlay */
167 if (!map->underlay)
168 return vxlan_fan_del_map(vxlan, map->overlay);
169
170 if (vxlan_fan_find_map(vxlan, map->overlay))
171 return -EEXIST;
172
173 fan_map = kmalloc(sizeof(*fan_map), GFP_KERNEL);
174 fan_map->underlay = map->underlay;
175 fan_map->overlay = map->overlay;
176 fan_map->underlay_prefix = map->underlay_prefix;
177 fan_map->overlay_mask = ntohl(overlay_mask);
178 fan_map->overlay_prefix = map->overlay_prefix;
179
180 list_add_tail_rcu(&fan_map->list, &vxlan->fan.fan_maps);
181
182 return 0;
183}
184
185static int vxlan_parse_fan_map(struct nlattr *data[], struct vxlan_dev *vxlan)
186{
187 struct ifla_fan_map *map;
188 struct nlattr *attr;
189 int rem, rv;
190
191 nla_for_each_nested(attr, data[IFLA_IPTUN_FAN_MAP], rem) {
192 map = nla_data(attr);
193 rv = vxlan_fan_add_map(vxlan, map);
194 if (rv)
195 return rv;
196 }
197
198 return 0;
199}
200
201static int vxlan_fan_build_rdst(struct vxlan_dev *vxlan, struct sk_buff *skb,
202 struct vxlan_rdst *fan_rdst)
203{
204 struct ip_fan_map *f_map;
205 union vxlan_addr *va;
206 u32 daddr, underlay;
207 struct arphdr *arp;
208 void *arp_ptr;
209 struct ethhdr *eth;
210 struct iphdr *iph;
211
212 eth = eth_hdr(skb);
213 switch (eth->h_proto) {
214 case htons(ETH_P_IP):
215 iph = ip_hdr(skb);
216 if (!iph)
217 return -EINVAL;
218 daddr = iph->daddr;
219 break;
220 case htons(ETH_P_ARP):
221 arp = arp_hdr(skb);
222 if (!arp)
223 return -EINVAL;
224 arp_ptr = arp + 1;
225 netdev_dbg(vxlan->dev,
226 "vfbr: arp sha %pM sip %pI4 tha %pM tip %pI4\n",
227 arp_ptr, arp_ptr + skb->dev->addr_len,
228 arp_ptr + skb->dev->addr_len + 4,
229 arp_ptr + (skb->dev->addr_len * 2) + 4);
230 arp_ptr += (skb->dev->addr_len * 2) + 4;
231 memcpy(&daddr, arp_ptr, 4);
232 break;
233 default:
234 netdev_dbg(vxlan->dev, "vfbr: unknown eth p %x\n", eth->h_proto);
235 return -EINVAL;
236 }
237
238 f_map = vxlan_fan_find_map(vxlan, daddr);
239 if (!f_map)
240 return -EINVAL;
241
242 daddr = ntohl(daddr);
243 underlay = ntohl(f_map->underlay);
244 if (!underlay)
245 return -EINVAL;
246
247 memset(fan_rdst, 0, sizeof(*fan_rdst));
248 va = &fan_rdst->remote_ip;
249 va->sa.sa_family = AF_INET;
250 fan_rdst->remote_vni = vxlan->default_dst.remote_vni;
251 va->sin.sin_addr.s_addr = htonl(underlay |
252 ((daddr & ~f_map->overlay_mask) >>
253 (32 - f_map->overlay_prefix -
254 (32 - f_map->underlay_prefix))));
255 netdev_dbg(vxlan->dev, "vfbr: daddr %x ul %x dst %x\n",
256 daddr, underlay, va->sin.sin_addr.s_addr);
257
258 return 0;
259}
260
e4c7ed41
CW		 261#if IS_ENABLED(CONFIG_IPV6)
262static inline
263bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
264{
f0ef3126
JB		 265 if (a->sa.sa_family != b->sa.sa_family)
266 return false;
267 if (a->sa.sa_family == AF_INET6)
268 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
269 else
270 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 271}
272
e4c7ed41
CW		 273static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
274{
f0ef3126 275 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 276 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 277 ip->sa.sa_family = AF_INET6;
278 return 0;
279 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 280 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 281 ip->sa.sa_family = AF_INET;
282 return 0;
283 } else {
284 return -EAFNOSUPPORT;
285 }
e4c7ed41
CW		 286}
287
288static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 289 const union vxlan_addr *ip)
e4c7ed41 290{
f0ef3126 291 if (ip->sa.sa_family == AF_INET6)
930345ea 292 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 293 else
930345ea 294 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 295}
296
297#else /* !CONFIG_IPV6 */
298
299static inline
300bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
301{
f0ef3126 302 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 303}
304
e4c7ed41
CW		 305static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
306{
f0ef3126
JB		 307 if (nla_len(nla) >= sizeof(struct in6_addr)) {
308 return -EAFNOSUPPORT;
309 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 310 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 311 ip->sa.sa_family = AF_INET;
312 return 0;
313 } else {
314 return -EAFNOSUPPORT;
315 }
e4c7ed41
CW		 316}
317
318static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 319 const union vxlan_addr *ip)
e4c7ed41 320{
930345ea 321 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 322}
323#endif
324
553675fb 325/* Virtual Network hash table head */
54bfd872 326static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 327{
54bfd872 328 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 329}
330
331/* Socket hash table head */
332static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 333{
334 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
335
553675fb 336 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
337}
338
3e61aa8f
SH		 339/* First remote destination for a forwarding entry.
340 * Guaranteed to be non-NULL because remotes are never deleted.
341 */
5ca5461c 342static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 343{
1274e1cc
RP		 344 if (rcu_access_pointer(fdb->nh))
345 return NULL;
5ca5461c 346 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
347}
348
349static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
350{
1274e1cc
RP		 351 if (rcu_access_pointer(fdb->nh))
352 return NULL;
5ca5461c 353 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 354}
355
78ec710e
FF		 356/* Find VXLAN socket based on network namespace, address family, UDP port,
357 * enabled unshareable flags and socket device binding (see l3mdev with
358 * non-default VRF).
ac5132d1
TG		 359 */
360static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
aab8cc36 361 __be16 port, u32 flags, int ifindex)
553675fb 362{
363 struct vxlan_sock *vs;
af33c1ad
TH		 364
365 flags &= VXLAN_F_RCV_FLAGS;
553675fb 366
367 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 368 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 369 vxlan_get_sk_family(vs) == family &&
aab8cc36
AB		 370 vs->flags == flags &&
371 vs->sock->sk->sk_bound_dev_if == ifindex)
553675fb 372 return vs;
373 }
374 return NULL;
d342894c 375}
376
49f810f0
MS		 377static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
378 __be32 vni)
d342894c 379{
69e76661 380 struct vxlan_dev_node *node;
d342894c 381
b9167b2e
JB		 382 /* For flow based devices, map all packets to VNI 0 */
383 if (vs->flags & VXLAN_F_COLLECT_METADATA)
384 vni = 0;
385
69e76661
JB		 386 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
387 if (node->vxlan->default_dst.remote_vni != vni)
49f810f0
MS		 388 continue;
389
390 if (IS_ENABLED(CONFIG_IPV6)) {
69e76661 391 const struct vxlan_config *cfg = &node->vxlan->cfg;
49f810f0
MS		 392
393 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
394 cfg->remote_ifindex != ifindex)
395 continue;
396 }
397
69e76661 398 return node->vxlan;
d342894c 399 }
400
401 return NULL;
402}
403
5cfccc5a 404/* Look up VNI in a per net namespace table */
49f810f0
MS		 405static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
406 __be32 vni, sa_family_t family,
407 __be16 port, u32 flags)
5cfccc5a
PS		 408{
409 struct vxlan_sock *vs;
410
aab8cc36 411 vs = vxlan_find_sock(net, family, port, flags, ifindex);
5cfccc5a
PS		 412 if (!vs)
413 return NULL;
414
49f810f0 415 return vxlan_vs_find_vni(vs, ifindex, vni);
5cfccc5a
PS		 416}
417
d342894c 418/* Fill in neighbour message in skbuff. */
419static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 420 const struct vxlan_fdb *fdb,
421 u32 portid, u32 seq, int type, unsigned int flags,
422 const struct vxlan_rdst *rdst)
d342894c 423{
424 unsigned long now = jiffies;
425 struct nda_cacheinfo ci;
1274e1cc 426 bool send_ip, send_eth;
d342894c 427 struct nlmsghdr *nlh;
1274e1cc 428 struct nexthop *nh;
d342894c 429 struct ndmsg *ndm;
06ec313e
IS		 430 int nh_family;
431 u32 nh_id;
d342894c 432
433 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
434 if (nlh == NULL)
435 return -EMSGSIZE;
436
437 ndm = nlmsg_data(nlh);
438 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 439
440 send_eth = send_ip = true;
441
06ec313e
IS		 442 rcu_read_lock();
443 nh = rcu_dereference(fdb->nh);
444 if (nh) {
445 nh_family = nexthop_get_family(nh);
446 nh_id = nh->id;
447 }
448 rcu_read_unlock();
449
e4f67add 450 if (type == RTM_GETNEIGH) {
1274e1cc
RP		 451 if (rdst) {
452 send_ip = !vxlan_addr_any(&rdst->remote_ip);
453 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
454 } else if (nh) {
06ec313e 455 ndm->ndm_family = nh_family;
1274e1cc 456 }
e4f67add
DS		 457 send_eth = !is_zero_ether_addr(fdb->eth_addr);
458 } else
459 ndm->ndm_family = AF_BRIDGE;
d342894c 460 ndm->ndm_state = fdb->state;
461 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 462 ndm->ndm_flags = fdb->flags;
1274e1cc 463 if (rdst && rdst->offloaded)
0efe1173 464 ndm->ndm_flags |= NTF_OFFLOADED;
545469f7 465 ndm->ndm_type = RTN_UNICAST;
d342894c 466
193523bf 467 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 468 nla_put_s32(skb, NDA_LINK_NETNSID,
38f507f1 469 peernet2id(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 470 goto nla_put_failure;
471
e4f67add 472 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 473 goto nla_put_failure;
1274e1cc 474 if (nh) {
06ec313e 475 if (nla_put_u32(skb, NDA_NH_ID, nh_id))
1274e1cc
RP		 476 goto nla_put_failure;
477 } else if (rdst) {
478 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST,
479 &rdst->remote_ip))
480 goto nla_put_failure;
481
482 if (rdst->remote_port &&
483 rdst->remote_port != vxlan->cfg.dst_port &&
484 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
485 goto nla_put_failure;
486 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
487 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
488 goto nla_put_failure;
489 if (rdst->remote_ifindex &&
490 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
491 goto nla_put_failure;
492 }
d342894c 493
dc5321d7 494 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
3ad7a4b1
RP		 495 nla_put_u32(skb, NDA_SRC_VNI,
496 be32_to_cpu(fdb->vni)))
497 goto nla_put_failure;
d342894c 498
499 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
500 ci.ndm_confirmed = 0;
501 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
502 ci.ndm_refcnt = 0;
503
504 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
505 goto nla_put_failure;
506
053c095a
JB		 507 nlmsg_end(skb, nlh);
508 return 0;
d342894c 509
510nla_put_failure:
511 nlmsg_cancel(skb, nlh);
512 return -EMSGSIZE;
513}
514
515static inline size_t vxlan_nlmsg_size(void)
516{
517 return NLMSG_ALIGN(sizeof(struct ndmsg))
518 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 519 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 520 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 521 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
522 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 523 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 524 + nla_total_size(sizeof(struct nda_cacheinfo));
525}
526
9a997353
PM		 527static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
528 struct vxlan_rdst *rd, int type)
d342894c 529{
530 struct net *net = dev_net(vxlan->dev);
531 struct sk_buff *skb;
532 int err = -ENOBUFS;
533
534 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
535 if (skb == NULL)
536 goto errout;
537
9e4b93f9 538 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 539 if (err < 0) {
540 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
541 WARN_ON(err == -EMSGSIZE);
542 kfree_skb(skb);
543 goto errout;
544 }
545
546 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
547 return;
548errout:
549 if (err < 0)
550 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
551}
552
ff23b91c
PM		 553static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
554 const struct vxlan_fdb *fdb,
555 const struct vxlan_rdst *rd,
4c59b7d1 556 struct netlink_ext_ack *extack,
ff23b91c
PM		 557 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
558{
559 fdb_info->info.dev = vxlan->dev;
4c59b7d1 560 fdb_info->info.extack = extack;
ff23b91c
PM		 561 fdb_info->remote_ip = rd->remote_ip;
562 fdb_info->remote_port = rd->remote_port;
563 fdb_info->remote_vni = rd->remote_vni;
564 fdb_info->remote_ifindex = rd->remote_ifindex;
565 memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
566 fdb_info->vni = fdb->vni;
567 fdb_info->offloaded = rd->offloaded;
568 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
569}
570
61f46fe8
PM		 571static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
572 struct vxlan_fdb *fdb,
573 struct vxlan_rdst *rd,
4c59b7d1
PM		 574 bool adding,
575 struct netlink_ext_ack *extack)
9a997353
PM		 576{
577 struct switchdev_notifier_vxlan_fdb_info info;
578 enum switchdev_notifier_type notifier_type;
61f46fe8 579 int ret;
9a997353
PM		 580
581 if (WARN_ON(!rd))
61f46fe8 582 return 0;
9a997353
PM		 583
584 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
585 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
4c59b7d1 586 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
61f46fe8 587 ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
6685987c 588 &info.info, extack);
61f46fe8 589 return notifier_to_errno(ret);
9a997353
PM		 590}
591
61f46fe8 592static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
4c59b7d1
PM		 593 struct vxlan_rdst *rd, int type, bool swdev_notify,
594 struct netlink_ext_ack *extack)
9a997353 595{
61f46fe8
PM		 596 int err;
597
1274e1cc 598 if (swdev_notify && rd) {
0e6160f3
PM		 599 switch (type) {
600 case RTM_NEWNEIGH:
61f46fe8 601 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
4c59b7d1 602 true, extack);
61f46fe8
PM		 603 if (err)
604 return err;
0e6160f3
PM		 605 break;
606 case RTM_DELNEIGH:
607 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
4c59b7d1 608 false, extack);
0e6160f3
PM		 609 break;
610 }
9a997353
PM		 611 }
612
613 __vxlan_fdb_notify(vxlan, fdb, rd, type);
61f46fe8 614 return 0;
9a997353
PM		 615}
616
e4c7ed41 617static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 618{
619 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 620 struct vxlan_fdb f = {
621 .state = NUD_STALE,
622 };
623 struct vxlan_rdst remote = {
e4c7ed41 624 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 625 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 626 };
3e61aa8f 627
4c59b7d1 628 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
e4f67add
DS		 629}
630
631static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
632{
bb3fd687
SH		 633 struct vxlan_fdb f = {
634 .state = NUD_STALE,
635 };
9e4b93f9 636 struct vxlan_rdst remote = { };
e4f67add 637
e4f67add
DS		 638 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
639
4c59b7d1 640 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
e4f67add
DS		 641}
642
d342894c 643/* Hash Ethernet address */
644static u32 eth_hash(const unsigned char *addr)
645{
646 u64 value = get_unaligned((u64 *)addr);
647
648 /* only want 6 bytes */
649#ifdef __BIG_ENDIAN
d342894c 650 value >>= 16;
321fb991 651#else
652 value <<= 16;
d342894c 653#endif
654 return hash_64(value, FDB_HASH_BITS);
655}
656
3ad7a4b1
RP		 657static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
658{
659 /* use 1 byte of OUI and 3 bytes of NIC */
660 u32 key = get_unaligned((u32 *)(addr + 2));
661
662 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
663}
664
fe1e0713
L		 665static u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
666{
667 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
668 return eth_vni_hash(mac, vni);
669 else
670 return eth_hash(mac);
671}
672
d342894c 673/* Hash chain to use given mac address */
674static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
3ad7a4b1 675 const u8 *mac, __be32 vni)
d342894c 676{
fe1e0713 677 return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
d342894c 678}
679
680/* Look up Ethernet address in forwarding table */
014be2c8 681static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 682 const u8 *mac, __be32 vni)
d342894c 683{
3ad7a4b1 684 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
d342894c 685 struct vxlan_fdb *f;
d342894c 686
b67bfe0d 687 hlist_for_each_entry_rcu(f, head, hlist) {
3ad7a4b1 688 if (ether_addr_equal(mac, f->eth_addr)) {
dc5321d7 689 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 690 if (vni == f->vni)
691 return f;
692 } else {
693 return f;
694 }
695 }
d342894c 696 }
697
698 return NULL;
699}
700
014be2c8 701static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 702 const u8 *mac, __be32 vni)
014be2c8
SS		 703{
704 struct vxlan_fdb *f;
705
3ad7a4b1 706 f = __vxlan_find_mac(vxlan, mac, vni);
016f3d18 707 if (f && f->used != jiffies)
014be2c8
SS		 708 f->used = jiffies;
709
710 return f;
711}
712
a5e7c10a
MR		 713/* caller should hold vxlan->hash_lock */
714static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 715 union vxlan_addr *ip, __be16 port,
54bfd872 716 __be32 vni, __u32 ifindex)
6681712d 717{
3e61aa8f 718 struct vxlan_rdst *rd;
6681712d 719
3e61aa8f 720 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 721 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 722 rd->remote_port == port &&
723 rd->remote_vni == vni &&
724 rd->remote_ifindex == ifindex)
a5e7c10a 725 return rd;
6681712d 726 }
3e61aa8f 727
a5e7c10a
MR		 728 return NULL;
729}
730
1941f1d6
PM		 731int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
732 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
733{
734 struct vxlan_dev *vxlan = netdev_priv(dev);
735 u8 eth_addr[ETH_ALEN + 2] = { 0 };
736 struct vxlan_rdst *rdst;
737 struct vxlan_fdb *f;
738 int rc = 0;
739
740 if (is_multicast_ether_addr(mac) ||
741 is_zero_ether_addr(mac))
742 return -EINVAL;
743
744 ether_addr_copy(eth_addr, mac);
745
746 rcu_read_lock();
747
748 f = __vxlan_find_mac(vxlan, eth_addr, vni);
749 if (!f) {
750 rc = -ENOENT;
751 goto out;
752 }
753
754 rdst = first_remote_rcu(f);
4c59b7d1 755 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
1941f1d6
PM		 756
757out:
758 rcu_read_unlock();
759 return rc;
760}
761EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
762
4f89f5b5
PM		 763static int vxlan_fdb_notify_one(struct notifier_block *nb,
764 const struct vxlan_dev *vxlan,
765 const struct vxlan_fdb *f,
4c59b7d1
PM		 766 const struct vxlan_rdst *rdst,
767 struct netlink_ext_ack *extack)
4f89f5b5
PM		 768{
769 struct switchdev_notifier_vxlan_fdb_info fdb_info;
770 int rc;
771
4c59b7d1 772 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
4f89f5b5
PM		 773 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
774 &fdb_info);
775 return notifier_to_errno(rc);
776}
777
778int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
4c59b7d1
PM		 779 struct notifier_block *nb,
780 struct netlink_ext_ack *extack)
4f89f5b5
PM		 781{
782 struct vxlan_dev *vxlan;
783 struct vxlan_rdst *rdst;
784 struct vxlan_fdb *f;
785 unsigned int h;
786 int rc = 0;
787
788 if (!netif_is_vxlan(dev))
789 return -EINVAL;
790 vxlan = netdev_priv(dev);
791
4f89f5b5 792 for (h = 0; h < FDB_HASH_SIZE; ++h) {
fe1e0713 793 spin_lock_bh(&vxlan->hash_lock[h]);
4f89f5b5
PM		 794 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
795 if (f->vni == vni) {
796 list_for_each_entry(rdst, &f->remotes, list) {
797 rc = vxlan_fdb_notify_one(nb, vxlan,
4c59b7d1
PM		 798 f, rdst,
799 extack);
4f89f5b5 800 if (rc)
fe1e0713 801 goto unlock;
4f89f5b5
PM		 802 }
803 }
804 }
fe1e0713 805 spin_unlock_bh(&vxlan->hash_lock[h]);
4f89f5b5 806 }
fe1e0713 807 return 0;
4f89f5b5 808
fe1e0713
L		 809unlock:
810 spin_unlock_bh(&vxlan->hash_lock[h]);
4f89f5b5
PM		 811 return rc;
812}
813EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
814
e5ff4b19
PM		 815void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
816{
817 struct vxlan_dev *vxlan;
818 struct vxlan_rdst *rdst;
819 struct vxlan_fdb *f;
820 unsigned int h;
821
822 if (!netif_is_vxlan(dev))
823 return;
824 vxlan = netdev_priv(dev);
825
e5ff4b19 826 for (h = 0; h < FDB_HASH_SIZE; ++h) {
fe1e0713 827 spin_lock_bh(&vxlan->hash_lock[h]);
e5ff4b19
PM		 828 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
829 if (f->vni == vni)
830 list_for_each_entry(rdst, &f->remotes, list)
831 rdst->offloaded = false;
fe1e0713 832 spin_unlock_bh(&vxlan->hash_lock[h]);
e5ff4b19 833 }
fe1e0713 834
e5ff4b19
PM		 835}
836EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
837
906dc186
TR		 838/* Replace destination of unicast mac */
839static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872 840 union vxlan_addr *ip, __be16 port, __be32 vni,
ccdfd4f7 841 __u32 ifindex, struct vxlan_rdst *oldrd)
906dc186
TR		 842{
843 struct vxlan_rdst *rd;
844
845 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
846 if (rd)
847 return 0;
848
849 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
850 if (!rd)
851 return 0;
0c1d70af 852
ccdfd4f7 853 *oldrd = *rd;
0c1d70af 854 dst_cache_reset(&rd->dst_cache);
e4c7ed41 855 rd->remote_ip = *ip;
906dc186
TR		 856 rd->remote_port = port;
857 rd->remote_vni = vni;
858 rd->remote_ifindex = ifindex;
6ad0b5a4 859 rd->offloaded = false;
906dc186
TR		 860 return 1;
861}
862
a5e7c10a
MR		 863/* Add/update destinations for multicast */
864static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 865 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 866 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 867{
868 struct vxlan_rdst *rd;
869
870 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
871 if (rd)
872 return 0;
873
6681712d
DS		 874 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
875 if (rd == NULL)
876 return -ENOBUFS;
0c1d70af
PA		 877
878 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
879 kfree(rd);
880 return -ENOBUFS;
881 }
882
e4c7ed41 883 rd->remote_ip = *ip;
6681712d 884 rd->remote_port = port;
0efe1173 885 rd->offloaded = false;
6681712d
DS		 886 rd->remote_vni = vni;
887 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 888
889 list_add_tail_rcu(&rd->list, &f->remotes);
890
9e4b93f9 891 *rdp = rd;
6681712d
DS		 892 return 1;
893}
894
dfd8645e
TH		 895static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
896 unsigned int off,
897 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 898 __be32 vni_field,
899 struct gro_remcsum *grc,
0ace2ca8 900 bool nopartial)
dfd8645e 901{
b7fe10e5 902 size_t start, offset;
dfd8645e
TH		 903
904 if (skb->remcsum_offload)
b7fe10e5 905 return vh;
dfd8645e
TH		 906
907 if (!NAPI_GRO_CB(skb)->csum_valid)
908 return NULL;
909
54bfd872
JB		 910 start = vxlan_rco_start(vni_field);
911 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 912
b7fe10e5
TH		 913 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
914 start, offset, grc, nopartial);
dfd8645e
TH		 915
916 skb->remcsum_offload = 1;
917
918 return vh;
919}
920
d4546c25
DM		 921static struct sk_buff *vxlan_gro_receive(struct sock *sk,
922 struct list_head *head,
923 struct sk_buff *skb)
dc01e7d3 924{
d4546c25
DM		 925 struct sk_buff *pp = NULL;
926 struct sk_buff *p;
dc01e7d3 927 struct vxlanhdr *vh, *vh2;
9b174d88 928 unsigned int hlen, off_vx;
dc01e7d3 929 int flush = 1;
5602c48c 930 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
54bfd872 931 __be32 flags;
26c4f7da
TH		 932 struct gro_remcsum grc;
933
934 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 935
936 off_vx = skb_gro_offset(skb);
937 hlen = off_vx + sizeof(*vh);
938 vh = skb_gro_header_fast(skb, off_vx);
939 if (skb_gro_header_hard(skb, hlen)) {
940 vh = skb_gro_header_slow(skb, hlen, off_vx);
941 if (unlikely(!vh))
942 goto out;
943 }
dc01e7d3 944
dfd8645e
TH		 945 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
946
54bfd872 947 flags = vh->vx_flags;
dfd8645e
TH		 948
949 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
950 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 951 vh->vx_vni, &grc,
0ace2ca8
TH		 952 !!(vs->flags &
953 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 954
955 if (!vh)
956 goto out;
957 }
958
b7fe10e5
TH		 959 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
960
d4546c25 961 list_for_each_entry(p, head, list) {
dc01e7d3
OG		 962 if (!NAPI_GRO_CB(p)->same_flow)
963 continue;
964
965 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 966 if (vh->vx_flags != vh2->vx_flags ||
967 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 968 NAPI_GRO_CB(p)->same_flow = 0;
969 continue;
970 }
dc01e7d3
OG		 971 }
972
fcd91dd4 973 pp = call_gro_receive(eth_gro_receive, head, skb);
c194cf93 974 flush = 0;
dc01e7d3 975
dc01e7d3 976out:
603d4cf8 977 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
dc01e7d3
OG		 978
979 return pp;
980}
981
5602c48c 982static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
dc01e7d3 983{
229740c6
JR		 984 /* Sets 'skb->inner_mac_header' since we are always called with
985 * 'skb->encapsulation' set.
986 */
9b174d88 987 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 988}
989
c7cdbe2e
RP		 990static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
991 __u16 state, __be32 src_vni,
992 __u16 ndm_flags)
25e20e73
RP		 993{
994 struct vxlan_fdb *f;
995
996 f = kmalloc(sizeof(*f), GFP_ATOMIC);
997 if (!f)
998 return NULL;
999 f->state = state;
1000 f->flags = ndm_flags;
1001 f->updated = f->used = jiffies;
1002 f->vni = src_vni;
1274e1cc 1003 f->nh = NULL;
79472fe8 1004 RCU_INIT_POINTER(f->vdev, vxlan);
1274e1cc 1005 INIT_LIST_HEAD(&f->nh_list);
25e20e73
RP		 1006 INIT_LIST_HEAD(&f->remotes);
1007 memcpy(f->eth_addr, mac, ETH_ALEN);
1008
1009 return f;
1010}
1011
7c31e54a
TY		 1012static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
1013 __be32 src_vni, struct vxlan_fdb *f)
1014{
1015 ++vxlan->addrcnt;
1016 hlist_add_head_rcu(&f->hlist,
1017 vxlan_fdb_head(vxlan, mac, src_vni));
1018}
1019
1274e1cc
RP		 1020static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
1021 u32 nhid, struct netlink_ext_ack *extack)
1022{
1023 struct nexthop *old_nh = rtnl_dereference(fdb->nh);
1274e1cc
RP		 1024 struct nexthop *nh;
1025 int err = -EINVAL;
1026
1027 if (old_nh && old_nh->id == nhid)
1028 return 0;
1029
1030 nh = nexthop_find_by_id(vxlan->net, nhid);
1031 if (!nh) {
1032 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
1033 goto err_inval;
1034 }
1035
1036 if (nh) {
1037 if (!nexthop_get(nh)) {
1038 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
1039 nh = NULL;
1040 goto err_inval;
1041 }
ce9ac056 1042 if (!nexthop_is_fdb(nh)) {
1274e1cc
RP		 1043 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
1044 goto err_inval;
1045 }
1046
50cb8769 1047 if (!nexthop_is_multipath(nh)) {
1274e1cc
RP		 1048 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
1049 goto err_inval;
1050 }
1051
1052 /* check nexthop group family */
1274e1cc
RP		 1053 switch (vxlan->default_dst.remote_ip.sa.sa_family) {
1054 case AF_INET:
50cb8769 1055 if (!nexthop_has_v4(nh)) {
1274e1cc
RP		 1056 err = -EAFNOSUPPORT;
1057 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
1058 goto err_inval;
1059 }
1060 break;
1061 case AF_INET6:
50cb8769 1062 if (nexthop_has_v4(nh)) {
1274e1cc
RP		 1063 err = -EAFNOSUPPORT;
1064 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
1065 goto err_inval;
1066 }
1067 }
1068 }
1069
1070 if (old_nh) {
1071 list_del_rcu(&fdb->nh_list);
1072 nexthop_put(old_nh);
1073 }
1074 rcu_assign_pointer(fdb->nh, nh);
1075 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
1076 return 1;
1077
1078err_inval:
1079 if (nh)
1080 nexthop_put(nh);
1081 return err;
1082}
1083
d342894c 1084static int vxlan_fdb_create(struct vxlan_dev *vxlan,
25e20e73
RP		 1085 const u8 *mac, union vxlan_addr *ip,
1086 __u16 state, __be16 port, __be32 src_vni,
45598c1c 1087 __be32 vni, __u32 ifindex, __u16 ndm_flags,
1274e1cc
RP		 1088 u32 nhid, struct vxlan_fdb **fdb,
1089 struct netlink_ext_ack *extack)
25e20e73
RP		 1090{
1091 struct vxlan_rdst *rd = NULL;
1092 struct vxlan_fdb *f;
1093 int rc;
1094
1095 if (vxlan->cfg.addrmax &&
1096 vxlan->addrcnt >= vxlan->cfg.addrmax)
1097 return -ENOSPC;
1098
1099 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
c7cdbe2e 1100 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
25e20e73
RP		 1101 if (!f)
1102 return -ENOMEM;
1103
1274e1cc
RP		 1104 if (nhid)
1105 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1106 else
1107 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1108 if (rc < 0)
1109 goto errout;
25e20e73 1110
25e20e73
RP		 1111 *fdb = f;
1112
1113 return 0;
1274e1cc
RP		 1114
1115errout:
1116 kfree(f);
1117 return rc;
25e20e73
RP		 1118}
1119
7c31e54a 1120static void __vxlan_fdb_free(struct vxlan_fdb *f)
c2b200e0 1121{
c2b200e0 1122 struct vxlan_rdst *rd, *nd;
1274e1cc
RP		 1123 struct nexthop *nh;
1124
1125 nh = rcu_dereference_raw(f->nh);
1126 if (nh) {
1127 rcu_assign_pointer(f->nh, NULL);
79472fe8 1128 rcu_assign_pointer(f->vdev, NULL);
1274e1cc
RP		 1129 nexthop_put(nh);
1130 }
c2b200e0
PM		 1131
1132 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
1133 dst_cache_destroy(&rd->dst_cache);
1134 kfree(rd);
1135 }
1136 kfree(f);
1137}
1138
7c31e54a
TY		 1139static void vxlan_fdb_free(struct rcu_head *head)
1140{
1141 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
1142
1143 __vxlan_fdb_free(f);
1144}
1145
c2b200e0
PM		 1146static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1147 bool do_notify, bool swdev_notify)
1148{
1149 struct vxlan_rdst *rd;
1150
1151 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
1152
1153 --vxlan->addrcnt;
1274e1cc
RP		 1154 if (do_notify) {
1155 if (rcu_access_pointer(f->nh))
1156 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
4c59b7d1 1157 swdev_notify, NULL);
1274e1cc
RP		 1158 else
1159 list_for_each_entry(rd, &f->remotes, list)
1160 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
1161 swdev_notify, NULL);
1162 }
c2b200e0
PM		 1163
1164 hlist_del_rcu(&f->hlist);
79472fe8 1165 list_del_rcu(&f->nh_list);
c2b200e0
PM		 1166 call_rcu(&f->rcu, vxlan_fdb_free);
1167}
1168
fc4aa1ca
PM		 1169static void vxlan_dst_free(struct rcu_head *head)
1170{
1171 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
1172
1173 dst_cache_destroy(&rd->dst_cache);
1174 kfree(rd);
1175}
1176
a76d1ca2
PM		 1177static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
1178 union vxlan_addr *ip,
1179 __u16 state, __u16 flags,
1180 __be16 port, __be32 vni,
1181 __u32 ifindex, __u16 ndm_flags,
1274e1cc 1182 struct vxlan_fdb *f, u32 nhid,
4c59b7d1
PM		 1183 bool swdev_notify,
1184 struct netlink_ext_ack *extack)
d342894c 1185{
45598c1c 1186 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
9e4b93f9 1187 struct vxlan_rdst *rd = NULL;
ccdfd4f7 1188 struct vxlan_rdst oldrd;
d342894c 1189 int notify = 0;
61f46fe8
PM		 1190 int rc = 0;
1191 int err;
d342894c 1192
1274e1cc
RP		 1193 if (nhid && !rcu_access_pointer(f->nh)) {
1194 NL_SET_ERR_MSG(extack,
1195 "Cannot replace an existing non nexthop fdb with a nexthop");
1196 return -EOPNOTSUPP;
1197 }
1198
1199 if (nhid && (flags & NLM_F_APPEND)) {
1200 NL_SET_ERR_MSG(extack,
1201 "Cannot append to a nexthop fdb");
1202 return -EOPNOTSUPP;
1203 }
1204
a76d1ca2
PM		 1205 /* Do not allow an externally learned entry to take over an entry added
1206 * by the user.
1207 */
1208 if (!(fdb_flags & NTF_EXT_LEARNED) ||
1209 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1210 if (f->state != state) {
1211 f->state = state;
1212 f->updated = jiffies;
1213 notify = 1;
ae884082 1214 }
a76d1ca2
PM		 1215 if (f->flags != fdb_flags) {
1216 f->flags = fdb_flags;
1217 f->updated = jiffies;
1218 notify = 1;
906dc186 1219 }
a76d1ca2 1220 }
6681712d 1221
a76d1ca2
PM		 1222 if ((flags & NLM_F_REPLACE)) {
1223 /* Only change unicasts */
1224 if (!(is_multicast_ether_addr(f->eth_addr) ||
1225 is_zero_ether_addr(f->eth_addr))) {
1274e1cc
RP		 1226 if (nhid) {
1227 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1228 if (rc < 0)
1229 return rc;
1230 } else {
1231 rc = vxlan_fdb_replace(f, ip, port, vni,
1232 ifindex, &oldrd);
1233 }
6681712d 1234 notify |= rc;
a76d1ca2 1235 } else {
1274e1cc 1236 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
906dc186 1237 return -EOPNOTSUPP;
a76d1ca2
PM		 1238 }
1239 }
1240 if ((flags & NLM_F_APPEND) &&
1241 (is_multicast_ether_addr(f->eth_addr) ||
1242 is_zero_ether_addr(f->eth_addr))) {
1243 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
906dc186 1244
25e20e73 1245 if (rc < 0)
17b46365 1246 return rc;
a76d1ca2 1247 notify |= rc;
d342894c 1248 }
1249
a76d1ca2
PM		 1250 if (ndm_flags & NTF_USE)
1251 f->used = jiffies;
1252
9e4b93f9
ND		 1253 if (notify) {
1254 if (rd == NULL)
1255 rd = first_remote_rtnl(f);
a76d1ca2 1256
61f46fe8 1257 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
4c59b7d1 1258 swdev_notify, extack);
61f46fe8
PM		 1259 if (err)
1260 goto err_notify;
9e4b93f9 1261 }
d342894c 1262
1263 return 0;
61f46fe8
PM		 1264
1265err_notify:
1274e1cc
RP		 1266 if (nhid)
1267 return err;
61f46fe8
PM		 1268 if ((flags & NLM_F_REPLACE) && rc)
1269 *rd = oldrd;
fc4aa1ca 1270 else if ((flags & NLM_F_APPEND) && rc) {
61f46fe8 1271 list_del_rcu(&rd->list);
fc4aa1ca
PM		 1272 call_rcu(&rd->rcu, vxlan_dst_free);
1273 }
61f46fe8 1274 return err;
d342894c 1275}
1276
a76d1ca2
PM		 1277static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
1278 const u8 *mac, union vxlan_addr *ip,
1279 __u16 state, __u16 flags,
1280 __be16 port, __be32 src_vni, __be32 vni,
1274e1cc 1281 __u32 ifindex, __u16 ndm_flags, u32 nhid,
4c59b7d1
PM		 1282 bool swdev_notify,
1283 struct netlink_ext_ack *extack)
a76d1ca2
PM		 1284{
1285 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1286 struct vxlan_fdb *f;
1287 int rc;
1288
1289 /* Disallow replace to add a multicast entry */
1290 if ((flags & NLM_F_REPLACE) &&
1291 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1292 return -EOPNOTSUPP;
1293
1294 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1295 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1274e1cc 1296 vni, ifindex, fdb_flags, nhid, &f, extack);
a76d1ca2
PM		 1297 if (rc < 0)
1298 return rc;
1299
7c31e54a 1300 vxlan_fdb_insert(vxlan, mac, src_vni, f);
61f46fe8 1301 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
4c59b7d1 1302 swdev_notify, extack);
61f46fe8
PM		 1303 if (rc)
1304 goto err_notify;
1305
a76d1ca2 1306 return 0;
61f46fe8
PM		 1307
1308err_notify:
1309 vxlan_fdb_destroy(vxlan, f, false, false);
1310 return rc;
a76d1ca2
PM		 1311}
1312
1313/* Add new entry to forwarding table -- assumes lock held */
1314static int vxlan_fdb_update(struct vxlan_dev *vxlan,
1315 const u8 *mac, union vxlan_addr *ip,
1316 __u16 state, __u16 flags,
1317 __be16 port, __be32 src_vni, __be32 vni,
1274e1cc 1318 __u32 ifindex, __u16 ndm_flags, u32 nhid,
4c59b7d1
PM		 1319 bool swdev_notify,
1320 struct netlink_ext_ack *extack)
a76d1ca2
PM		 1321{
1322 struct vxlan_fdb *f;
1323
1324 f = __vxlan_find_mac(vxlan, mac, src_vni);
1325 if (f) {
1326 if (flags & NLM_F_EXCL) {
1327 netdev_dbg(vxlan->dev,
1328 "lost race to create %pM\n", mac);
1329 return -EEXIST;
1330 }
1331
1332 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1333 vni, ifindex, ndm_flags, f,
1274e1cc 1334 nhid, swdev_notify, extack);
a76d1ca2
PM		 1335 } else {
1336 if (!(flags & NLM_F_CREATE))
1337 return -ENOENT;
1338
1339 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1340 port, src_vni, vni, ifindex,
1274e1cc
RP		 1341 ndm_flags, nhid, swdev_notify,
1342 extack);
a76d1ca2
PM		 1343 }
1344}
1345
35cf2845 1346static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
0e6160f3 1347 struct vxlan_rdst *rd, bool swdev_notify)
35cf2845
LR		 1348{
1349 list_del_rcu(&rd->list);
4c59b7d1 1350 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
35cf2845
LR		 1351 call_rcu(&rd->rcu, vxlan_dst_free);
1352}
1353
f0b074be 1354static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
3ad7a4b1 1355 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1274e1cc 1356 __be32 *vni, u32 *ifindex, u32 *nhid)
d342894c 1357{
6681712d 1358 struct net *net = dev_net(vxlan->dev);
e4c7ed41 1359 int err;
d342894c 1360
72b48682
RP		 1361 if (tb[NDA_NH_ID] && (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] ||
1362 tb[NDA_PORT]))
1363 return -EINVAL;
1364
f0b074be 1365 if (tb[NDA_DST]) {
e4c7ed41
CW		 1366 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1367 if (err)
1368 return err;
f0b074be 1369 } else {
e4c7ed41 1370 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
98c81476 1371
e4c7ed41
CW		 1372 if (remote->sa.sa_family == AF_INET) {
1373 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1374 ip->sa.sa_family = AF_INET;
1375#if IS_ENABLED(CONFIG_IPV6)
1376 } else {
1377 ip->sin6.sin6_addr = in6addr_any;
1378 ip->sa.sa_family = AF_INET6;
1379#endif
1380 }
f0b074be 1381 }
d342894c 1382
6681712d 1383 if (tb[NDA_PORT]) {
73cf3317 1384 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 1385 return -EINVAL;
f0b074be
MR		 1386 *port = nla_get_be16(tb[NDA_PORT]);
1387 } else {
0dfbdf41 1388 *port = vxlan->cfg.dst_port;
f0b074be 1389 }
6681712d
DS		 1390
1391 if (tb[NDA_VNI]) {
1392 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
1393 return -EINVAL;
54bfd872 1394 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 1395 } else {
1396 *vni = vxlan->default_dst.remote_vni;
1397 }
6681712d 1398
3ad7a4b1
RP		 1399 if (tb[NDA_SRC_VNI]) {
1400 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
1401 return -EINVAL;
1402 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1403 } else {
1404 *src_vni = vxlan->default_dst.remote_vni;
1405 }
1406
6681712d 1407 if (tb[NDA_IFINDEX]) {
5abb0029 1408 struct net_device *tdev;
6681712d
DS		 1409
1410 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
1411 return -EINVAL;
f0b074be 1412 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 1413 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 1414 if (!tdev)
6681712d 1415 return -EADDRNOTAVAIL;
f0b074be
MR		 1416 } else {
1417 *ifindex = 0;
1418 }
1419
1274e1cc
RP		 1420 if (tb[NDA_NH_ID])
1421 *nhid = nla_get_u32(tb[NDA_NH_ID]);
1422 else
1423 *nhid = 0;
1424
f0b074be
MR		 1425 return 0;
1426}
1427
1428/* Add static entry (via netlink) */
1429static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1430 struct net_device *dev,
87b0984e
PM		 1431 const unsigned char *addr, u16 vid, u16 flags,
1432 struct netlink_ext_ack *extack)
f0b074be
MR		 1433{
1434 struct vxlan_dev *vxlan = netdev_priv(dev);
1435 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 1436 union vxlan_addr ip;
f0b074be 1437 __be16 port;
3ad7a4b1 1438 __be32 src_vni, vni;
1274e1cc 1439 u32 ifindex, nhid;
fe1e0713 1440 u32 hash_index;
f0b074be
MR		 1441 int err;
1442
1443 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1444 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1445 ndm->ndm_state);
1446 return -EINVAL;
1447 }
1448
1274e1cc 1449 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
f0b074be
MR		 1450 return -EINVAL;
1451
1274e1cc
RP		 1452 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1453 &nhid);
f0b074be
MR		 1454 if (err)
1455 return err;
6681712d 1456
5933a7bb
MR		 1457 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1458 return -EAFNOSUPPORT;
1459
fe1e0713
L		 1460 hash_index = fdb_head_index(vxlan, addr, src_vni);
1461 spin_lock_bh(&vxlan->hash_lock[hash_index]);
25e20e73 1462 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
45598c1c
PM		 1463 port, src_vni, vni, ifindex,
1464 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1274e1cc 1465 nhid, true, extack);
fe1e0713 1466 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
d342894c 1467
1468 return err;
1469}
1470
3ad7a4b1
RP		 1471static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1472 const unsigned char *addr, union vxlan_addr ip,
fc39c38b 1473 __be16 port, __be32 src_vni, __be32 vni,
0e6160f3 1474 u32 ifindex, bool swdev_notify)
d342894c 1475{
bc7892ba 1476 struct vxlan_rdst *rd = NULL;
1274e1cc 1477 struct vxlan_fdb *f;
3ad7a4b1 1478 int err = -ENOENT;
bc7892ba 1479
3ad7a4b1 1480 f = vxlan_find_mac(vxlan, addr, src_vni);
bc7892ba 1481 if (!f)
3ad7a4b1 1482 return err;
bc7892ba 1483
e4c7ed41
CW		 1484 if (!vxlan_addr_any(&ip)) {
1485 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 1486 if (!rd)
1487 goto out;
1488 }
1489
bc7892ba
MR		 1490 /* remove a destination if it's not the only one on the list,
1491 * otherwise destroy the fdb entry
1492 */
1493 if (rd && !list_is_singular(&f->remotes)) {
0e6160f3 1494 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
bc7892ba 1495 goto out;
d342894c 1496 }
bc7892ba 1497
0e6160f3 1498 vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
bc7892ba
MR		 1499
1500out:
3ad7a4b1
RP		 1501 return 0;
1502}
1503
1504/* Delete entry (via netlink) */
1505static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1506 struct net_device *dev,
1507 const unsigned char *addr, u16 vid)
1508{
1509 struct vxlan_dev *vxlan = netdev_priv(dev);
1510 union vxlan_addr ip;
1511 __be32 src_vni, vni;
1274e1cc 1512 u32 ifindex, nhid;
fe1e0713 1513 u32 hash_index;
1274e1cc 1514 __be16 port;
3ad7a4b1
RP		 1515 int err;
1516
1274e1cc
RP		 1517 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1518 &nhid);
3ad7a4b1
RP		 1519 if (err)
1520 return err;
1521
fe1e0713
L		 1522 hash_index = fdb_head_index(vxlan, addr, src_vni);
1523 spin_lock_bh(&vxlan->hash_lock[hash_index]);
0e6160f3
PM		 1524 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1525 true);
fe1e0713 1526 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
d342894c 1527
1528 return err;
1529}
1530
1531/* Dump forwarding table */
1532static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1533 struct net_device *dev,
d297653d 1534 struct net_device *filter_dev, int *idx)
d342894c 1535{
1536 struct vxlan_dev *vxlan = netdev_priv(dev);
1537 unsigned int h;
d297653d 1538 int err = 0;
d342894c 1539
1540 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1541 struct vxlan_fdb *f;
d342894c 1542
b5141915 1543 rcu_read_lock();
b67bfe0d 1544 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 1545 struct vxlan_rdst *rd;
6681712d 1546
1274e1cc 1547 if (rcu_access_pointer(f->nh)) {
b18e9834
RP		 1548 if (*idx < cb->args[2])
1549 goto skip_nh;
1274e1cc
RP		 1550 err = vxlan_fdb_info(skb, vxlan, f,
1551 NETLINK_CB(cb->skb).portid,
1552 cb->nlh->nlmsg_seq,
1553 RTM_NEWNEIGH,
1554 NLM_F_MULTI, NULL);
b5141915
IS		 1555 if (err < 0) {
1556 rcu_read_unlock();
1274e1cc 1557 goto out;
b5141915 1558 }
b18e9834
RP		 1559skip_nh:
1560 *idx += 1;
1274e1cc
RP		 1561 continue;
1562 }
1563
3e61aa8f 1564 list_for_each_entry_rcu(rd, &f->remotes, list) {
d297653d 1565 if (*idx < cb->args[2])
07a51cd3
AW		 1566 goto skip;
1567
6681712d
DS		 1568 err = vxlan_fdb_info(skb, vxlan, f,
1569 NETLINK_CB(cb->skb).portid,
1570 cb->nlh->nlmsg_seq,
1571 RTM_NEWNEIGH,
1572 NLM_F_MULTI, rd);
b5141915
IS		 1573 if (err < 0) {
1574 rcu_read_unlock();
3e61aa8f 1575 goto out;
b5141915 1576 }
3e61aa8f 1577skip:
d297653d 1578 *idx += 1;
07a51cd3 1579 }
d342894c 1580 }
b5141915 1581 rcu_read_unlock();
d342894c 1582 }
3e61aa8f 1583out:
d297653d 1584 return err;
d342894c 1585}
1586
474c3c89
RP		 1587static int vxlan_fdb_get(struct sk_buff *skb,
1588 struct nlattr *tb[],
1589 struct net_device *dev,
1590 const unsigned char *addr,
1591 u16 vid, u32 portid, u32 seq,
1592 struct netlink_ext_ack *extack)
1593{
1594 struct vxlan_dev *vxlan = netdev_priv(dev);
1595 struct vxlan_fdb *f;
1596 __be32 vni;
1597 int err;
1598
1599 if (tb[NDA_VNI])
1600 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1601 else
1602 vni = vxlan->default_dst.remote_vni;
1603
1604 rcu_read_lock();
1605
1606 f = __vxlan_find_mac(vxlan, addr, vni);
1607 if (!f) {
1608 NL_SET_ERR_MSG(extack, "Fdb entry not found");
1609 err = -ENOENT;
1610 goto errout;
1611 }
1612
1613 err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1614 RTM_NEWNEIGH, 0, first_remote_rcu(f));
1615errout:
1616 rcu_read_unlock();
1617 return err;
1618}
1619
d342894c 1620/* Watch incoming packets to learn mapping between Ethernet address
1621 * and Tunnel endpoint.
c4b49512 1622 * Return true if packet is bogus and should be dropped.
d342894c 1623 */
26a41ae6 1624static bool vxlan_snoop(struct net_device *dev,
3ad7a4b1 1625 union vxlan_addr *src_ip, const u8 *src_mac,
87613de9 1626 u32 src_ifindex, __be32 vni)
d342894c 1627{
1628 struct vxlan_dev *vxlan = netdev_priv(dev);
1629 struct vxlan_fdb *f;
87613de9
MS		 1630 u32 ifindex = 0;
1631
1632#if IS_ENABLED(CONFIG_IPV6)
1633 if (src_ip->sa.sa_family == AF_INET6 &&
1634 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1635 ifindex = src_ifindex;
1636#endif
d342894c 1637
3ad7a4b1 1638 f = vxlan_find_mac(vxlan, src_mac, vni);
d342894c 1639 if (likely(f)) {
5ca5461c 1640 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 1641
87613de9
MS		 1642 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1643 rdst->remote_ifindex == ifindex))
26a41ae6 1644 return false;
1645
1646 /* Don't migrate static entries, drop packets */
e0090a9e 1647 if (f->state & (NUD_PERMANENT | NUD_NOARP))
26a41ae6 1648 return true;
d342894c 1649
1274e1cc
RP		 1650 /* Don't override an fdb with nexthop with a learnt entry */
1651 if (rcu_access_pointer(f->nh))
1652 return true;
1653
d342894c 1654 if (net_ratelimit())
1655 netdev_info(dev,
e4c7ed41 1656 "%pM migrated from %pIS to %pIS\n",
a4870f79 1657 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 1658
e4c7ed41 1659 rdst->remote_ip = *src_ip;
d342894c 1660 f->updated = jiffies;
4c59b7d1 1661 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
d342894c 1662 } else {
fe1e0713
L		 1663 u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1664
d342894c 1665 /* learned new entry */
fe1e0713 1666 spin_lock(&vxlan->hash_lock[hash_index]);
3bf74b1a 1667
1668 /* close off race between vxlan_flush and incoming packets */
1669 if (netif_running(dev))
25e20e73 1670 vxlan_fdb_update(vxlan, src_mac, src_ip,
3bf74b1a 1671 NUD_REACHABLE,
1672 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 1673 vxlan->cfg.dst_port,
3ad7a4b1 1674 vni,
3bf74b1a 1675 vxlan->default_dst.remote_vni,
1274e1cc 1676 ifindex, NTF_SELF, 0, true, NULL);
fe1e0713 1677 spin_unlock(&vxlan->hash_lock[hash_index]);
d342894c 1678 }
26a41ae6 1679
1680 return false;
d342894c 1681}
1682
d342894c 1683/* See if multicast group is already in use by other ID */
95ab0991 1684static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1685{
553675fb 1686 struct vxlan_dev *vxlan;
c6fcc4fc 1687 struct vxlan_sock *sock4;
4053ab1b
AB		 1688#if IS_ENABLED(CONFIG_IPV6)
1689 struct vxlan_sock *sock6;
1690#endif
b1be00a6 1691 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1692
c6fcc4fc 1693 sock4 = rtnl_dereference(dev->vn4_sock);
1694
95ab0991
G		 1695 /* The vxlan_sock is only used by dev, leaving group has
1696 * no effect on other vxlan devices.
1697 */
66af846f 1698 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
95ab0991 1699 return false;
b1be00a6 1700#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1701 sock6 = rtnl_dereference(dev->vn6_sock);
66af846f 1702 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
b1be00a6
JB		 1703 return false;
1704#endif
95ab0991 1705
553675fb 1706 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1707 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1708 continue;
d342894c 1709
c6fcc4fc 1710 if (family == AF_INET &&
1711 rtnl_dereference(vxlan->vn4_sock) != sock4)
95ab0991 1712 continue;
b1be00a6 1713#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1714 if (family == AF_INET6 &&
1715 rtnl_dereference(vxlan->vn6_sock) != sock6)
b1be00a6
JB		 1716 continue;
1717#endif
95ab0991
G		 1718
1719 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1720 &dev->default_dst.remote_ip))
1721 continue;
1722
1723 if (vxlan->default_dst.remote_ifindex !=
1724 dev->default_dst.remote_ifindex)
1725 continue;
1726
1727 return true;
553675fb 1728 }
d342894c 1729
1730 return false;
1731}
1732
544a773a 1733static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
7c47cedf 1734{
b1be00a6 1735 struct vxlan_net *vn;
012a5729 1736
b1be00a6 1737 if (!vs)
544a773a 1738 return false;
66af846f 1739 if (!refcount_dec_and_test(&vs->refcnt))
544a773a 1740 return false;
d342894c 1741
b1be00a6 1742 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1743 spin_lock(&vn->sock_lock);
7c47cedf 1744 hlist_del_rcu(&vs->hlist);
e7b3db5e 1745 udp_tunnel_notify_del_rx_port(vs->sock,
b9adcd69
AD		 1746 (vs->flags & VXLAN_F_GPE) ?
1747 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 1748 UDP_TUNNEL_TYPE_VXLAN);
1c51a915
SH		 1749 spin_unlock(&vn->sock_lock);
1750
544a773a 1751 return true;
d342894c 1752}
1753
b1be00a6
JB		 1754static void vxlan_sock_release(struct vxlan_dev *vxlan)
1755{
c6fcc4fc 1756 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
b1be00a6 1757#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1758 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1759
57d88182 1760 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
544a773a
HFS		 1761#endif
1762
57d88182 1763 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
544a773a
HFS		 1764 synchronize_net();
1765
a53cb29b
MB		 1766 vxlan_vs_del_dev(vxlan);
1767
c6fcc4fc 1768 if (__vxlan_sock_release_prep(sock4)) {
1769 udp_tunnel_sock_release(sock4->sock);
1770 kfree(sock4);
544a773a
HFS		 1771 }
1772
1773#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1774 if (__vxlan_sock_release_prep(sock6)) {
1775 udp_tunnel_sock_release(sock6->sock);
1776 kfree(sock6);
544a773a 1777 }
b1be00a6
JB		 1778#endif
1779}
1780
56ef9c90 1781/* Update multicast group membership when first VNI on
c4b49512 1782 * multicast address is brought up
7c47cedf 1783 */
56ef9c90 1784static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1785{
b1be00a6 1786 struct sock *sk;
e4c7ed41
CW		 1787 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1788 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1789 int ret = -EINVAL;
d342894c 1790
e4c7ed41 1791 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1792 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1793 struct ip_mreqn mreq = {
1794 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1795 .imr_ifindex = ifindex,
1796 };
1797
c6fcc4fc 1798 sk = sock4->sock->sk;
b1be00a6 1799 lock_sock(sk);
56ef9c90 1800 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1801 release_sock(sk);
e4c7ed41
CW		 1802#if IS_ENABLED(CONFIG_IPV6)
1803 } else {
c6fcc4fc 1804 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1805
1806 sk = sock6->sock->sk;
b1be00a6 1807 lock_sock(sk);
56ef9c90
MRL		 1808 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1809 &ip->sin6.sin6_addr);
b1be00a6 1810 release_sock(sk);
e4c7ed41
CW		 1811#endif
1812 }
3fc2de2f 1813
56ef9c90 1814 return ret;
3fc2de2f 1815}
1816
1817/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1818static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1819{
b1be00a6 1820 struct sock *sk;
e4c7ed41
CW		 1821 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1822 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1823 int ret = -EINVAL;
3fc2de2f 1824
e4c7ed41 1825 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1826 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1827 struct ip_mreqn mreq = {
1828 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1829 .imr_ifindex = ifindex,
1830 };
1831
c6fcc4fc 1832 sk = sock4->sock->sk;
b1be00a6 1833 lock_sock(sk);
56ef9c90 1834 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1835 release_sock(sk);
e4c7ed41
CW		 1836#if IS_ENABLED(CONFIG_IPV6)
1837 } else {
c6fcc4fc 1838 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1839
1840 sk = sock6->sock->sk;
b1be00a6 1841 lock_sock(sk);
56ef9c90
MRL		 1842 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1843 &ip->sin6.sin6_addr);
b1be00a6 1844 release_sock(sk);
e4c7ed41
CW		 1845#endif
1846 }
d342894c 1847
56ef9c90 1848 return ret;
d342894c 1849}
1850
f14ecebb
JB		 1851static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1852 struct sk_buff *skb, u32 vxflags)
dfd8645e 1853{
7d34fa75 1854 size_t start, offset;
dfd8645e 1855
f14ecebb
JB		 1856 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1857 goto out;
b7fe10e5 1858
f14ecebb
JB		 1859 start = vxlan_rco_start(unparsed->vx_vni);
1860 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1861
7d34fa75 1862 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1863 return false;
dfd8645e 1864
be5cfeab
JB		 1865 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1866 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1867out:
1868 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1869 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1870 return true;
dfd8645e
TH		 1871}
1872
f14ecebb 1873static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1874 struct sk_buff *skb, u32 vxflags,
10a5af23 1875 struct vxlan_metadata *md)
3288af08 1876{
f14ecebb 1877 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1878 struct metadata_dst *tun_dst;
f14ecebb
JB		 1879
1880 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1881 goto out;
3288af08 1882
3288af08
JB		 1883 md->gbp = ntohs(gbp->policy_id);
1884
10a5af23 1885 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1886 if (tun_dst) {
3288af08 1887 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1888 tun_dst->u.tun_info.options_len = sizeof(*md);
1889 }
3288af08
JB		 1890 if (gbp->dont_learn)
1891 md->gbp |= VXLAN_GBP_DONT_LEARN;
1892
1893 if (gbp->policy_applied)
1894 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1895
64f87d36
JB		 1896 /* In flow-based mode, GBP is carried in dst_metadata */
1897 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1898 skb->mark = md->gbp;
f14ecebb
JB		 1899out:
1900 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1901}
1902
e1e5314d 1903static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
61618eea 1904 __be16 *protocol,
e1e5314d
JB		 1905 struct sk_buff *skb, u32 vxflags)
1906{
1907 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1908
1909 /* Need to have Next Protocol set for interfaces in GPE mode. */
1910 if (!gpe->np_applied)
1911 return false;
1912 /* "The initial version is 0. If a receiver does not support the
1913 * version indicated it MUST drop the packet.
1914 */
1915 if (gpe->version != 0)
1916 return false;
1917 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1918 * processing MUST occur." However, we don't implement OAM
1919 * processing, thus drop the packet.
1920 */
1921 if (gpe->oam_flag)
1922 return false;
1923
fa20e0e3
JB		 1924 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1925 if (!*protocol)
e1e5314d 1926 return false;
e1e5314d
JB		 1927
1928 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1929 return true;
1930}
1931
1ab016e2
JB		 1932static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1933 struct vxlan_sock *vs,
3ad7a4b1 1934 struct sk_buff *skb, __be32 vni)
614732ea 1935{
614732ea 1936 union vxlan_addr saddr;
87613de9 1937 u32 ifindex = skb->dev->ifindex;
614732ea 1938
614732ea 1939 skb_reset_mac_header(skb);
614732ea
TG		 1940 skb->protocol = eth_type_trans(skb, vxlan->dev);
1941 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1942
1943 /* Ignore packet loops (and multicast echo) */
1944 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1945 return false;
614732ea 1946
760c6805 1947 /* Get address from the outer IP header */
ce212d0f 1948 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1949 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1950 saddr.sa.sa_family = AF_INET;
1951#if IS_ENABLED(CONFIG_IPV6)
1952 } else {
1ab016e2 1953 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1954 saddr.sa.sa_family = AF_INET6;
1955#endif
1956 }
1957
dc5321d7 1958 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
87613de9 1959 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1ab016e2
JB		 1960 return false;
1961
1962 return true;
1963}
1964
760c6805
JB		 1965static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1966 struct sk_buff *skb)
1967{
1968 int err = 0;
1969
1970 if (vxlan_get_sk_family(vs) == AF_INET)
1971 err = IP_ECN_decapsulate(oiph, skb);
1972#if IS_ENABLED(CONFIG_IPV6)
1973 else
1974 err = IP6_ECN_decapsulate(oiph, skb);
1975#endif
1976
1977 if (unlikely(err) && log_ecn_error) {
1978 if (vxlan_get_sk_family(vs) == AF_INET)
1979 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1980 &((struct iphdr *)oiph)->saddr,
1981 ((struct iphdr *)oiph)->tos);
1982 else
1983 net_info_ratelimited("non-ECT from %pI6\n",
1984 &((struct ipv6hdr *)oiph)->saddr);
1985 }
1986 return err <= 1;
1987}
1988
d342894c 1989/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1990static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1991{
c9e78efb 1992 struct vxlan_dev *vxlan;
5cfccc5a 1993 struct vxlan_sock *vs;
f14ecebb 1994 struct vxlanhdr unparsed;
ee122c79
TG		 1995 struct vxlan_metadata _md;
1996 struct vxlan_metadata *md = &_md;
61618eea 1997 __be16 protocol = htons(ETH_P_TEB);
e1e5314d 1998 bool raw_proto = false;
f2d1968e 1999 void *oiph;
3ad7a4b1 2000 __be32 vni = 0;
d342894c 2001
e1e5314d 2002 /* Need UDP and VXLAN header to be present */
7ce04758 2003 if (!pskb_may_pull(skb, VXLAN_HLEN))
e5aed006 2004 goto drop;
d342894c 2005
f14ecebb 2006 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 2007 /* VNI flag always required to be set */
2008 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
2009 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
2010 ntohl(vxlan_hdr(skb)->vx_flags),
2011 ntohl(vxlan_hdr(skb)->vx_vni));
2012 /* Return non vxlan pkt */
e5aed006 2013 goto drop;
d342894c 2014 }
288b01c8
JB		 2015 unparsed.vx_flags &= ~VXLAN_HF_VNI;
2016 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 2017
559835ea 2018 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 2019 if (!vs)
d342894c 2020 goto drop;
d342894c 2021
3ad7a4b1
RP		 2022 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
2023
49f810f0 2024 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
c9e78efb
JB		 2025 if (!vxlan)
2026 goto drop;
2027
e1e5314d
JB		 2028 /* For backwards compatibility, only allow reserved fields to be
2029 * used by VXLAN extensions if explicitly requested.
2030 */
2031 if (vs->flags & VXLAN_F_GPE) {
2032 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
2033 goto drop;
2034 raw_proto = true;
2035 }
2036
2037 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
2038 !net_eq(vxlan->net, dev_net(vxlan->dev))))
98c81476 2039 goto drop;
c9e78efb 2040
2ae2904b 2041 if (vs->flags & VXLAN_F_REMCSUM_RX)
0189399c 2042 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags)))
2ae2904b
FF		 2043 goto drop;
2044
ee122c79 2045 if (vxlan_collect_metadata(vs)) {
10a5af23 2046 struct metadata_dst *tun_dst;
07dabf20 2047
c29a70d2 2048 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
d817f432 2049 key32_to_tunnel_id(vni), sizeof(*md));
c29a70d2 2050
ee122c79
TG		 2051 if (!tun_dst)
2052 goto drop;
2053
0f1b7354 2054 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 2055
2056 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 2057 } else {
2058 memset(md, 0, sizeof(*md));
2059 }
2060
f14ecebb 2061 if (vs->flags & VXLAN_F_GBP)
10a5af23 2062 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB		 2063 /* Note that GBP and GPE can never be active together. This is
2064 * ensured in vxlan_dev_configure.
2065 */
3511494c 2066
f14ecebb 2067 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 2068 /* If there are any unprocessed flags remaining treat
2069 * this as a malformed packet. This behavior diverges from
2070 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
2071 * in reserved fields are to be ignored. The approach here
c4b49512 2072 * maintains compatibility with previous stack code, and also
3bf39475
TH		 2073 * is more robust and provides a little more security in
2074 * adding extensions to VXLAN.
2075 */
288b01c8 2076 goto drop;
3bf39475
TH		 2077 }
2078
e1e5314d 2079 if (!raw_proto) {
3ad7a4b1 2080 if (!vxlan_set_mac(vxlan, vs, skb, vni))
e1e5314d
JB		 2081 goto drop;
2082 } else {
8be0cfa4 2083 skb_reset_mac_header(skb);
e1e5314d
JB		 2084 skb->dev = vxlan->dev;
2085 skb->pkt_type = PACKET_HOST;
2086 }
f2d1968e 2087
f2d1968e
JB		 2088 oiph = skb_network_header(skb);
2089 skb_reset_network_header(skb);
2090
2091 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
2092 ++vxlan->dev->stats.rx_frame_errors;
2093 ++vxlan->dev->stats.rx_errors;
2094 goto drop;
2095 }
2096
59cbf56f
ED		 2097 rcu_read_lock();
2098
2099 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
2100 rcu_read_unlock();
2101 atomic_long_inc(&vxlan->dev->rx_dropped);
2102 goto drop;
2103 }
2104
1f8dda1d 2105 dev_sw_netstats_rx_add(vxlan->dev, skb->len);
f2d1968e 2106 gro_cells_receive(&vxlan->gro_cells, skb);
59cbf56f
ED		 2107
2108 rcu_read_unlock();
2109
5cfccc5a
PS		 2110 return 0;
2111
2112drop:
288b01c8
JB		 2113 /* Consume bad packet */
2114 kfree_skb(skb);
2115 return 0;
5cfccc5a
PS		 2116}
2117
c3a43b9f
SB		 2118/* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
2119static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
2120{
2121 struct vxlan_dev *vxlan;
2122 struct vxlan_sock *vs;
2123 struct vxlanhdr *hdr;
2124 __be32 vni;
2125
8399a693 2126 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
c3a43b9f
SB		 2127 return -EINVAL;
2128
2129 hdr = vxlan_hdr(skb);
2130
2131 if (!(hdr->vx_flags & VXLAN_HF_VNI))
2132 return -EINVAL;
2133
2134 vs = rcu_dereference_sk_user_data(sk);
2135 if (!vs)
2136 return -ENOENT;
2137
2138 vni = vxlan_vni(hdr->vx_vni);
2139 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
2140 if (!vxlan)
2141 return -ENOENT;
2142
2143 return 0;
2144}
2145
3ad7a4b1 2146static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
e4f67add
DS		 2147{
2148 struct vxlan_dev *vxlan = netdev_priv(dev);
2149 struct arphdr *parp;
2150 u8 *arpptr, *sha;
2151 __be32 sip, tip;
2152 struct neighbour *n;
2153
2154 if (dev->flags & IFF_NOARP)
2155 goto out;
2156
2157 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
2158 dev->stats.tx_dropped++;
2159 goto out;
2160 }
2161 parp = arp_hdr(skb);
2162
2163 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
2164 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
2165 parp->ar_pro != htons(ETH_P_IP) ||
2166 parp->ar_op != htons(ARPOP_REQUEST) ||
2167 parp->ar_hln != dev->addr_len ||
2168 parp->ar_pln != 4)
2169 goto out;
2170 arpptr = (u8 *)parp + sizeof(struct arphdr);
2171 sha = arpptr;
2172 arpptr += dev->addr_len; /* sha */
2173 memcpy(&sip, arpptr, sizeof(sip));
2174 arpptr += sizeof(sip);
2175 arpptr += dev->addr_len; /* tha */
2176 memcpy(&tip, arpptr, sizeof(tip));
2177
2178 if (ipv4_is_loopback(tip) ||
2179 ipv4_is_multicast(tip))
2180 goto out;
2181
2182 n = neigh_lookup(&arp_tbl, &tip, dev);
2183
2184 if (n) {
e4f67add
DS		 2185 struct vxlan_fdb *f;
2186 struct sk_buff *reply;
2187
2188 if (!(n->nud_state & NUD_CONNECTED)) {
2189 neigh_release(n);
2190 goto out;
2191 }
2192
3ad7a4b1 2193 f = vxlan_find_mac(vxlan, n->ha, vni);
e4c7ed41 2194 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 2195 /* bridge-local neighbor */
2196 neigh_release(n);
2197 goto out;
2198 }
2199
2200 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
2201 n->ha, sha);
2202
2203 neigh_release(n);
2204
7346135d
DS		 2205 if (reply == NULL)
2206 goto out;
2207
e4f67add
DS		 2208 skb_reset_mac_header(reply);
2209 __skb_pull(reply, skb_network_offset(reply));
2210 reply->ip_summed = CHECKSUM_UNNECESSARY;
2211 reply->pkt_type = PACKET_HOST;
2212
2213 if (netif_rx_ni(reply) == NET_RX_DROP)
2214 dev->stats.rx_dropped++;
dc5321d7 2215 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
e4c7ed41
CW		 2216 union vxlan_addr ipa = {
2217 .sin.sin_addr.s_addr = tip,
a45e92a5 2218 .sin.sin_family = AF_INET,
e4c7ed41
CW		 2219 };
2220
2221 vxlan_ip_miss(dev, &ipa);
2222 }
e4f67add
DS		 2223out:
2224 consume_skb(skb);
2225 return NETDEV_TX_OK;
2226}
2227
f564f45c 2228#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 2229static struct sk_buff *vxlan_na_create(struct sk_buff *request,
2230 struct neighbour *n, bool isrouter)
2231{
2232 struct net_device *dev = request->dev;
2233 struct sk_buff *reply;
2234 struct nd_msg *ns, *na;
2235 struct ipv6hdr *pip6;
2236 u8 *daddr;
2237 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
2238 int ns_olen;
2239 int i, len;
2240
f1fb08f6 2241 if (dev == NULL || !pskb_may_pull(request, request->len))
4b29dba9
DS		 2242 return NULL;
2243
2244 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
2245 sizeof(*na) + na_olen + dev->needed_tailroom;
2246 reply = alloc_skb(len, GFP_ATOMIC);
2247 if (reply == NULL)
2248 return NULL;
2249
2250 reply->protocol = htons(ETH_P_IPV6);
2251 reply->dev = dev;
2252 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
2253 skb_push(reply, sizeof(struct ethhdr));
6297b91c 2254 skb_reset_mac_header(reply);
4b29dba9 2255
f1fb08f6 2256 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
4b29dba9
DS		 2257
2258 daddr = eth_hdr(request)->h_source;
f1fb08f6
VB		 2259 ns_olen = request->len - skb_network_offset(request) -
2260 sizeof(struct ipv6hdr) - sizeof(*ns);
4b29dba9 2261 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
8066e6b4
IS		 2262 if (!ns->opt[i + 1]) {
2263 kfree_skb(reply);
2264 return NULL;
2265 }
4b29dba9
DS		 2266 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
2267 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
2268 break;
2269 }
2270 }
2271
2272 /* Ethernet header */
2273 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
2274 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
2275 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
2276 reply->protocol = htons(ETH_P_IPV6);
2277
2278 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 2279 skb_reset_network_header(reply);
4b29dba9
DS		 2280 skb_put(reply, sizeof(struct ipv6hdr));
2281
2282 /* IPv6 header */
2283
2284 pip6 = ipv6_hdr(reply);
2285 memset(pip6, 0, sizeof(struct ipv6hdr));
2286 pip6->version = 6;
2287 pip6->priority = ipv6_hdr(request)->priority;
2288 pip6->nexthdr = IPPROTO_ICMPV6;
2289 pip6->hop_limit = 255;
2290 pip6->daddr = ipv6_hdr(request)->saddr;
2291 pip6->saddr = *(struct in6_addr *)n->primary_key;
2292
2293 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 2294 skb_reset_transport_header(reply);
4b29dba9 2295
4b29dba9 2296 /* Neighbor Advertisement */
b080db58 2297 na = skb_put_zero(reply, sizeof(*na) + na_olen);
4b29dba9
DS		 2298 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
2299 na->icmph.icmp6_router = isrouter;
2300 na->icmph.icmp6_override = 1;
2301 na->icmph.icmp6_solicited = 1;
2302 na->target = ns->target;
2303 ether_addr_copy(&na->opt[2], n->ha);
2304 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
2305 na->opt[1] = na_olen >> 3;
2306
2307 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
2308 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
2309 csum_partial(na, sizeof(*na)+na_olen, 0));
2310
2311 pip6->payload_len = htons(sizeof(*na)+na_olen);
2312
2313 skb_push(reply, sizeof(struct ipv6hdr));
2314
2315 reply->ip_summed = CHECKSUM_UNNECESSARY;
2316
2317 return reply;
2318}
2319
3ad7a4b1 2320static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
f564f45c
CW		 2321{
2322 struct vxlan_dev *vxlan = netdev_priv(dev);
8dcd81a9 2323 const struct in6_addr *daddr;
8bff3685 2324 const struct ipv6hdr *iphdr;
4b29dba9 2325 struct inet6_dev *in6_dev;
8bff3685
XL		 2326 struct neighbour *n;
2327 struct nd_msg *msg;
f564f45c
CW		 2328
2329 in6_dev = __in6_dev_get(dev);
2330 if (!in6_dev)
2331 goto out;
2332
f564f45c 2333 iphdr = ipv6_hdr(skb);
f564f45c 2334 daddr = &iphdr->daddr;
f1fb08f6 2335 msg = (struct nd_msg *)(iphdr + 1);
f564f45c 2336
4b29dba9
DS		 2337 if (ipv6_addr_loopback(daddr) ||
2338 ipv6_addr_is_multicast(&msg->target))
2339 goto out;
2340
2341 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 2342
2343 if (n) {
2344 struct vxlan_fdb *f;
4b29dba9 2345 struct sk_buff *reply;
f564f45c
CW		 2346
2347 if (!(n->nud_state & NUD_CONNECTED)) {
2348 neigh_release(n);
2349 goto out;
2350 }
2351
3ad7a4b1 2352 f = vxlan_find_mac(vxlan, n->ha, vni);
f564f45c
CW		 2353 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2354 /* bridge-local neighbor */
2355 neigh_release(n);
2356 goto out;
2357 }
2358
4b29dba9
DS		 2359 reply = vxlan_na_create(skb, n,
2360 !!(f ? f->flags & NTF_ROUTER : 0));
2361
f564f45c 2362 neigh_release(n);
4b29dba9
DS		 2363
2364 if (reply == NULL)
2365 goto out;
2366
2367 if (netif_rx_ni(reply) == NET_RX_DROP)
2368 dev->stats.rx_dropped++;
2369
dc5321d7 2370 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 2371 union vxlan_addr ipa = {
2372 .sin6.sin6_addr = msg->target,
a45e92a5 2373 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 2374 };
2375
f564f45c
CW		 2376 vxlan_ip_miss(dev, &ipa);
2377 }
2378
2379out:
2380 consume_skb(skb);
2381 return NETDEV_TX_OK;
2382}
2383#endif
2384
e4f67add
DS		 2385static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2386{
2387 struct vxlan_dev *vxlan = netdev_priv(dev);
2388 struct neighbour *n;
e4f67add
DS		 2389
2390 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2391 return false;
2392
2393 n = NULL;
2394 switch (ntohs(eth_hdr(skb)->h_proto)) {
2395 case ETH_P_IP:
e15a00aa
CW		 2396 {
2397 struct iphdr *pip;
2398
e4f67add
DS		 2399 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2400 return false;
2401 pip = ip_hdr(skb);
2402 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
dc5321d7 2403 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e4c7ed41
CW		 2404 union vxlan_addr ipa = {
2405 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 2406 .sin.sin_family = AF_INET,
e4c7ed41
CW		 2407 };
2408
2409 vxlan_ip_miss(dev, &ipa);
2410 return false;
2411 }
2412
e4f67add 2413 break;
e15a00aa
CW		 2414 }
2415#if IS_ENABLED(CONFIG_IPV6)
2416 case ETH_P_IPV6:
2417 {
2418 struct ipv6hdr *pip6;
2419
2420 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2421 return false;
2422 pip6 = ipv6_hdr(skb);
2423 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
dc5321d7 2424 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e15a00aa
CW		 2425 union vxlan_addr ipa = {
2426 .sin6.sin6_addr = pip6->daddr,
a45e92a5 2427 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 2428 };
2429
2430 vxlan_ip_miss(dev, &ipa);
2431 return false;
2432 }
2433
2434 break;
2435 }
2436#endif
e4f67add
DS		 2437 default:
2438 return false;
2439 }
2440
2441 if (n) {
2442 bool diff;
2443
7367d0b5 2444 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 2445 if (diff) {
2446 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2447 dev->addr_len);
2448 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2449 }
2450 neigh_release(n);
2451 return diff;
e4c7ed41
CW		 2452 }
2453
e4f67add
DS		 2454 return false;
2455}
2456
af33c1ad 2457static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 2458 struct vxlan_metadata *md)
2459{
2460 struct vxlanhdr_gbp *gbp;
2461
db79a621
TG		 2462 if (!md->gbp)
2463 return;
2464
3511494c 2465 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 2466 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 2467
2468 if (md->gbp & VXLAN_GBP_DONT_LEARN)
2469 gbp->dont_learn = 1;
2470
2471 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
2472 gbp->policy_applied = 1;
2473
2474 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
2475}
2476
e1e5314d
JB		 2477static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
2478 __be16 protocol)
2479{
2480 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2481
2482 gpe->np_applied = 1;
fa20e0e3
JB		 2483 gpe->next_protocol = tun_p_from_eth_p(protocol);
2484 if (!gpe->next_protocol)
2485 return -EPFNOSUPPORT;
2486 return 0;
e1e5314d
JB		 2487}
2488
f491e56d
JB		 2489static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2490 int iphdr_len, __be32 vni,
2491 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 2492 bool udp_sum)
e4c7ed41 2493{
e4c7ed41 2494 struct vxlanhdr *vxh;
e4c7ed41
CW		 2495 int min_headroom;
2496 int err;
dfd8645e 2497 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 2498 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 2499
af33c1ad 2500 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 2501 skb->ip_summed == CHECKSUM_PARTIAL) {
2502 int csum_start = skb_checksum_start_offset(skb);
2503
2504 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2505 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2506 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 2507 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 2508 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 2509 }
e4c7ed41 2510
e4c7ed41 2511 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
4a4f86cc 2512 + VXLAN_HLEN + iphdr_len;
649c5b8b
PS		 2513
2514 /* Need space for new headers (invalidates iph ptr) */
2515 err = skb_cow_head(skb, min_headroom);
e1e5314d 2516 if (unlikely(err))
c46b7897 2517 return err;
649c5b8b 2518
aed069df
AD		 2519 err = iptunnel_handle_offloads(skb, type);
2520 if (err)
c46b7897 2521 return err;
b736a623 2522
d58ff351 2523 vxh = __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 2524 vxh->vx_flags = VXLAN_HF_VNI;
2525 vxh->vx_vni = vxlan_vni_field(vni);
49560532 2526
dfd8645e 2527 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 2528 unsigned int start;
dfd8645e 2529
54bfd872
JB		 2530 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2531 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2532 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 2533
2534 if (!skb_is_gso(skb)) {
2535 skb->ip_summed = CHECKSUM_NONE;
2536 skb->encapsulation = 0;
2537 }
2538 }
2539
af33c1ad
TH		 2540 if (vxflags & VXLAN_F_GBP)
2541 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB		 2542 if (vxflags & VXLAN_F_GPE) {
2543 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2544 if (err < 0)
c46b7897 2545 return err;
e1e5314d
JB		 2546 inner_protocol = skb->protocol;
2547 }
3511494c 2548
e1e5314d 2549 skb_set_inner_protocol(skb, inner_protocol);
039f5062 2550 return 0;
49560532 2551}
49560532 2552
655c3de1 2553static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2554 struct vxlan_sock *sock4,
1a8496ba 2555 struct sk_buff *skb, int oif, u8 tos,
4ecb1d83 2556 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
0c1d70af 2557 struct dst_cache *dst_cache,
db3c6139 2558 const struct ip_tunnel_info *info)
1a8496ba 2559{
db3c6139 2560 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 2561 struct rtable *rt = NULL;
2562 struct flowi4 fl4;
2563
655c3de1 2564 if (!sock4)
2565 return ERR_PTR(-EIO);
2566
db3c6139
DB		 2567 if (tos && !info)
2568 use_cache = false;
2569 if (use_cache) {
0c1d70af
PA		 2570 rt = dst_cache_get_ip4(dst_cache, saddr);
2571 if (rt)
2572 return rt;
2573 }
2574
1a8496ba
JB		 2575 memset(&fl4, 0, sizeof(fl4));
2576 fl4.flowi4_oif = oif;
2577 fl4.flowi4_tos = RT_TOS(tos);
2578 fl4.flowi4_mark = skb->mark;
2579 fl4.flowi4_proto = IPPROTO_UDP;
2580 fl4.daddr = daddr;
272d96a5 2581 fl4.saddr = *saddr;
4ecb1d83
MP		 2582 fl4.fl4_dport = dport;
2583 fl4.fl4_sport = sport;
1a8496ba
JB		 2584
2585 rt = ip_route_output_key(vxlan->net, &fl4);
478db1f1 2586 if (!IS_ERR(rt)) {
655c3de1 2587 if (rt->dst.dev == dev) {
2588 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2589 ip_rt_put(rt);
2590 return ERR_PTR(-ELOOP);
2591 }
2592
1a8496ba 2593 *saddr = fl4.saddr;
0c1d70af
PA		 2594 if (use_cache)
2595 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
655c3de1 2596 } else {
2597 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2598 return ERR_PTR(-ENETUNREACH);
0c1d70af 2599 }
1a8496ba
JB		 2600 return rt;
2601}
2602
e5d4b29f
JB		 2603#if IS_ENABLED(CONFIG_IPV6)
2604static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
655c3de1 2605 struct net_device *dev,
03dc52a8 2606 struct vxlan_sock *sock6,
1400615d 2607 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 2608 __be32 label,
e5d4b29f 2609 const struct in6_addr *daddr,
0c1d70af 2610 struct in6_addr *saddr,
4ecb1d83 2611 __be16 dport, __be16 sport,
db3c6139
DB		 2612 struct dst_cache *dst_cache,
2613 const struct ip_tunnel_info *info)
e5d4b29f 2614{
db3c6139 2615 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 2616 struct dst_entry *ndst;
2617 struct flowi6 fl6;
e5d4b29f 2618
c6fcc4fc 2619 if (!sock6)
2620 return ERR_PTR(-EIO);
2621
1400615d
DB		 2622 if (tos && !info)
2623 use_cache = false;
db3c6139 2624 if (use_cache) {
0c1d70af
PA		 2625 ndst = dst_cache_get_ip6(dst_cache, saddr);
2626 if (ndst)
2627 return ndst;
2628 }
2629
e5d4b29f
JB		 2630 memset(&fl6, 0, sizeof(fl6));
2631 fl6.flowi6_oif = oif;
2632 fl6.daddr = *daddr;
272d96a5 2633 fl6.saddr = *saddr;
eaa93bf4 2634 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB		 2635 fl6.flowi6_mark = skb->mark;
2636 fl6.flowi6_proto = IPPROTO_UDP;
4ecb1d83
MP		 2637 fl6.fl6_dport = dport;
2638 fl6.fl6_sport = sport;
e5d4b29f 2639
6c8991f4
SD		 2640 ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
2641 &fl6, NULL);
a10b24b8 2642 if (IS_ERR(ndst)) {
655c3de1 2643 netdev_dbg(dev, "no route to %pI6\n", daddr);
2644 return ERR_PTR(-ENETUNREACH);
2645 }
2646
2647 if (unlikely(ndst->dev == dev)) {
2648 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2649 dst_release(ndst);
2650 return ERR_PTR(-ELOOP);
2651 }
e5d4b29f
JB		 2652
2653 *saddr = fl6.saddr;
db3c6139 2654 if (use_cache)
0c1d70af 2655 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 2656 return ndst;
2657}
2658#endif
2659
9dcc71e1
SS		 2660/* Bypass encapsulation if the destination is local */
2661static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
fc68c995
SB		 2662 struct vxlan_dev *dst_vxlan, __be32 vni,
2663 bool snoop)
9dcc71e1 2664{
8f84985f 2665 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 2666 union vxlan_addr loopback;
2667 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
4179cb5a 2668 struct net_device *dev;
ce6502a8 2669 int len = skb->len;
9dcc71e1 2670
8f84985f
LR		 2671 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2672 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 2673 skb->pkt_type = PACKET_HOST;
2674 skb->encapsulation = 0;
2675 skb->dev = dst_vxlan->dev;
2676 __skb_pull(skb, skb_network_offset(skb));
2677
e4c7ed41
CW		 2678 if (remote_ip->sa.sa_family == AF_INET) {
2679 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2680 loopback.sa.sa_family = AF_INET;
2681#if IS_ENABLED(CONFIG_IPV6)
2682 } else {
2683 loopback.sin6.sin6_addr = in6addr_loopback;
2684 loopback.sa.sa_family = AF_INET6;
2685#endif
2686 }
2687
4179cb5a
ED		 2688 rcu_read_lock();
2689 dev = skb->dev;
2690 if (unlikely(!(dev->flags & IFF_UP))) {
2691 kfree_skb(skb);
2692 goto drop;
2693 }
2694
fc68c995 2695 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
4179cb5a 2696 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
9dcc71e1
SS		 2697
2698 u64_stats_update_begin(&tx_stats->syncp);
2699 tx_stats->tx_packets++;
ce6502a8 2700 tx_stats->tx_bytes += len;
9dcc71e1
SS		 2701 u64_stats_update_end(&tx_stats->syncp);
2702
2703 if (netif_rx(skb) == NET_RX_SUCCESS) {
2704 u64_stats_update_begin(&rx_stats->syncp);
2705 rx_stats->rx_packets++;
ce6502a8 2706 rx_stats->rx_bytes += len;
9dcc71e1
SS		 2707 u64_stats_update_end(&rx_stats->syncp);
2708 } else {
4179cb5a 2709drop:
ce6502a8 2710 dev->stats.rx_dropped++;
9dcc71e1 2711 }
4179cb5a 2712 rcu_read_unlock();
9dcc71e1
SS		 2713}
2714
fee1fad7 2715static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
49f810f0
MS		 2716 struct vxlan_dev *vxlan,
2717 union vxlan_addr *daddr,
2718 __be16 dst_port, int dst_ifindex, __be32 vni,
2719 struct dst_entry *dst,
fee1fad7 2720 u32 rt_flags)
2721{
2722#if IS_ENABLED(CONFIG_IPV6)
2723 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2724 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2725 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2726 */
2727 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2728#endif
2729 /* Bypass encapsulation if the destination is local */
2730 if (rt_flags & RTCF_LOCAL &&
2731 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2732 struct vxlan_dev *dst_vxlan;
2733
2734 dst_release(dst);
49f810f0 2735 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
fee1fad7 2736 daddr->sa.sa_family, dst_port,
dc5321d7 2737 vxlan->cfg.flags);
fee1fad7 2738 if (!dst_vxlan) {
2739 dev->stats.tx_errors++;
2740 kfree_skb(skb);
2741
2742 return -ENOENT;
2743 }
fc68c995 2744 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
fee1fad7 2745 return 1;
2746 }
2747
2748 return 0;
2749}
2750
4ad16930 2751static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
3ad7a4b1
RP		 2752 __be32 default_vni, struct vxlan_rdst *rdst,
2753 bool did_rsc)
d342894c 2754{
d71785ff 2755 struct dst_cache *dst_cache;
3093fbe7 2756 struct ip_tunnel_info *info;
d342894c 2757 struct vxlan_dev *vxlan = netdev_priv(dev);
0770b53b 2758 const struct iphdr *old_iph = ip_hdr(skb);
e4c7ed41 2759 union vxlan_addr *dst;
272d96a5 2760 union vxlan_addr remote_ip, local_ip;
ee122c79
TG		 2761 struct vxlan_metadata _md;
2762 struct vxlan_metadata *md = &_md;
e4c7ed41 2763 __be16 src_port = 0, dst_port;
655c3de1 2764 struct dst_entry *ndst = NULL;
e7f70af1 2765 __be32 vni, label;
d342894c 2766 __u8 tos, ttl;
49f810f0 2767 int ifindex;
0e6fbc5b 2768 int err;
dc5321d7 2769 u32 flags = vxlan->cfg.flags;
b4ed5cad 2770 bool udp_sum = false;
f491e56d 2771 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 2772
61adedf3 2773 info = skb_tunnel_info(skb);
3093fbe7 2774
ee122c79 2775 if (rdst) {
0770b53b 2776 dst = &rdst->remote_ip;
2777 if (vxlan_addr_any(dst)) {
2778 if (did_rsc) {
2779 /* short-circuited back to local bridge */
fc68c995
SB		 2780 vxlan_encap_bypass(skb, vxlan, vxlan,
2781 default_vni, true);
0770b53b 2782 return;
2783 }
2784 goto drop;
2785 }
2786
0dfbdf41 2787 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
3ad7a4b1 2788 vni = (rdst->remote_vni) ? : default_vni;
49f810f0 2789 ifindex = rdst->remote_ifindex;
1158632b 2790 local_ip = vxlan->cfg.saddr;
d71785ff 2791 dst_cache = &rdst->dst_cache;
0770b53b 2792 md->gbp = skb->mark;
72f6d71e
HL		 2793 if (flags & VXLAN_F_TTL_INHERIT) {
2794 ttl = ip_tunnel_get_ttl(old_iph, skb);
2795 } else {
2796 ttl = vxlan->cfg.ttl;
2797 if (!ttl && vxlan_addr_multicast(dst))
2798 ttl = 1;
2799 }
0770b53b 2800
2801 tos = vxlan->cfg.tos;
2802 if (tos == 1)
2803 tos = ip_tunnel_get_dsfield(old_iph, skb);
2804
2805 if (dst->sa.sa_family == AF_INET)
2806 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2807 else
2808 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2809 label = vxlan->cfg.label;
ee122c79 2810 } else {
435be28b
JK		 2811 if (!info) {
2812 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2813 dev->name);
2814 goto drop;
2815 }
b1be00a6 2816 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
272d96a5 2817 if (remote_ip.sa.sa_family == AF_INET) {
a725e514 2818 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
272d96a5 2819 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2820 } else {
a725e514 2821 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
272d96a5 2822 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2823 }
ee122c79 2824 dst = &remote_ip;
0770b53b 2825 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2826 vni = tunnel_id_to_key32(info->key.tun_id);
49f810f0 2827 ifindex = 0;
d71785ff 2828 dst_cache = &info->dst_cache;
eadf52cf
XL		 2829 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2830 if (info->options_len < sizeof(*md))
2831 goto drop;
0770b53b 2832 md = ip_tunnel_info_opts(info);
eadf52cf 2833 }
a725e514
JB		 2834 ttl = info->key.ttl;
2835 tos = info->key.tos;
e7f70af1 2836 label = info->key.label;
b4ed5cad 2837 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514 2838 }
0770b53b 2839 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2840 vxlan->cfg.port_max, true);
a725e514 2841
56de859e 2842 rcu_read_lock();
e4c7ed41 2843 if (dst->sa.sa_family == AF_INET) {
c6fcc4fc 2844 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
c46b7897 2845 struct rtable *rt;
0770b53b 2846 __be16 df = 0;
c6fcc4fc 2847
aab8cc36
AB		 2848 if (!ifindex)
2849 ifindex = sock4->sock->sk->sk_bound_dev_if;
2850
49f810f0 2851 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
272d96a5 2852 dst->sin.sin_addr.s_addr,
1158632b 2853 &local_ip.sin.sin_addr.s_addr,
4ecb1d83 2854 dst_port, src_port,
d71785ff 2855 dst_cache, info);
8ebd115b
DM		 2856 if (IS_ERR(rt)) {
2857 err = PTR_ERR(rt);
c46b7897 2858 goto tx_error;
8ebd115b 2859 }
e4c7ed41 2860
4f94c353
JV		 2861 if (fan_has_map(&vxlan->fan) && rt->rt_flags & RTCF_LOCAL) {
2862 netdev_dbg(dev, "discard fan to localhost %pI4\n",
2863 &dst->sin.sin_addr.s_addr);
2864 ip_rt_put(rt);
2865 goto tx_free;
2866 }
2867
fee1fad7 2868 if (!info) {
b4d30697 2869 /* Bypass encapsulation if the destination is local */
fee1fad7 2870 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2871 dst_port, ifindex, vni,
2872 &rt->dst, rt->rt_flags);
fee1fad7 2873 if (err)
56de859e 2874 goto out_unlock;
b4d30697
SB		 2875
2876 if (vxlan->cfg.df == VXLAN_DF_SET) {
2877 df = htons(IP_DF);
2878 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2879 struct ethhdr *eth = eth_hdr(skb);
2880
2881 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2882 (ntohs(eth->h_proto) == ETH_P_IP &&
2883 old_iph->frag_off & htons(IP_DF)))
2884 df = htons(IP_DF);
2885 }
fee1fad7 2886 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
6ceb31ca 2887 df = htons(IP_DF);
fee1fad7 2888 }
6ceb31ca 2889
c46b7897 2890 ndst = &rt->dst;
fc68c995
SB		 2891 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM,
2892 netif_is_any_bridge_port(dev));
2893 if (err < 0) {
2894 goto tx_error;
2895 } else if (err) {
2896 if (info) {
8a734168 2897 struct ip_tunnel_info *unclone;
fc68c995
SB		 2898 struct in_addr src, dst;
2899
8a734168
AT		 2900 unclone = skb_tunnel_info_unclone(skb);
2901 if (unlikely(!unclone))
2902 goto tx_error;
2903
fc68c995
SB		 2904 src = remote_ip.sin.sin_addr;
2905 dst = local_ip.sin.sin_addr;
8a734168
AT		 2906 unclone->key.u.ipv4.src = src.s_addr;
2907 unclone->key.u.ipv4.dst = dst.s_addr;
fc68c995
SB		 2908 }
2909 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2910 dst_release(ndst);
2911 goto out_unlock;
2912 }
a93bf0ff 2913
a0dced17 2914 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2915 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
c46b7897 2916 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
54bfd872 2917 vni, md, flags, udp_sum);
f491e56d 2918 if (err < 0)
c46b7897 2919 goto tx_error;
f491e56d 2920
1158632b 2921 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
f491e56d
JB		 2922 dst->sin.sin_addr.s_addr, tos, ttl, df,
2923 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2924#if IS_ENABLED(CONFIG_IPV6)
2925 } else {
c6fcc4fc 2926 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
e4c7ed41 2927
aab8cc36
AB		 2928 if (!ifindex)
2929 ifindex = sock6->sock->sk->sk_bound_dev_if;
2930
49f810f0 2931 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
272d96a5 2932 label, &dst->sin6.sin6_addr,
1158632b 2933 &local_ip.sin6.sin6_addr,
4ecb1d83 2934 dst_port, src_port,
db3c6139 2935 dst_cache, info);
e5d4b29f 2936 if (IS_ERR(ndst)) {
8ebd115b 2937 err = PTR_ERR(ndst);
c46b7897 2938 ndst = NULL;
9dcc71e1 2939 goto tx_error;
e4c7ed41 2940 }
655c3de1 2941
fee1fad7 2942 if (!info) {
2943 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
49560532 2944
fee1fad7 2945 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2946 dst_port, ifindex, vni,
2947 ndst, rt6i_flags);
fee1fad7 2948 if (err)
56de859e 2949 goto out_unlock;
fee1fad7 2950 }
35e2d115 2951
fc68c995
SB		 2952 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM,
2953 netif_is_any_bridge_port(dev));
2954 if (err < 0) {
2955 goto tx_error;
2956 } else if (err) {
2957 if (info) {
8a734168 2958 struct ip_tunnel_info *unclone;
fc68c995
SB		 2959 struct in6_addr src, dst;
2960
8a734168
AT		 2961 unclone = skb_tunnel_info_unclone(skb);
2962 if (unlikely(!unclone))
2963 goto tx_error;
2964
fc68c995
SB		 2965 src = remote_ip.sin6.sin6_addr;
2966 dst = local_ip.sin6.sin6_addr;
8a734168
AT		 2967 unclone->key.u.ipv6.src = src;
2968 unclone->key.u.ipv6.dst = dst;
fc68c995
SB		 2969 }
2970
2971 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2972 dst_release(ndst);
2973 goto out_unlock;
2974 }
a93bf0ff 2975
a0dced17 2976 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2977 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2978 skb_scrub_packet(skb, xnet);
2979 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2980 vni, md, flags, udp_sum);
c46b7897 2981 if (err < 0)
2982 goto tx_error;
2983
0770b53b 2984 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
1158632b 2985 &local_ip.sin6.sin6_addr,
272d96a5 2986 &dst->sin6.sin6_addr, tos, ttl,
e7f70af1 2987 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2988#endif
2989 }
56de859e
JK		 2990out_unlock:
2991 rcu_read_unlock();
4ad16930 2992 return;
d342894c 2993
2994drop:
2995 dev->stats.tx_dropped++;
c46b7897 2996 dev_kfree_skb(skb);
2997 return;
d342894c 2998
2999tx_error:
56de859e 3000 rcu_read_unlock();
655c3de1 3001 if (err == -ELOOP)
3002 dev->stats.collisions++;
3003 else if (err == -ENETUNREACH)
3004 dev->stats.tx_carrier_errors++;
c46b7897 3005 dst_release(ndst);
d342894c 3006 dev->stats.tx_errors++;
4f94c353 3007tx_free:
c46b7897 3008 kfree_skb(skb);
d342894c 3009}
3010
1274e1cc
RP		 3011static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
3012 struct vxlan_fdb *f, __be32 vni, bool did_rsc)
3013{
3014 struct vxlan_rdst nh_rdst;
3015 struct nexthop *nh;
3016 bool do_xmit;
3017 u32 hash;
3018
3019 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
3020 hash = skb_get_hash(skb);
3021
3022 rcu_read_lock();
3023 nh = rcu_dereference(f->nh);
3024 if (!nh) {
3025 rcu_read_unlock();
3026 goto drop;
3027 }
3028 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
3029 rcu_read_unlock();
3030
3031 if (likely(do_xmit))
3032 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
3033 else
3034 goto drop;
3035
3036 return;
3037
3038drop:
3039 dev->stats.tx_dropped++;
3040 dev_kfree_skb(skb);
3041}
3042
6681712d
DS		 3043/* Transmit local packets over Vxlan
3044 *
3045 * Outer IP header inherits ECN and DF from inner header.
3046 * Outer UDP destination is the VXLAN assigned port.
3047 * source port is based on hash of flow
3048 */
3049static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
3050{
3051 struct vxlan_dev *vxlan = netdev_priv(dev);
8bff3685 3052 struct vxlan_rdst *rdst, *fdst = NULL;
3093fbe7 3053 const struct ip_tunnel_info *info;
6681712d 3054 bool did_rsc = false;
6681712d 3055 struct vxlan_fdb *f;
8bff3685 3056 struct ethhdr *eth;
3ad7a4b1 3057 __be32 vni = 0;
6681712d 3058
61adedf3 3059 info = skb_tunnel_info(skb);
3093fbe7 3060
6681712d 3061 skb_reset_mac_header(skb);
6681712d 3062
dc5321d7 3063 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 3064 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
3065 info->mode & IP_TUNNEL_INFO_TX) {
3066 vni = tunnel_id_to_key32(info->key.tun_id);
3067 } else {
3068 if (info && info->mode & IP_TUNNEL_INFO_TX)
3069 vxlan_xmit_one(skb, dev, vni, NULL, false);
3070 else
3071 kfree_skb(skb);
3072 return NETDEV_TX_OK;
3073 }
47e5d1b0
JB		 3074 }
3075
dc5321d7 3076 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
47e5d1b0 3077 eth = eth_hdr(skb);
f564f45c 3078 if (ntohs(eth->h_proto) == ETH_P_ARP)
3ad7a4b1 3079 return arp_reduce(dev, skb, vni);
f564f45c 3080#if IS_ENABLED(CONFIG_IPV6)
8bff3685
XL		 3081 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
3082 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
3083 sizeof(struct nd_msg)) &&
3084 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
3085 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
3086
3087 if (m->icmph.icmp6_code == 0 &&
3088 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
f1fb08f6 3089 return neigh_reduce(dev, skb, vni);
f564f45c
CW		 3090 }
3091#endif
3092 }
6681712d 3093
4f94c353
JV		 3094 if (fan_has_map(&vxlan->fan)) {
3095 struct vxlan_rdst fan_rdst;
3096
3097 netdev_dbg(vxlan->dev, "vxlan_xmit p %x d %pM\n",
3098 eth->h_proto, eth->h_dest);
3099 if (vxlan_fan_build_rdst(vxlan, skb, &fan_rdst)) {
3100 dev->stats.tx_dropped++;
3101 kfree_skb(skb);
3102 return NETDEV_TX_OK;
3103 }
3104 vxlan_xmit_one(skb, dev, vni, &fan_rdst, 0);
3105 return NETDEV_TX_OK;
3106 }
3107
47e5d1b0 3108 eth = eth_hdr(skb);
3ad7a4b1 3109 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 3110 did_rsc = false;
3111
dc5321d7 3112 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
e15a00aa
CW		 3113 (ntohs(eth->h_proto) == ETH_P_IP ||
3114 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 3115 did_rsc = route_shortcircuit(dev, skb);
3116 if (did_rsc)
3ad7a4b1 3117 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 3118 }
3119
6681712d 3120 if (f == NULL) {
3ad7a4b1 3121 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 3122 if (f == NULL) {
dc5321d7 3123 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
afbd8bae
MR		 3124 !is_multicast_ether_addr(eth->h_dest))
3125 vxlan_fdb_miss(vxlan, eth->h_dest);
3126
3127 dev->stats.tx_dropped++;
8f646c92 3128 kfree_skb(skb);
afbd8bae
MR		 3129 return NETDEV_TX_OK;
3130 }
3131 }
6681712d 3132
1274e1cc
RP		 3133 if (rcu_access_pointer(f->nh)) {
3134 vxlan_xmit_nh(skb, dev, f,
3135 (vni ? : vxlan->default_dst.remote_vni), did_rsc);
3136 } else {
3137 list_for_each_entry_rcu(rdst, &f->remotes, list) {
3138 struct sk_buff *skb1;
6681712d 3139
1274e1cc
RP		 3140 if (!fdst) {
3141 fdst = rdst;
3142 continue;
3143 }
3144 skb1 = skb_clone(skb, GFP_ATOMIC);
3145 if (skb1)
3146 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
8f646c92 3147 }
1274e1cc
RP		 3148 if (fdst)
3149 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
3150 else
3151 kfree_skb(skb);
6681712d
DS		 3152 }
3153
4ad16930 3154 return NETDEV_TX_OK;
6681712d
DS		 3155}
3156
d342894c 3157/* Walk the forwarding table and purge stale entries */
df7e828c 3158static void vxlan_cleanup(struct timer_list *t)
d342894c 3159{
df7e828c 3160 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
d342894c 3161 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
3162 unsigned int h;
3163
3164 if (!netif_running(vxlan->dev))
3165 return;
3166
d342894c 3167 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3168 struct hlist_node *p, *n;
14e1d0fa 3169
fe1e0713 3170 spin_lock(&vxlan->hash_lock[h]);
d342894c 3171 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3172 struct vxlan_fdb *f
3173 = container_of(p, struct vxlan_fdb, hlist);
3174 unsigned long timeout;
3175
efb5f68f 3176 if (f->state & (NUD_PERMANENT | NUD_NOARP))
d342894c 3177 continue;
3178
def499c9
RP		 3179 if (f->flags & NTF_EXT_LEARNED)
3180 continue;
3181
0dfbdf41 3182 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 3183 if (time_before_eq(timeout, jiffies)) {
3184 netdev_dbg(vxlan->dev,
3185 "garbage collect %pM\n",
3186 f->eth_addr);
3187 f->state = NUD_STALE;
0e6160f3 3188 vxlan_fdb_destroy(vxlan, f, true, true);
d342894c 3189 } else if (time_before(timeout, next_timer))
3190 next_timer = timeout;
3191 }
fe1e0713 3192 spin_unlock(&vxlan->hash_lock[h]);
d342894c 3193 }
d342894c 3194
3195 mod_timer(&vxlan->age_timer, next_timer);
3196}
3197
a53cb29b
MB		 3198static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
3199{
3200 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3201
3202 spin_lock(&vn->sock_lock);
69e76661
JB		 3203 hlist_del_init_rcu(&vxlan->hlist4.hlist);
3204#if IS_ENABLED(CONFIG_IPV6)
3205 hlist_del_init_rcu(&vxlan->hlist6.hlist);
3206#endif
a53cb29b
MB		 3207 spin_unlock(&vn->sock_lock);
3208}
3209
69e76661
JB		 3210static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
3211 struct vxlan_dev_node *node)
9c2e24e1 3212{
56ef9c90 3213 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 3214 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 3215
69e76661 3216 node->vxlan = vxlan;
56ef9c90 3217 spin_lock(&vn->sock_lock);
69e76661 3218 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
56ef9c90 3219 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 3220}
3221
d342894c 3222/* Setup stats when device is created */
3223static int vxlan_init(struct net_device *dev)
3224{
384d91c2
TY		 3225 struct vxlan_dev *vxlan = netdev_priv(dev);
3226 int err;
3227
1c213bd2 3228 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 3229 if (!dev->tstats)
d342894c 3230 return -ENOMEM;
3231
384d91c2
TY		 3232 err = gro_cells_init(&vxlan->gro_cells, dev);
3233 if (err) {
3234 free_percpu(dev->tstats);
3235 return err;
3236 }
3237
d342894c 3238 return 0;
3239}
3240
3ad7a4b1 3241static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
afbd8bae
MR		 3242{
3243 struct vxlan_fdb *f;
fe1e0713 3244 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
afbd8bae 3245
fe1e0713 3246 spin_lock_bh(&vxlan->hash_lock[hash_index]);
3ad7a4b1 3247 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 3248 if (f)
0e6160f3 3249 vxlan_fdb_destroy(vxlan, f, true, true);
fe1e0713 3250 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
afbd8bae
MR		 3251}
3252
ebf4063e
SH		 3253static void vxlan_uninit(struct net_device *dev)
3254{
3255 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 3256
ad6c9986
SB		 3257 gro_cells_destroy(&vxlan->gro_cells);
3258
3ad7a4b1 3259 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
afbd8bae 3260
ebf4063e
SH		 3261 free_percpu(dev->tstats);
3262}
3263
d342894c 3264/* Start ageing timer and join group when device is brought up */
3265static int vxlan_open(struct net_device *dev)
3266{
3267 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 3268 int ret;
56ef9c90 3269
205f356d
JB		 3270 ret = vxlan_sock_add(vxlan);
3271 if (ret < 0)
3272 return ret;
d342894c 3273
79d4a94f 3274 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 3275 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 3276 if (ret == -EADDRINUSE)
3277 ret = 0;
56ef9c90 3278 if (ret) {
205f356d 3279 vxlan_sock_release(vxlan);
56ef9c90
MRL		 3280 return ret;
3281 }
d342894c 3282 }
3283
0dfbdf41 3284 if (vxlan->cfg.age_interval)
d342894c 3285 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
3286
56ef9c90 3287 return ret;
d342894c 3288}
3289
3290/* Purge the forwarding table */
8b3f9337 3291static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
d342894c 3292{
31fec5aa 3293 unsigned int h;
d342894c 3294
d342894c 3295 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3296 struct hlist_node *p, *n;
fe1e0713
L		 3297
3298 spin_lock_bh(&vxlan->hash_lock[h]);
d342894c 3299 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3300 struct vxlan_fdb *f
3301 = container_of(p, struct vxlan_fdb, hlist);
8b3f9337
RP		 3302 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
3303 continue;
afbd8bae 3304 /* the all_zeros_mac entry is deleted at vxlan_uninit */
fda2ec62
TY		 3305 if (is_zero_ether_addr(f->eth_addr) &&
3306 f->vni == vxlan->cfg.vni)
3307 continue;
3308 vxlan_fdb_destroy(vxlan, f, true, true);
d342894c 3309 }
fe1e0713 3310 spin_unlock_bh(&vxlan->hash_lock[h]);
d342894c 3311 }
d342894c 3312}
3313
3314/* Cleanup timer and forwarding table on shutdown */
3315static int vxlan_stop(struct net_device *dev)
3316{
3317 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 3318 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 3319 int ret = 0;
d342894c 3320
24c0e683 3321 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 3322 !vxlan_group_used(vn, vxlan))
56ef9c90 3323 ret = vxlan_igmp_leave(vxlan);
d342894c 3324
3325 del_timer_sync(&vxlan->age_timer);
3326
8b3f9337 3327 vxlan_flush(vxlan, false);
205f356d 3328 vxlan_sock_release(vxlan);
d342894c 3329
56ef9c90 3330 return ret;
d342894c 3331}
3332
d342894c 3333/* Stub, nothing needs to be done. */
3334static void vxlan_set_multicast_list(struct net_device *dev)
3335{
3336}
3337
91572088 3338static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
345010b5 3339{
91572088
JW		 3340 struct vxlan_dev *vxlan = netdev_priv(dev);
3341 struct vxlan_rdst *dst = &vxlan->default_dst;
3342 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3343 dst->remote_ifindex);
ce44a4ae 3344 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
345010b5 3345
91572088
JW		 3346 /* This check is different than dev->max_mtu, because it looks at
3347 * the lowerdev->mtu, rather than the static dev->max_mtu
3348 */
3349 if (lowerdev) {
3350 int max_mtu = lowerdev->mtu -
3351 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
3352 if (new_mtu > max_mtu)
72564b59 3353 return -EINVAL;
72564b59
DW		 3354 }
3355
345010b5
DB		 3356 dev->mtu = new_mtu;
3357 return 0;
3358}
3359
fc4099f1
PS		 3360static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3361{
3362 struct vxlan_dev *vxlan = netdev_priv(dev);
3363 struct ip_tunnel_info *info = skb_tunnel_info(skb);
3364 __be16 sport, dport;
3365
3366 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3367 vxlan->cfg.port_max, true);
3368 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3369
239e944f 3370 if (ip_tunnel_info_af(info) == AF_INET) {
c6fcc4fc 3371 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
1a8496ba
JB		 3372 struct rtable *rt;
3373
655c3de1 3374 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
1a8496ba 3375 info->key.u.ipv4.dst,
22f0708a
PA		 3376 &info->key.u.ipv4.src, dport, sport,
3377 &info->dst_cache, info);
1a8496ba
JB		 3378 if (IS_ERR(rt))
3379 return PTR_ERR(rt);
3380 ip_rt_put(rt);
239e944f
JB		 3381 } else {
3382#if IS_ENABLED(CONFIG_IPV6)
03dc52a8 3383 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
239e944f
JB		 3384 struct dst_entry *ndst;
3385
655c3de1 3386 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
e7f70af1 3387 info->key.label, &info->key.u.ipv6.dst,
22f0708a
PA		 3388 &info->key.u.ipv6.src, dport, sport,
3389 &info->dst_cache, info);
239e944f
JB		 3390 if (IS_ERR(ndst))
3391 return PTR_ERR(ndst);
3392 dst_release(ndst);
239e944f
JB		 3393#else /* !CONFIG_IPV6 */
3394 return -EPFNOSUPPORT;
3395#endif
3396 }
1a8496ba
JB		 3397 info->key.tp_src = sport;
3398 info->key.tp_dst = dport;
239e944f 3399 return 0;
fc4099f1
PS		 3400}
3401
0c867c9b 3402static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 3403 .ndo_init = vxlan_init,
ebf4063e 3404 .ndo_uninit = vxlan_uninit,
d342894c 3405 .ndo_open = vxlan_open,
3406 .ndo_stop = vxlan_stop,
3407 .ndo_start_xmit = vxlan_xmit,
b220a4a7 3408 .ndo_get_stats64 = dev_get_tstats64,
d342894c 3409 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 3410 .ndo_change_mtu = vxlan_change_mtu,
d342894c 3411 .ndo_validate_addr = eth_validate_addr,
3412 .ndo_set_mac_address = eth_mac_addr,
3413 .ndo_fdb_add = vxlan_fdb_add,
3414 .ndo_fdb_del = vxlan_fdb_delete,
3415 .ndo_fdb_dump = vxlan_fdb_dump,
474c3c89 3416 .ndo_fdb_get = vxlan_fdb_get,
fc4099f1 3417 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
8f1af75d 3418 .ndo_change_proto_down = dev_change_proto_down_generic,
d342894c 3419};
3420
e1e5314d
JB		 3421static const struct net_device_ops vxlan_netdev_raw_ops = {
3422 .ndo_init = vxlan_init,
3423 .ndo_uninit = vxlan_uninit,
3424 .ndo_open = vxlan_open,
3425 .ndo_stop = vxlan_stop,
3426 .ndo_start_xmit = vxlan_xmit,
b220a4a7 3427 .ndo_get_stats64 = dev_get_tstats64,
e1e5314d
JB		 3428 .ndo_change_mtu = vxlan_change_mtu,
3429 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3430};
3431
d342894c 3432/* Info for udev, that this is a virtual tunnel endpoint */
3433static struct device_type vxlan_type = {
3434 .name = "vxlan",
3435};
3436
e5de25dc 3437/* Calls the ndo_udp_tunnel_add of the caller in order to
35e42379 3438 * supply the listening VXLAN udp ports. Callers are expected
e5de25dc 3439 * to implement the ndo_udp_tunnel_add.
53cf5275 3440 */
2d2b13fc 3441static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
53cf5275
JG		 3442{
3443 struct vxlan_sock *vs;
3444 struct net *net = dev_net(dev);
3445 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
35e42379 3446 unsigned int i;
53cf5275
JG		 3447
3448 spin_lock(&vn->sock_lock);
3449 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2d2b13fc
SD		 3450 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3451 unsigned short type;
3452
3453 if (vs->flags & VXLAN_F_GPE)
3454 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3455 else
3456 type = UDP_TUNNEL_TYPE_VXLAN;
3457
3458 if (push)
3459 udp_tunnel_push_rx_port(dev, vs->sock, type);
3460 else
3461 udp_tunnel_drop_rx_port(dev, vs->sock, type);
3462 }
53cf5275
JG		 3463 }
3464 spin_unlock(&vn->sock_lock);
3465}
53cf5275 3466
d342894c 3467/* Initialize the device structure. */
3468static void vxlan_setup(struct net_device *dev)
3469{
3470 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 3471 unsigned int h;
d342894c 3472
65226ef8
JB		 3473 eth_hw_addr_random(dev);
3474 ether_setup(dev);
3475
cf124db5 3476 dev->needs_free_netdev = true;
d342894c 3477 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3478
d342894c 3479 dev->features |= NETIF_F_LLTX;
d6727fe3 3480 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 3481 dev->features |= NETIF_F_RXCSUM;
05c0db08 3482 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 3483
1eaa8178 3484 dev->vlan_features = dev->features;
0afb1666 3485 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 3486 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
02875878 3487 netif_keep_dst(dev);
0c867c9b 3488 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 3489
a985343b
MS		 3490 /* MTU range: 68 - 65535 */
3491 dev->min_mtu = ETH_MIN_MTU;
3492 dev->max_mtu = ETH_MAX_MTU;
3493
553675fb 3494 INIT_LIST_HEAD(&vxlan->next);
d342894c 3495
df7e828c 3496 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
d342894c 3497
3498 vxlan->dev = dev;
3499
fe1e0713
L		 3500 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3501 spin_lock_init(&vxlan->hash_lock[h]);
d342894c 3502 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
fe1e0713 3503 }
4f94c353
JV		 3504
3505 INIT_LIST_HEAD(&vxlan->fan.fan_maps);
d342894c 3506}
3507
0c867c9b
JB		 3508static void vxlan_ether_setup(struct net_device *dev)
3509{
0c867c9b
JB		 3510 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3511 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3512 dev->netdev_ops = &vxlan_netdev_ether_ops;
3513}
3514
e1e5314d
JB		 3515static void vxlan_raw_setup(struct net_device *dev)
3516{
65226ef8 3517 dev->header_ops = NULL;
e1e5314d
JB		 3518 dev->type = ARPHRD_NONE;
3519 dev->hard_header_len = 0;
3520 dev->addr_len = 0;
e1e5314d
JB		 3521 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3522 dev->netdev_ops = &vxlan_netdev_raw_ops;
3523}
3524
d342894c 3525static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3526 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
c593642c 3527 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) },
e4c7ed41 3528 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 3529 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
c593642c 3530 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
e4c7ed41 3531 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 3532 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
3533 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 3534 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 3535 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
3536 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
3537 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 3538 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 3539 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
3540 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
3541 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
3542 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 3543 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 3544 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 3545 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
3546 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
3547 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 3548 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3549 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 3550 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 3551 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 3552 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
72f6d71e 3553 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
b4d30697 3554 [IFLA_VXLAN_DF] = { .type = NLA_U8 },
d342894c 3555};
3556
a8b8a889
MS		 3557static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3558 struct netlink_ext_ack *extack)
d342894c 3559{
3560 if (tb[IFLA_ADDRESS]) {
3561 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
653ef6a3
GM		 3562 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3563 "Provided link layer address is not Ethernet");
d342894c 3564 return -EINVAL;
3565 }
3566
3567 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
653ef6a3
GM		 3568 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3569 "Provided Ethernet address is not unicast");
d342894c 3570 return -EADDRNOTAVAIL;
3571 }
3572 }
3573
a985343b 3574 if (tb[IFLA_MTU]) {
019b13ae 3575 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
a985343b 3576
653ef6a3
GM		 3577 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3578 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3579 "MTU must be between 68 and 65535");
a985343b 3580 return -EINVAL;
653ef6a3 3581 }
a985343b
MS		 3582 }
3583
653ef6a3
GM		 3584 if (!data) {
3585 NL_SET_ERR_MSG(extack,
3586 "Required attributes not provided to perform the operation");
d342894c 3587 return -EINVAL;
653ef6a3 3588 }
d342894c 3589
3590 if (data[IFLA_VXLAN_ID]) {
a985343b
MS		 3591 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3592
653ef6a3 3593 if (id >= VXLAN_N_VID) {
cc8e7c69 3594 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
653ef6a3 3595 "VXLAN ID must be lower than 16777216");
d342894c 3596 return -ERANGE;
653ef6a3 3597 }
d342894c 3598 }
3599
05f47d69 3600 if (data[IFLA_VXLAN_PORT_RANGE]) {
3601 const struct ifla_vxlan_port_range *p
3602 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3603
3604 if (ntohs(p->high) < ntohs(p->low)) {
cc8e7c69 3605 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
653ef6a3 3606 "Invalid source port range");
05f47d69 3607 return -EINVAL;
3608 }
3609 }
3610
b4d30697
SB		 3611 if (data[IFLA_VXLAN_DF]) {
3612 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3613
3614 if (df < 0 || df > VXLAN_DF_MAX) {
cc8e7c69 3615 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
b4d30697
SB		 3616 "Invalid DF attribute");
3617 return -EINVAL;
3618 }
3619 }
3620
d342894c 3621 return 0;
3622}
3623
1b13c97f
YB		 3624static void vxlan_get_drvinfo(struct net_device *netdev,
3625 struct ethtool_drvinfo *drvinfo)
3626{
3627 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3628 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3629}
3630
36fe3a61
MS		 3631static int vxlan_get_link_ksettings(struct net_device *dev,
3632 struct ethtool_link_ksettings *cmd)
3633{
3634 struct vxlan_dev *vxlan = netdev_priv(dev);
3635 struct vxlan_rdst *dst = &vxlan->default_dst;
3636 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3637 dst->remote_ifindex);
3638
3639 if (!lowerdev) {
3640 cmd->base.duplex = DUPLEX_UNKNOWN;
3641 cmd->base.port = PORT_OTHER;
3642 cmd->base.speed = SPEED_UNKNOWN;
3643
3644 return 0;
3645 }
3646
3647 return __ethtool_get_link_ksettings(lowerdev, cmd);
3648}
3649
1b13c97f 3650static const struct ethtool_ops vxlan_ethtool_ops = {
36fe3a61
MS		 3651 .get_drvinfo = vxlan_get_drvinfo,
3652 .get_link = ethtool_op_get_link,
3653 .get_link_ksettings = vxlan_get_link_ksettings,
1b13c97f
YB		 3654};
3655
3ee64f39 3656static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
aab8cc36 3657 __be16 port, u32 flags, int ifindex)
553675fb 3658{
e4c7ed41 3659 struct socket *sock;
3ee64f39
TH		 3660 struct udp_port_cfg udp_conf;
3661 int err;
e4c7ed41 3662
3ee64f39 3663 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 3664
3ee64f39
TH		 3665 if (ipv6) {
3666 udp_conf.family = AF_INET6;
3ee64f39 3667 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 3668 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 3669 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 3670 } else {
3671 udp_conf.family = AF_INET;
e4c7ed41 3672 }
e4c7ed41 3673
3ee64f39 3674 udp_conf.local_udp_port = port;
aab8cc36 3675 udp_conf.bind_ifindex = ifindex;
553675fb 3676
3ee64f39
TH		 3677 /* Open UDP socket */
3678 err = udp_sock_create(net, &udp_conf, &sock);
3679 if (err < 0)
3680 return ERR_PTR(err);
e4c7ed41 3681
39deb2c7 3682 return sock;
e4c7ed41
CW		 3683}
3684
3685/* Create new listen socket if needed */
b1be00a6 3686static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
aab8cc36
AB		 3687 __be16 port, u32 flags,
3688 int ifindex)
e4c7ed41
CW		 3689{
3690 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3691 struct vxlan_sock *vs;
3692 struct socket *sock;
e4c7ed41 3693 unsigned int h;
acbf74a7 3694 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 3695
dc01e7d3 3696 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 3697 if (!vs)
3698 return ERR_PTR(-ENOMEM);
3699
3700 for (h = 0; h < VNI_HASH_SIZE; ++h)
3701 INIT_HLIST_HEAD(&vs->vni_list[h]);
3702
aab8cc36 3703 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
39deb2c7 3704 if (IS_ERR(sock)) {
553675fb 3705 kfree(vs);
e50fddc8 3706 return ERR_CAST(sock);
553675fb 3707 }
e4c7ed41
CW		 3708
3709 vs->sock = sock;
66af846f 3710 refcount_set(&vs->refcnt, 1);
af33c1ad 3711 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 3712
9c2e24e1
PS		 3713 spin_lock(&vn->sock_lock);
3714 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
e7b3db5e 3715 udp_tunnel_notify_add_rx_port(sock,
b9adcd69
AD		 3716 (vs->flags & VXLAN_F_GPE) ?
3717 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 3718 UDP_TUNNEL_TYPE_VXLAN);
9c2e24e1 3719 spin_unlock(&vn->sock_lock);
553675fb 3720
3721 /* Mark socket as an encapsulation socket. */
5602c48c 3722 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ		 3723 tunnel_cfg.sk_user_data = vs;
3724 tunnel_cfg.encap_type = 1;
f2d1968e 3725 tunnel_cfg.encap_rcv = vxlan_rcv;
c3a43b9f 3726 tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
acbf74a7 3727 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH		 3728 tunnel_cfg.gro_receive = vxlan_gro_receive;
3729 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ		 3730
3731 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 3732
9c2e24e1
PS		 3733 return vs;
3734}
3735
b1be00a6 3736static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 3737{
205f356d
JB		 3738 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3739 struct vxlan_sock *vs = NULL;
69e76661 3740 struct vxlan_dev_node *node;
aab8cc36
AB		 3741 int l3mdev_index = 0;
3742
3743 if (vxlan->cfg.remote_ifindex)
3744 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3745 vxlan->net, vxlan->cfg.remote_ifindex);
9c2e24e1 3746
205f356d 3747 if (!vxlan->cfg.no_share) {
56ef9c90 3748 spin_lock(&vn->sock_lock);
205f356d 3749 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
aab8cc36
AB		 3750 vxlan->cfg.dst_port, vxlan->cfg.flags,
3751 l3mdev_index);
66af846f 3752 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
56ef9c90 3753 spin_unlock(&vn->sock_lock);
205f356d 3754 return -EBUSY;
56ef9c90
MRL		 3755 }
3756 spin_unlock(&vn->sock_lock);
3757 }
205f356d 3758 if (!vs)
b1be00a6 3759 vs = vxlan_socket_create(vxlan->net, ipv6,
aab8cc36
AB		 3760 vxlan->cfg.dst_port, vxlan->cfg.flags,
3761 l3mdev_index);
205f356d
JB		 3762 if (IS_ERR(vs))
3763 return PTR_ERR(vs);
b1be00a6 3764#if IS_ENABLED(CONFIG_IPV6)
69e76661 3765 if (ipv6) {
c6fcc4fc 3766 rcu_assign_pointer(vxlan->vn6_sock, vs);
69e76661
JB		 3767 node = &vxlan->hlist6;
3768 } else
b1be00a6 3769#endif
69e76661 3770 {
c6fcc4fc 3771 rcu_assign_pointer(vxlan->vn4_sock, vs);
69e76661
JB		 3772 node = &vxlan->hlist4;
3773 }
3774 vxlan_vs_add_dev(vs, vxlan, node);
205f356d 3775 return 0;
553675fb 3776}
3777
b1be00a6
JB		 3778static int vxlan_sock_add(struct vxlan_dev *vxlan)
3779{
dc5321d7
MS		 3780 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3781 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
d074bf96 3782 bool ipv4 = !ipv6 || metadata;
b1be00a6
JB		 3783 int ret = 0;
3784
c6fcc4fc 3785 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
b1be00a6 3786#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 3787 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
d074bf96 3788 if (ipv6) {
b1be00a6 3789 ret = __vxlan_sock_add(vxlan, true);
d074bf96
JB		 3790 if (ret < 0 && ret != -EAFNOSUPPORT)
3791 ipv4 = false;
3792 }
b1be00a6 3793#endif
d074bf96 3794 if (ipv4)
b1be00a6
JB		 3795 ret = __vxlan_sock_add(vxlan, false);
3796 if (ret < 0)
3797 vxlan_sock_release(vxlan);
3798 return ret;
3799}
3800
a985343b
MS		 3801static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3802 struct net_device **lower,
653ef6a3
GM		 3803 struct vxlan_dev *old,
3804 struct netlink_ext_ack *extack)
d342894c 3805{
33564bbb 3806 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
a985343b 3807 struct vxlan_dev *tmp;
e4c7ed41 3808 bool use_ipv6 = false;
d342894c 3809
a985343b
MS		 3810 if (conf->flags & VXLAN_F_GPE) {
3811 /* For now, allow GPE only together with
3812 * COLLECT_METADATA. This can be relaxed later; in such
3813 * case, the other side of the PtP link will have to be
3814 * provided.
3815 */
3816 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3817 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
653ef6a3
GM		 3818 NL_SET_ERR_MSG(extack,
3819 "VXLAN GPE does not support this combination of attributes");
a985343b 3820 return -EINVAL;
3555621d 3821 }
e1e5314d 3822 }
0c867c9b 3823
ce44a4ae
MS		 3824 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3825 /* Unless IPv6 is explicitly requested, assume IPv4 */
a985343b 3826 conf->remote_ip.sa.sa_family = AF_INET;
ce44a4ae
MS		 3827 conf->saddr.sa.sa_family = AF_INET;
3828 } else if (!conf->remote_ip.sa.sa_family) {
3829 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3830 } else if (!conf->saddr.sa.sa_family) {
3831 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3832 }
3833
653ef6a3
GM		 3834 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3835 NL_SET_ERR_MSG(extack,
3836 "Local and remote address must be from the same family");
ce44a4ae 3837 return -EINVAL;
653ef6a3 3838 }
e4c7ed41 3839
653ef6a3
GM		 3840 if (vxlan_addr_multicast(&conf->saddr)) {
3841 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
0f22a3c6 3842 return -EINVAL;
653ef6a3 3843 }
0f22a3c6 3844
ce44a4ae 3845 if (conf->saddr.sa.sa_family == AF_INET6) {
653ef6a3
GM		 3846 if (!IS_ENABLED(CONFIG_IPV6)) {
3847 NL_SET_ERR_MSG(extack,
3848 "IPv6 support not enabled in the kernel");
057ba29b 3849 return -EPFNOSUPPORT;
653ef6a3 3850 }
e4c7ed41 3851 use_ipv6 = true;
a985343b 3852 conf->flags |= VXLAN_F_IPV6;
0f22a3c6
MS		 3853
3854 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3855 int local_type =
3856 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3857 int remote_type =
3858 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3859
3860 if (local_type & IPV6_ADDR_LINKLOCAL) {
3861 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
653ef6a3
GM		 3862 (remote_type != IPV6_ADDR_ANY)) {
3863 NL_SET_ERR_MSG(extack,
3864 "Invalid combination of local and remote address scopes");
0f22a3c6 3865 return -EINVAL;
653ef6a3 3866 }
0f22a3c6
MS		 3867
3868 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3869 } else {
3870 if (remote_type ==
653ef6a3
GM		 3871 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3872 NL_SET_ERR_MSG(extack,
3873 "Invalid combination of local and remote address scopes");
0f22a3c6 3874 return -EINVAL;
653ef6a3 3875 }
0f22a3c6
MS		 3876
3877 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3878 }
3879 }
057ba29b 3880 }
d342894c 3881
653ef6a3
GM		 3882 if (conf->label && !use_ipv6) {
3883 NL_SET_ERR_MSG(extack,
3884 "Label attribute only applies to IPv6 VXLAN devices");
e7f70af1 3885 return -EINVAL;
653ef6a3 3886 }
e7f70af1 3887
a985343b
MS		 3888 if (conf->remote_ifindex) {
3889 struct net_device *lowerdev;
34e02aa1 3890
a985343b 3891 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
653ef6a3
GM		 3892 if (!lowerdev) {
3893 NL_SET_ERR_MSG(extack,
3894 "Invalid local interface, device not found");
34e02aa1 3895 return -ENODEV;
653ef6a3 3896 }
d342894c 3897
e4c7ed41
CW		 3898#if IS_ENABLED(CONFIG_IPV6)
3899 if (use_ipv6) {
3900 struct inet6_dev *idev = __in6_dev_get(lowerdev);
653ef6a3
GM		 3901 if (idev && idev->cnf.disable_ipv6) {
3902 NL_SET_ERR_MSG(extack,
3903 "IPv6 support disabled by administrator");
e4c7ed41 3904 return -EPERM;
653ef6a3 3905 }
e4c7ed41
CW		 3906 }
3907#endif
3908
a985343b
MS		 3909 *lower = lowerdev;
3910 } else {
653ef6a3
GM		 3911 if (vxlan_addr_multicast(&conf->remote_ip)) {
3912 NL_SET_ERR_MSG(extack,
3913 "Local interface required for multicast remote destination");
3914
a985343b 3915 return -EINVAL;
653ef6a3 3916 }
1ba56fb4 3917
0f22a3c6 3918#if IS_ENABLED(CONFIG_IPV6)
653ef6a3
GM		 3919 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3920 NL_SET_ERR_MSG(extack,
3921 "Local interface required for link-local local/remote addresses");
0f22a3c6 3922 return -EINVAL;
653ef6a3 3923 }
0f22a3c6
MS		 3924#endif
3925
a985343b 3926 *lower = NULL;
9dc2ad10 3927 }
d342894c 3928
a985343b
MS		 3929 if (!conf->dst_port) {
3930 if (conf->flags & VXLAN_F_GPE)
3931 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3932 else
3933 conf->dst_port = htons(vxlan_port);
d6acfeb1
FM		 3934 }
3935
a985343b
MS		 3936 if (!conf->age_interval)
3937 conf->age_interval = FDB_AGE_DEFAULT;
91572088 3938
a985343b
MS		 3939 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3940 if (tmp == old)
3941 continue;
91572088 3942
49f810f0
MS		 3943 if (tmp->cfg.vni != conf->vni)
3944 continue;
3945 if (tmp->cfg.dst_port != conf->dst_port)
3946 continue;
3947 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
ce44a4ae 3948 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
49f810f0
MS		 3949 continue;
3950
3951 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3952 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3953 continue;
3954
653ef6a3
GM		 3955 NL_SET_ERR_MSG(extack,
3956 "A VXLAN device with the specified VNI already exists");
49f810f0 3957 return -EEXIST;
a985343b 3958 }
91572088 3959
a985343b
MS		 3960 return 0;
3961}
3962
3963static void vxlan_config_apply(struct net_device *dev,
3964 struct vxlan_config *conf,
889ce937
SD		 3965 struct net_device *lowerdev,
3966 struct net *src_net,
3967 bool changelink)
a985343b
MS		 3968{
3969 struct vxlan_dev *vxlan = netdev_priv(dev);
3970 struct vxlan_rdst *dst = &vxlan->default_dst;
3971 unsigned short needed_headroom = ETH_HLEN;
3972 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3973 int max_mtu = ETH_MAX_MTU;
3974
3975 if (!changelink) {
3976 if (conf->flags & VXLAN_F_GPE)
3977 vxlan_raw_setup(dev);
3978 else
3979 vxlan_ether_setup(dev);
3980
3981 if (conf->mtu)
3982 dev->mtu = conf->mtu;
889ce937
SD		 3983
3984 vxlan->net = src_net;
a985343b
MS		 3985 }
3986
3987 dst->remote_vni = conf->vni;
3988
3989 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3990
3991 if (lowerdev) {
3992 dst->remote_ifindex = conf->remote_ifindex;
3993
3994 dev->gso_max_size = lowerdev->gso_max_size;
3995 dev->gso_max_segs = lowerdev->gso_max_segs;
91572088 3996
a985343b 3997 needed_headroom = lowerdev->hard_header_len;
0a35dc41 3998 needed_headroom += lowerdev->needed_headroom;
91572088 3999
a5e74021 4000 dev->needed_tailroom = lowerdev->needed_tailroom;
91572088 4001
a985343b
MS		 4002 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
4003 VXLAN_HEADROOM);
f870c1ff
AK		 4004 if (max_mtu < ETH_MIN_MTU)
4005 max_mtu = ETH_MIN_MTU;
4006
4007 if (!changelink && !conf->mtu)
4008 dev->mtu = max_mtu;
7e059158
DW		 4009 }
4010
a985343b
MS		 4011 if (dev->mtu > max_mtu)
4012 dev->mtu = max_mtu;
4013
b1be00a6
JB		 4014 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
4015 needed_headroom += VXLAN6_HEADROOM;
4016 else
4017 needed_headroom += VXLAN_HEADROOM;
4018 dev->needed_headroom = needed_headroom;
4019
0dfbdf41 4020 memcpy(&vxlan->cfg, conf, sizeof(*conf));
a985343b 4021}
0dfbdf41 4022
a985343b 4023static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
653ef6a3
GM		 4024 struct vxlan_config *conf, bool changelink,
4025 struct netlink_ext_ack *extack)
a985343b
MS		 4026{
4027 struct vxlan_dev *vxlan = netdev_priv(dev);
4028 struct net_device *lowerdev;
4029 int ret;
0dfbdf41 4030
653ef6a3 4031 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
a985343b
MS		 4032 if (ret)
4033 return ret;
8bcdc4f3 4034
889ce937 4035 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
0dfbdf41 4036
0dfbdf41
TG		 4037 return 0;
4038}
4039
c80498e3 4040static int __vxlan_dev_create(struct net *net, struct net_device *dev,
653ef6a3
GM		 4041 struct vxlan_config *conf,
4042 struct netlink_ext_ack *extack)
c80498e3
ND		 4043{
4044 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4045 struct vxlan_dev *vxlan = netdev_priv(dev);
0ce1822c 4046 struct net_device *remote_dev = NULL;
0241b836 4047 struct vxlan_fdb *f = NULL;
6db92468 4048 bool unregister = false;
0ce1822c 4049 struct vxlan_rdst *dst;
c80498e3
ND		 4050 int err;
4051
0ce1822c 4052 dst = &vxlan->default_dst;
653ef6a3 4053 err = vxlan_dev_configure(net, dev, conf, false, extack);
c80498e3
ND		 4054 if (err)
4055 return err;
4056
4057 dev->ethtool_ops = &vxlan_ethtool_ops;
4058
4059 /* create an fdb entry for a valid default destination */
0ce1822c 4060 if (!vxlan_addr_any(&dst->remote_ip)) {
0241b836 4061 err = vxlan_fdb_create(vxlan, all_zeros_mac,
0ce1822c 4062 &dst->remote_ip,
c80498e3 4063 NUD_REACHABLE | NUD_PERMANENT,
c80498e3 4064 vxlan->cfg.dst_port,
0ce1822c
TY		 4065 dst->remote_vni,
4066 dst->remote_vni,
4067 dst->remote_ifindex,
1274e1cc 4068 NTF_SELF, 0, &f, extack);
c80498e3
ND		 4069 if (err)
4070 return err;
4071 }
4072
4073 err = register_netdevice(dev);
0241b836
RP		 4074 if (err)
4075 goto errout;
6db92468 4076 unregister = true;
0241b836 4077
0ce1822c
TY		 4078 if (dst->remote_ifindex) {
4079 remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
832e0979
ZC		 4080 if (!remote_dev) {
4081 err = -ENODEV;
0ce1822c 4082 goto errout;
832e0979 4083 }
0ce1822c
TY		 4084
4085 err = netdev_upper_dev_link(remote_dev, dev, extack);
4086 if (err)
4087 goto errout;
4088 }
4089
0241b836 4090 err = rtnl_configure_link(dev, NULL);
2eabcb8a 4091 if (err < 0)
0ce1822c 4092 goto unlink;
c80498e3 4093
61f46fe8 4094 if (f) {
0ce1822c 4095 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
7c31e54a
TY		 4096
4097 /* notify default fdb entry */
61f46fe8 4098 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
4c59b7d1 4099 RTM_NEWNEIGH, true, extack);
7c31e54a
TY		 4100 if (err) {
4101 vxlan_fdb_destroy(vxlan, f, false, false);
0ce1822c
TY		 4102 if (remote_dev)
4103 netdev_upper_dev_unlink(remote_dev, dev);
7c31e54a
TY		 4104 goto unregister;
4105 }
61f46fe8 4106 }
0241b836 4107
c80498e3 4108 list_add(&vxlan->next, &vn->vxlan_list);
0ce1822c
TY		 4109 if (remote_dev)
4110 dst->remote_dev = remote_dev;
c80498e3 4111 return 0;
0ce1822c
TY		 4112unlink:
4113 if (remote_dev)
4114 netdev_upper_dev_unlink(remote_dev, dev);
0241b836 4115errout:
6db92468
PM		 4116 /* unregister_netdevice() destroys the default FDB entry with deletion
4117 * notification. But the addition notification was not sent yet, so
4118 * destroy the entry by hand here.
4119 */
0241b836 4120 if (f)
7c31e54a
TY		 4121 __vxlan_fdb_free(f);
4122unregister:
6db92468
PM		 4123 if (unregister)
4124 unregister_netdevice(dev);
0241b836 4125 return err;
c80498e3
ND		 4126}
4127
70fb0828
RP		 4128/* Set/clear flags based on attribute */
4129static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
4130 int attrtype, unsigned long mask, bool changelink,
4131 bool changelink_supported,
4132 struct netlink_ext_ack *extack)
4133{
4134 unsigned long flags;
4135
4136 if (!tb[attrtype])
4137 return 0;
4138
4139 if (changelink && !changelink_supported) {
4140 vxlan_flag_attr_error(attrtype, extack);
4141 return -EOPNOTSUPP;
4142 }
4143
4144 if (vxlan_policy[attrtype].type == NLA_FLAG)
4145 flags = conf->flags | mask;
4146 else if (nla_get_u8(tb[attrtype]))
4147 flags = conf->flags | mask;
4148 else
4149 flags = conf->flags & ~mask;
4150
4151 conf->flags = flags;
4152
4153 return 0;
4154}
4155
8bcdc4f3
RP		 4156static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
4157 struct net_device *dev, struct vxlan_config *conf,
70fb0828 4158 bool changelink, struct netlink_ext_ack *extack)
0dfbdf41 4159{
8bcdc4f3 4160 struct vxlan_dev *vxlan = netdev_priv(dev);
70fb0828 4161 int err = 0;
0dfbdf41 4162
8bcdc4f3 4163 memset(conf, 0, sizeof(*conf));
e277de5f 4164
8bcdc4f3
RP		 4165 /* if changelink operation, start with old existing cfg */
4166 if (changelink)
4167 memcpy(conf, &vxlan->cfg, sizeof(*conf));
4168
4169 if (data[IFLA_VXLAN_ID]) {
4170 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
4171
70fb0828
RP		 4172 if (changelink && (vni != conf->vni)) {
4173 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
8bcdc4f3 4174 return -EOPNOTSUPP;
70fb0828 4175 }
8bcdc4f3
RP		 4176 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
4177 }
0dfbdf41
TG		 4178
4179 if (data[IFLA_VXLAN_GROUP]) {
70fb0828
RP		 4180 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
4181 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
ce44a4ae 4182 return -EOPNOTSUPP;
70fb0828 4183 }
ce44a4ae 4184
8bcdc4f3 4185 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
ce44a4ae 4186 conf->remote_ip.sa.sa_family = AF_INET;
0dfbdf41 4187 } else if (data[IFLA_VXLAN_GROUP6]) {
70fb0828
RP		 4188 if (!IS_ENABLED(CONFIG_IPV6)) {
4189 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
0dfbdf41 4190 return -EPFNOSUPPORT;
70fb0828 4191 }
0dfbdf41 4192
70fb0828
RP		 4193 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
4194 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
ce44a4ae 4195 return -EOPNOTSUPP;
70fb0828 4196 }
ce44a4ae 4197
8bcdc4f3
RP		 4198 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
4199 conf->remote_ip.sa.sa_family = AF_INET6;
0dfbdf41
TG		 4200 }
4201
4f94c353
JV		 4202 if (data[IFLA_VXLAN_FAN_MAP]) {
4203 err = vxlan_parse_fan_map(data, vxlan);
4204 if (err)
4205 return err;
4206 }
4207
0dfbdf41 4208 if (data[IFLA_VXLAN_LOCAL]) {
70fb0828
RP		 4209 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
4210 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
ce44a4ae 4211 return -EOPNOTSUPP;
70fb0828 4212 }
ce44a4ae 4213
8bcdc4f3
RP		 4214 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
4215 conf->saddr.sa.sa_family = AF_INET;
0dfbdf41 4216 } else if (data[IFLA_VXLAN_LOCAL6]) {
70fb0828
RP		 4217 if (!IS_ENABLED(CONFIG_IPV6)) {
4218 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
0dfbdf41 4219 return -EPFNOSUPPORT;
70fb0828 4220 }
0dfbdf41 4221
70fb0828
RP		 4222 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
4223 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
ce44a4ae 4224 return -EOPNOTSUPP;
70fb0828 4225 }
ce44a4ae 4226
0dfbdf41 4227 /* TODO: respect scope id */
8bcdc4f3
RP		 4228 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
4229 conf->saddr.sa.sa_family = AF_INET6;
0dfbdf41
TG		 4230 }
4231
4232 if (data[IFLA_VXLAN_LINK])
8bcdc4f3 4233 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
0dfbdf41 4234
d342894c 4235 if (data[IFLA_VXLAN_TOS])
8bcdc4f3 4236 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 4237
afb97186 4238 if (data[IFLA_VXLAN_TTL])
8bcdc4f3 4239 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 4240
72f6d71e 4241 if (data[IFLA_VXLAN_TTL_INHERIT]) {
70fb0828
RP		 4242 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4243 VXLAN_F_TTL_INHERIT, changelink, false,
4244 extack);
4245 if (err)
4246 return err;
4247
72f6d71e
HL		 4248 }
4249
e7f70af1 4250 if (data[IFLA_VXLAN_LABEL])
8bcdc4f3 4251 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
e7f70af1
DB		 4252 IPV6_FLOWLABEL_MASK;
4253
8bcdc4f3 4254 if (data[IFLA_VXLAN_LEARNING]) {
70fb0828
RP		 4255 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4256 VXLAN_F_LEARN, changelink, true,
4257 extack);
4258 if (err)
4259 return err;
8bcdc4f3
RP		 4260 } else if (!changelink) {
4261 /* default to learn on a new device */
4262 conf->flags |= VXLAN_F_LEARN;
4263 }
d342894c 4264
40051c4d 4265 if (data[IFLA_VXLAN_AGEING])
8bcdc4f3 4266 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
d342894c 4267
8bcdc4f3 4268 if (data[IFLA_VXLAN_PROXY]) {
70fb0828
RP		 4269 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4270 VXLAN_F_PROXY, changelink, false,
4271 extack);
4272 if (err)
4273 return err;
8bcdc4f3 4274 }
e4f67add 4275
8bcdc4f3 4276 if (data[IFLA_VXLAN_RSC]) {
70fb0828
RP		 4277 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4278 VXLAN_F_RSC, changelink, false,
4279 extack);
4280 if (err)
4281 return err;
8bcdc4f3 4282 }
e4f67add 4283
8bcdc4f3 4284 if (data[IFLA_VXLAN_L2MISS]) {
70fb0828
RP		 4285 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4286 VXLAN_F_L2MISS, changelink, false,
4287 extack);
4288 if (err)
4289 return err;
8bcdc4f3 4290 }
e4f67add 4291
8bcdc4f3 4292 if (data[IFLA_VXLAN_L3MISS]) {
70fb0828
RP		 4293 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4294 VXLAN_F_L3MISS, changelink, false,
4295 extack);
4296 if (err)
4297 return err;
8bcdc4f3 4298 }
e4f67add 4299
8bcdc4f3 4300 if (data[IFLA_VXLAN_LIMIT]) {
70fb0828
RP		 4301 if (changelink) {
4302 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4303 "Cannot change limit");
8bcdc4f3 4304 return -EOPNOTSUPP;
70fb0828 4305 }
8bcdc4f3
RP		 4306 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4307 }
d342894c 4308
8bcdc4f3 4309 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
70fb0828
RP		 4310 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4311 VXLAN_F_COLLECT_METADATA, changelink, false,
4312 extack);
4313 if (err)
4314 return err;
8bcdc4f3 4315 }
f8a9b1bc 4316
05f47d69 4317 if (data[IFLA_VXLAN_PORT_RANGE]) {
8bcdc4f3
RP		 4318 if (!changelink) {
4319 const struct ifla_vxlan_port_range *p
4320 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4321 conf->port_min = ntohs(p->low);
4322 conf->port_max = ntohs(p->high);
4323 } else {
70fb0828
RP		 4324 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4325 "Cannot change port range");
8bcdc4f3
RP		 4326 return -EOPNOTSUPP;
4327 }
05f47d69 4328 }
4329
8bcdc4f3 4330 if (data[IFLA_VXLAN_PORT]) {
70fb0828
RP		 4331 if (changelink) {
4332 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4333 "Cannot change port");
8bcdc4f3 4334 return -EOPNOTSUPP;
70fb0828 4335 }
8bcdc4f3
RP		 4336 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4337 }
823aa873 4338
8bcdc4f3 4339 if (data[IFLA_VXLAN_UDP_CSUM]) {
70fb0828
RP		 4340 if (changelink) {
4341 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4342 "Cannot change UDP_CSUM flag");
8bcdc4f3 4343 return -EOPNOTSUPP;
70fb0828 4344 }
8bcdc4f3
RP		 4345 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4346 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4347 }
359a0ea9 4348
8bcdc4f3 4349 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
70fb0828
RP		 4350 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4351 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4352 false, extack);
4353 if (err)
4354 return err;
8bcdc4f3 4355 }
359a0ea9 4356
8bcdc4f3 4357 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
70fb0828
RP		 4358 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4359 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4360 false, extack);
4361 if (err)
4362 return err;
8bcdc4f3 4363 }
359a0ea9 4364
8bcdc4f3 4365 if (data[IFLA_VXLAN_REMCSUM_TX]) {
70fb0828
RP		 4366 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4367 VXLAN_F_REMCSUM_TX, changelink, false,
4368 extack);
4369 if (err)
4370 return err;
8bcdc4f3 4371 }
dfd8645e 4372
8bcdc4f3 4373 if (data[IFLA_VXLAN_REMCSUM_RX]) {
70fb0828
RP		 4374 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4375 VXLAN_F_REMCSUM_RX, changelink, false,
4376 extack);
4377 if (err)
4378 return err;
8bcdc4f3
RP		 4379 }
4380
4381 if (data[IFLA_VXLAN_GBP]) {
70fb0828
RP		 4382 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4383 VXLAN_F_GBP, changelink, false, extack);
4384 if (err)
4385 return err;
8bcdc4f3
RP		 4386 }
4387
4388 if (data[IFLA_VXLAN_GPE]) {
70fb0828
RP		 4389 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4390 VXLAN_F_GPE, changelink, false,
4391 extack);
4392 if (err)
4393 return err;
8bcdc4f3
RP		 4394 }
4395
4396 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
70fb0828
RP		 4397 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4398 VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4399 false, extack);
4400 if (err)
4401 return err;
8bcdc4f3
RP		 4402 }
4403
4404 if (tb[IFLA_MTU]) {
70fb0828
RP		 4405 if (changelink) {
4406 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4407 "Cannot change mtu");
8bcdc4f3 4408 return -EOPNOTSUPP;
70fb0828 4409 }
8bcdc4f3
RP		 4410 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4411 }
4412
b4d30697
SB		 4413 if (data[IFLA_VXLAN_DF])
4414 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4415
8bcdc4f3
RP		 4416 return 0;
4417}
4418
4419static int vxlan_newlink(struct net *src_net, struct net_device *dev,
7a3f4a18
MS		 4420 struct nlattr *tb[], struct nlattr *data[],
4421 struct netlink_ext_ack *extack)
8bcdc4f3 4422{
8bcdc4f3
RP		 4423 struct vxlan_config conf;
4424 int err;
4425
70fb0828 4426 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
8bcdc4f3
RP		 4427 if (err)
4428 return err;
4429
653ef6a3 4430 return __vxlan_dev_create(src_net, dev, &conf, extack);
8bcdc4f3 4431}
dfd8645e 4432
8bcdc4f3 4433static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
ad744b22
MS		 4434 struct nlattr *data[],
4435 struct netlink_ext_ack *extack)
8bcdc4f3
RP		 4436{
4437 struct vxlan_dev *vxlan = netdev_priv(dev);
8db9427d 4438 struct net_device *lowerdev;
8bcdc4f3 4439 struct vxlan_config conf;
0ce1822c 4440 struct vxlan_rdst *dst;
8bcdc4f3
RP		 4441 int err;
4442
0ce1822c 4443 dst = &vxlan->default_dst;
70fb0828 4444 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
8bcdc4f3
RP		 4445 if (err)
4446 return err;
3511494c 4447
8db9427d
PM		 4448 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4449 vxlan, extack);
8bcdc4f3
RP		 4450 if (err)
4451 return err;
0ace2ca8 4452
c6761cf5
TY		 4453 if (dst->remote_dev == lowerdev)
4454 lowerdev = NULL;
4455
0ce1822c
TY		 4456 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4457 extack);
4458 if (err)
4459 return err;
4460
8bcdc4f3 4461 /* handle default dst entry */
038a5a99 4462 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
fe1e0713
L		 4463 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4464
4465 spin_lock_bh(&vxlan->hash_lock[hash_index]);
038a5a99 4466 if (!vxlan_addr_any(&conf.remote_ip)) {
ce5e098f 4467 err = vxlan_fdb_update(vxlan, all_zeros_mac,
038a5a99 4468 &conf.remote_ip,
8bcdc4f3 4469 NUD_REACHABLE | NUD_PERMANENT,
ce5e098f 4470 NLM_F_APPEND | NLM_F_CREATE,
8bcdc4f3 4471 vxlan->cfg.dst_port,
038a5a99
PM		 4472 conf.vni, conf.vni,
4473 conf.remote_ifindex,
1274e1cc 4474 NTF_SELF, 0, true, extack);
8bcdc4f3 4475 if (err) {
fe1e0713 4476 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
0ce1822c
TY		 4477 netdev_adjacent_change_abort(dst->remote_dev,
4478 lowerdev, dev);
8bcdc4f3
RP		 4479 return err;
4480 }
4481 }
1cdc98c2
PM		 4482 if (!vxlan_addr_any(&dst->remote_ip))
4483 __vxlan_fdb_delete(vxlan, all_zeros_mac,
4484 dst->remote_ip,
4485 vxlan->cfg.dst_port,
4486 dst->remote_vni,
4487 dst->remote_vni,
4488 dst->remote_ifindex,
4489 true);
fe1e0713 4490 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
8bcdc4f3 4491 }
ce577668 4492
038a5a99
PM		 4493 if (conf.age_interval != vxlan->cfg.age_interval)
4494 mod_timer(&vxlan->age_timer, jiffies);
4495
0ce1822c 4496 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
845e0ebb 4497 if (lowerdev && lowerdev != dst->remote_dev)
0ce1822c 4498 dst->remote_dev = lowerdev;
038a5a99 4499 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
8bcdc4f3 4500 return 0;
d342894c 4501}
4502
4503static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4504{
4505 struct vxlan_dev *vxlan = netdev_priv(dev);
4506
8b3f9337
RP		 4507 vxlan_flush(vxlan, true);
4508
553675fb 4509 list_del(&vxlan->next);
d342894c 4510 unregister_netdevice_queue(dev, head);
0ce1822c
TY		 4511 if (vxlan->default_dst.remote_dev)
4512 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
d342894c 4513}
4514
4515static size_t vxlan_get_size(const struct net_device *dev)
4516{
4517
4518 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 4519 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 4520 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 4521 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 4522 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
8fd78069 4523 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
d342894c 4524 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
b4d30697 4525 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */
e7f70af1 4526 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 4527 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 4528 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
4529 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
4530 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
4531 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 4532 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 4533 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
4534 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 4535 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 4536 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4537 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4538 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4539 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 4540 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4541 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4f94c353 4542 nla_total_size(sizeof(struct ip_fan_map) * 256) +
d342894c 4543 0;
4544}
4545
4546static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4547{
4548 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 4549 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 4550 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 4551 .low = htons(vxlan->cfg.port_min),
4552 .high = htons(vxlan->cfg.port_max),
05f47d69 4553 };
d342894c 4554
54bfd872 4555 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 4556 goto nla_put_failure;
4557
e4c7ed41
CW		 4558 if (!vxlan_addr_any(&dst->remote_ip)) {
4559 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 4560 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4561 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 4562 goto nla_put_failure;
4563#if IS_ENABLED(CONFIG_IPV6)
4564 } else {
930345ea
JB		 4565 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4566 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 4567 goto nla_put_failure;
4568#endif
4569 }
4570 }
d342894c 4571
c7995c43 4572 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 4573 goto nla_put_failure;
4574
0dfbdf41
TG		 4575 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4576 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 4577 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 4578 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 4579 goto nla_put_failure;
4580#if IS_ENABLED(CONFIG_IPV6)
4581 } else {
930345ea 4582 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 4583 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 4584 goto nla_put_failure;
4585#endif
4586 }
4587 }
d342894c 4588
4f94c353
JV		 4589 if (fan_has_map(&vxlan->fan)) {
4590 struct nlattr *fan_nest;
4591 struct ip_fan_map *fan_map;
4592
4593 fan_nest = nla_nest_start(skb, IFLA_VXLAN_FAN_MAP);
4594 if (!fan_nest)
4595 goto nla_put_failure;
4596 list_for_each_entry_rcu(fan_map, &vxlan->fan.fan_maps, list) {
4597 struct ifla_fan_map map;
4598
4599 map.underlay = fan_map->underlay;
4600 map.underlay_prefix = fan_map->underlay_prefix;
4601 map.overlay = fan_map->overlay;
4602 map.overlay_prefix = fan_map->overlay_prefix;
4603 if (nla_put(skb, IFLA_FAN_MAPPING, sizeof(map), &map))
4604 goto nla_put_failure;
4605 }
4606 nla_nest_end(skb, fan_nest);
4607 }
4608
0dfbdf41 4609 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
8fd78069
HL		 4610 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4611 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
0dfbdf41 4612 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
b4d30697 4613 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
e7f70af1 4614 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add 4615 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
98c81476 4616 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
e4f67add 4617 nla_put_u8(skb, IFLA_VXLAN_PROXY,
98c81476 4618 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
dc5321d7
MS		 4619 nla_put_u8(skb, IFLA_VXLAN_RSC,
4620 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
e4f67add 4621 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
98c81476 4622 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
e4f67add 4623 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
98c81476 4624 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
da8b43c0 4625 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
dc5321d7 4626 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 4627 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4628 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4629 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 4630 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
98c81476 4631 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9 4632 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
98c81476 4633 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
359a0ea9 4634 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
98c81476 4635 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
dfd8645e 4636 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
98c81476 4637 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
dfd8645e 4638 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
98c81476 4639 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
d342894c 4640 goto nla_put_failure;
4641
05f47d69 4642 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4643 goto nla_put_failure;
4644
dc5321d7 4645 if (vxlan->cfg.flags & VXLAN_F_GBP &&
3511494c
TG		 4646 nla_put_flag(skb, IFLA_VXLAN_GBP))
4647 goto nla_put_failure;
4648
dc5321d7 4649 if (vxlan->cfg.flags & VXLAN_F_GPE &&
e1e5314d
JB		 4650 nla_put_flag(skb, IFLA_VXLAN_GPE))
4651 goto nla_put_failure;
4652
dc5321d7 4653 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
0ace2ca8
TH		 4654 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4655 goto nla_put_failure;
4656
d342894c 4657 return 0;
4658
4659nla_put_failure:
4660 return -EMSGSIZE;
4661}
4662
1728d4fa
ND		 4663static struct net *vxlan_get_link_net(const struct net_device *dev)
4664{
4665 struct vxlan_dev *vxlan = netdev_priv(dev);
4666
4667 return vxlan->net;
4668}
4669
d342894c 4670static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4671 .kind = "vxlan",
4672 .maxtype = IFLA_VXLAN_MAX,
4673 .policy = vxlan_policy,
4674 .priv_size = sizeof(struct vxlan_dev),
4675 .setup = vxlan_setup,
4676 .validate = vxlan_validate,
4677 .newlink = vxlan_newlink,
8bcdc4f3 4678 .changelink = vxlan_changelink,
d342894c 4679 .dellink = vxlan_dellink,
4680 .get_size = vxlan_get_size,
4681 .fill_info = vxlan_fill_info,
1728d4fa 4682 .get_link_net = vxlan_get_link_net,
d342894c 4683};
4684
cf5da330
ND		 4685struct net_device *vxlan_dev_create(struct net *net, const char *name,
4686 u8 name_assign_type,
4687 struct vxlan_config *conf)
4688{
4689 struct nlattr *tb[IFLA_MAX + 1];
4690 struct net_device *dev;
4691 int err;
4692
4693 memset(&tb, 0, sizeof(tb));
4694
4695 dev = rtnl_create_link(net, name, name_assign_type,
d0522f1c 4696 &vxlan_link_ops, tb, NULL);
cf5da330
ND		 4697 if (IS_ERR(dev))
4698 return dev;
4699
653ef6a3 4700 err = __vxlan_dev_create(net, dev, conf, NULL);
cf5da330
ND		 4701 if (err < 0) {
4702 free_netdev(dev);
4703 return ERR_PTR(err);
4704 }
4705
4706 err = rtnl_configure_link(dev, NULL);
4707 if (err < 0) {
4708 LIST_HEAD(list_kill);
4709
4710 vxlan_dellink(dev, &list_kill);
4711 unregister_netdevice_many(&list_kill);
4712 return ERR_PTR(err);
4713 }
4714
4715 return dev;
4716}
4717EXPORT_SYMBOL_GPL(vxlan_dev_create);
4718
acaf4e70
DB		 4719static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4720 struct net_device *dev)
4721{
4722 struct vxlan_dev *vxlan, *next;
4723 LIST_HEAD(list_kill);
4724
4725 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4726 struct vxlan_rdst *dst = &vxlan->default_dst;
4727
4728 /* In case we created vxlan device with carrier
4729 * and we loose the carrier due to module unload
4730 * we also need to remove vxlan device. In other
4731 * cases, it's not necessary and remote_ifindex
4732 * is 0 here, so no matches.
4733 */
4734 if (dst->remote_ifindex == dev->ifindex)
4735 vxlan_dellink(vxlan->dev, &list_kill);
4736 }
4737
4738 unregister_netdevice_many(&list_kill);
4739}
4740
b7aade15
HFS		 4741static int vxlan_netdevice_event(struct notifier_block *unused,
4742 unsigned long event, void *ptr)
acaf4e70
DB		 4743{
4744 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 4745 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 4746
04584957 4747 if (event == NETDEV_UNREGISTER) {
cc4e3835
JK		 4748 if (!dev->udp_tunnel_nic_info)
4749 vxlan_offload_rx_ports(dev, false);
acaf4e70 4750 vxlan_handle_lowerdev_unregister(vn, dev);
04584957 4751 } else if (event == NETDEV_REGISTER) {
cc4e3835
JK		 4752 if (!dev->udp_tunnel_nic_info)
4753 vxlan_offload_rx_ports(dev, true);
04584957
SD		 4754 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
4755 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
2d2b13fc 4756 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
04584957 4757 }
acaf4e70
DB		 4758
4759 return NOTIFY_DONE;
4760}
4761
4762static struct notifier_block vxlan_notifier_block __read_mostly = {
b7aade15 4763 .notifier_call = vxlan_netdevice_event,
acaf4e70
DB		 4764};
4765
0efe1173
PM		 4766static void
4767vxlan_fdb_offloaded_set(struct net_device *dev,
4768 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4769{
4770 struct vxlan_dev *vxlan = netdev_priv(dev);
4771 struct vxlan_rdst *rdst;
4772 struct vxlan_fdb *f;
fe1e0713
L		 4773 u32 hash_index;
4774
4775 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
0efe1173 4776
fe1e0713 4777 spin_lock_bh(&vxlan->hash_lock[hash_index]);
0efe1173
PM		 4778
4779 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4780 if (!f)
4781 goto out;
4782
4783 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4784 fdb_info->remote_port,
4785 fdb_info->remote_vni,
4786 fdb_info->remote_ifindex);
4787 if (!rdst)
4788 goto out;
4789
4790 rdst->offloaded = fdb_info->offloaded;
4791
4792out:
fe1e0713 4793 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
0efe1173
PM		 4794}
4795
5728ae0d
PM		 4796static int
4797vxlan_fdb_external_learn_add(struct net_device *dev,
4798 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4799{
4800 struct vxlan_dev *vxlan = netdev_priv(dev);
4c59b7d1 4801 struct netlink_ext_ack *extack;
fe1e0713 4802 u32 hash_index;
5728ae0d
PM		 4803 int err;
4804
fe1e0713 4805 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4c59b7d1
PM		 4806 extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4807
fe1e0713 4808 spin_lock_bh(&vxlan->hash_lock[hash_index]);
5728ae0d
PM		 4809 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4810 NUD_REACHABLE,
4811 NLM_F_CREATE | NLM_F_REPLACE,
4812 fdb_info->remote_port,
4813 fdb_info->vni,
4814 fdb_info->remote_vni,
4815 fdb_info->remote_ifindex,
4816 NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
1274e1cc 4817 0, false, extack);
fe1e0713 4818 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
5728ae0d
PM		 4819
4820 return err;
4821}
4822
4823static int
4824vxlan_fdb_external_learn_del(struct net_device *dev,
4825 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4826{
4827 struct vxlan_dev *vxlan = netdev_priv(dev);
4828 struct vxlan_fdb *f;
fe1e0713 4829 u32 hash_index;
5728ae0d
PM		 4830 int err = 0;
4831
fe1e0713
L		 4832 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4833 spin_lock_bh(&vxlan->hash_lock[hash_index]);
5728ae0d
PM		 4834
4835 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4836 if (!f)
4837 err = -ENOENT;
4838 else if (f->flags & NTF_EXT_LEARNED)
4839 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4840 fdb_info->remote_ip,
4841 fdb_info->remote_port,
4842 fdb_info->vni,
4843 fdb_info->remote_vni,
4844 fdb_info->remote_ifindex,
4845 false);
4846
fe1e0713 4847 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
5728ae0d
PM		 4848
4849 return err;
4850}
4851
0efe1173
PM		 4852static int vxlan_switchdev_event(struct notifier_block *unused,
4853 unsigned long event, void *ptr)
4854{
4855 struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
5728ae0d
PM		 4856 struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4857 int err = 0;
0efe1173
PM		 4858
4859 switch (event) {
4860 case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4861 vxlan_fdb_offloaded_set(dev, ptr);
4862 break;
5728ae0d
PM		 4863 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4864 fdb_info = ptr;
4865 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4866 if (err) {
4867 err = notifier_from_errno(err);
4868 break;
4869 }
4870 fdb_info->offloaded = true;
4871 vxlan_fdb_offloaded_set(dev, fdb_info);
4872 break;
4873 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4874 fdb_info = ptr;
4875 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4876 if (err) {
4877 err = notifier_from_errno(err);
4878 break;
4879 }
4880 fdb_info->offloaded = false;
4881 vxlan_fdb_offloaded_set(dev, fdb_info);
4882 break;
0efe1173
PM		 4883 }
4884
5728ae0d 4885 return err;
0efe1173
PM		 4886}
4887
4888static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4889 .notifier_call = vxlan_switchdev_event,
4890};
4891
79472fe8
RP		 4892static void vxlan_fdb_nh_flush(struct nexthop *nh)
4893{
4894 struct vxlan_fdb *fdb;
4895 struct vxlan_dev *vxlan;
4896 u32 hash_index;
4897
4898 rcu_read_lock();
4899 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4900 vxlan = rcu_dereference(fdb->vdev);
4901 WARN_ON(!vxlan);
4902 hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4903 vxlan->default_dst.remote_vni);
4904 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4905 if (!hlist_unhashed(&fdb->hlist))
4906 vxlan_fdb_destroy(vxlan, fdb, false, false);
4907 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4908 }
4909 rcu_read_unlock();
4910}
4911
c7cdbe2e
RP		 4912static int vxlan_nexthop_event(struct notifier_block *nb,
4913 unsigned long event, void *ptr)
4914{
1ec69d18
IS		 4915 struct nh_notifier_info *info = ptr;
4916 struct nexthop *nh;
4917
4918 if (event != NEXTHOP_EVENT_DEL)
4919 return NOTIFY_DONE;
c7cdbe2e 4920
1ec69d18
IS		 4921 nh = nexthop_find_by_id(info->net, info->id);
4922 if (!nh)
c7cdbe2e
RP		 4923 return NOTIFY_DONE;
4924
79472fe8 4925 vxlan_fdb_nh_flush(nh);
c7cdbe2e
RP		 4926
4927 return NOTIFY_DONE;
4928}
4929
d342894c 4930static __net_init int vxlan_init_net(struct net *net)
4931{
4932 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 4933 unsigned int h;
d342894c 4934
553675fb 4935 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 4936 spin_lock_init(&vn->sock_lock);
626d667b 4937 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event;
d342894c 4938
553675fb 4939 for (h = 0; h < PORT_HASH_SIZE; ++h)
4940 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 4941
ce7e9c8a
IS		 4942 return register_nexthop_notifier(net, &vn->nexthop_notifier_block,
4943 NULL);
d342894c 4944}
4945
4f94c353
JV		 4946#ifdef CONFIG_SYSCTL
4947static struct ctl_table_header *vxlan_fan_header;
4948static unsigned int vxlan_fan_version = 4;
4949
4950static struct ctl_table vxlan_fan_sysctls[] = {
4951 {
4952 .procname = "vxlan",
4953 .data = &vxlan_fan_version,
4954 .maxlen = sizeof(vxlan_fan_version),
4955 .mode = 0444,
4956 .proc_handler = proc_dointvec,
4957 },
4958 {},
4959};
4960#endif /* CONFIG_SYSCTL */
4961
57b61127 4962static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
f01ec1c0
ND		 4963{
4964 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4965 struct vxlan_dev *vxlan, *next;
4966 struct net_device *dev, *aux;
f01ec1c0 4967
f01ec1c0
ND		 4968 for_each_netdev_safe(net, dev, aux)
4969 if (dev->rtnl_link_ops == &vxlan_link_ops)
57b61127 4970 unregister_netdevice_queue(dev, head);
f01ec1c0
ND		 4971
4972 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4973 /* If vxlan->dev is in the same netns, it has already been added
4974 * to the list by the previous loop.
4975 */
cc4807bb 4976 if (!net_eq(dev_net(vxlan->dev), net))
57b61127 4977 unregister_netdevice_queue(vxlan->dev, head);
f01ec1c0
ND		 4978 }
4979
f01ec1c0
ND		 4980}
4981
57b61127
HY		 4982static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4983{
4984 struct net *net;
4985 LIST_HEAD(list);
68ca2b68 4986 unsigned int h;
57b61127
HY		 4987
4988 rtnl_lock();
626d667b
IS		 4989 list_for_each_entry(net, net_list, exit_list) {
4990 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4991
4992 unregister_nexthop_notifier(net, &vn->nexthop_notifier_block);
4993 }
57b61127
HY		 4994 list_for_each_entry(net, net_list, exit_list)
4995 vxlan_destroy_tunnels(net, &list);
4996
4997 unregister_netdevice_many(&list);
4998 rtnl_unlock();
68ca2b68
TY		 4999
5000 list_for_each_entry(net, net_list, exit_list) {
5001 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
5002
5003 for (h = 0; h < PORT_HASH_SIZE; ++h)
5004 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
5005 }
57b61127
HY		 5006}
5007
d342894c 5008static struct pernet_operations vxlan_net_ops = {
5009 .init = vxlan_init_net,
57b61127 5010 .exit_batch = vxlan_exit_batch_net,
d342894c 5011 .id = &vxlan_net_id,
5012 .size = sizeof(struct vxlan_net),
5013};
5014
5015static int __init vxlan_init_module(void)
5016{
5017 int rc;
5018
5019 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
5020
783c1463 5021 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 5022 if (rc)
5023 goto out1;
5024
acaf4e70 5025 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 5026 if (rc)
5027 goto out2;
5028
0efe1173 5029 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
acaf4e70
DB		 5030 if (rc)
5031 goto out3;
d342894c 5032
0efe1173
PM		 5033 rc = rtnl_link_register(&vxlan_link_ops);
5034 if (rc)
5035 goto out4;
5036
4f94c353
JV		 5037#ifdef CONFIG_SYSCTL
5038 vxlan_fan_header = register_net_sysctl(&init_net, "net/fan",
5039 vxlan_fan_sysctls);
5040 if (!vxlan_fan_header) {
5041 rc = -ENOMEM;
5042 goto sysctl_failed;
5043 }
5044#endif /* CONFIG_SYSCTL */
5045
acaf4e70 5046 return 0;
4f94c353
JV		 5047#ifdef CONFIG_SYSCTL
5048sysctl_failed:
5049 rtnl_link_unregister(&vxlan_link_ops);
5050#endif /* CONFIG_SYSCTL */
0efe1173
PM		 5051out4:
5052 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
acaf4e70
DB		 5053out3:
5054 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 5055out2:
783c1463 5056 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 5057out1:
5058 return rc;
5059}
7332a13b 5060late_initcall(vxlan_init_module);
d342894c 5061
5062static void __exit vxlan_cleanup_module(void)
5063{
4f94c353
JV		 5064#ifdef CONFIG_SYSCTL
5065 unregister_net_sysctl_table(vxlan_fan_header);
5066#endif /* CONFIG_SYSCTL */
b7153984 5067 rtnl_link_unregister(&vxlan_link_ops);
0efe1173 5068 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
acaf4e70 5069 unregister_netdevice_notifier(&vxlan_notifier_block);
783c1463
DB		 5070 unregister_pernet_subsys(&vxlan_net_ops);
5071 /* rcu_barrier() is called by netns */
d342894c 5072}
5073module_exit(vxlan_cleanup_module);
5074
5075MODULE_LICENSE("GPL");
5076MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 5077MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 5078MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 5079MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Hirsute Linux kernel mirror