]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
UBUNTU: Ubuntu-4.15.0-96.97
[mirror_ubuntu-bionic-kernel.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
d342894c 14#include <linux/module.h>
15#include <linux/errno.h>
16#include <linux/slab.h>
d342894c 17#include <linux/udp.h>
18#include <linux/igmp.h>
d57420a1 19#include <linux/inetdevice.h>
d342894c 20#include <linux/if_ether.h>
1b13c97f 21#include <linux/ethtool.h>
e4f67add
DS		 22#include <net/arp.h>
23#include <net/ndisc.h>
d342894c 24#include <net/ip.h>
25#include <net/icmp.h>
d342894c 26#include <net/rtnetlink.h>
d342894c 27#include <net/inet_ecn.h>
28#include <net/net_namespace.h>
29#include <net/netns/generic.h>
fa20e0e3 30#include <net/tun_proto.h>
012a5729 31#include <net/vxlan.h>
40d29af0 32
e4c7ed41 33#if IS_ENABLED(CONFIG_IPV6)
e4c7ed41 34#include <net/ip6_tunnel.h>
660d98ca 35#include <net/ip6_checksum.h>
e4c7ed41 36#endif
d342894c 37
38#define VXLAN_VERSION "0.1"
39
553675fb 40#define PORT_HASH_BITS 8
41#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 42#define FDB_AGE_DEFAULT 300 /* 5 min */
43#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
44
23c578bf 45/* UDP port for VXLAN traffic.
46 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 47 * for compatibility with early adopters.
23c578bf 48 */
9daaa397
SH		 49static unsigned short vxlan_port __read_mostly = 8472;
50module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 51MODULE_PARM_DESC(udp_port, "Destination UDP port");
52
53static bool log_ecn_error = true;
54module_param(log_ecn_error, bool, 0644);
55MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
56
c7d03a00 57static unsigned int vxlan_net_id;
0dfbdf41 58static struct rtnl_link_ops vxlan_link_ops;
553675fb 59
7256eac1 60static const u8 all_zeros_mac[ETH_ALEN + 2];
afbd8bae 61
205f356d 62static int vxlan_sock_add(struct vxlan_dev *vxlan);
614732ea 63
a53cb29b
MB		 64static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
65
553675fb 66/* per-network namespace private data for this module */
67struct vxlan_net {
68 struct list_head vxlan_list;
69 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 70 spinlock_t sock_lock;
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;
ae884082 83 u8 flags; /* see ndm_flags */
d342894c 84};
85
d342894c 86/* salt for hash table */
87static u32 vxlan_salt __read_mostly;
88
ee122c79
TG		 89static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
90{
e7030878
TG		 91 return vs->flags & VXLAN_F_COLLECT_METADATA ||
92 ip_tunnel_collect_metadata();
ee122c79
TG		 93}
94
d57420a1
JV		 95static struct ip_fan_map *vxlan_fan_find_map(struct vxlan_dev *vxlan, __be32 daddr)
96{
97 struct ip_fan_map *fan_map;
98
99 rcu_read_lock();
100 list_for_each_entry_rcu(fan_map, &vxlan->fan.fan_maps, list) {
101 if (fan_map->overlay ==
102 (daddr & inet_make_mask(fan_map->overlay_prefix))) {
103 rcu_read_unlock();
104 return fan_map;
105 }
106 }
107 rcu_read_unlock();
108
109 return NULL;
110}
111
112static void vxlan_fan_flush_map(struct vxlan_dev *vxlan)
113{
114 struct ip_fan_map *fan_map;
115
116 list_for_each_entry_rcu(fan_map, &vxlan->fan.fan_maps, list) {
117 list_del_rcu(&fan_map->list);
118 kfree_rcu(fan_map, rcu);
119 }
120}
121
122static int vxlan_fan_del_map(struct vxlan_dev *vxlan, __be32 overlay)
123{
124 struct ip_fan_map *fan_map;
125
126 fan_map = vxlan_fan_find_map(vxlan, overlay);
127 if (!fan_map)
128 return -ENOENT;
129
130 list_del_rcu(&fan_map->list);
131 kfree_rcu(fan_map, rcu);
132
133 return 0;
134}
135
136static int vxlan_fan_add_map(struct vxlan_dev *vxlan, struct ifla_fan_map *map)
137{
138 __be32 overlay_mask, underlay_mask;
139 struct ip_fan_map *fan_map;
140
141 overlay_mask = inet_make_mask(map->overlay_prefix);
142 underlay_mask = inet_make_mask(map->underlay_prefix);
143
144 netdev_dbg(vxlan->dev, "vfam: map: o %x/%d u %x/%d om %x um %x\n",
145 map->overlay, map->overlay_prefix,
146 map->underlay, map->underlay_prefix,
147 overlay_mask, underlay_mask);
148
149 if ((map->overlay & ~overlay_mask) || (map->underlay & ~underlay_mask))
150 return -EINVAL;
151
152 if (!(map->overlay & overlay_mask) && (map->underlay & underlay_mask))
153 return -EINVAL;
154
155 /* Special case: overlay 0 and underlay 0: flush all mappings */
156 if (!map->overlay && !map->underlay) {
157 vxlan_fan_flush_map(vxlan);
158 return 0;
159 }
160
161 /* Special case: overlay set and underlay 0: clear map for overlay */
162 if (!map->underlay)
163 return vxlan_fan_del_map(vxlan, map->overlay);
164
165 if (vxlan_fan_find_map(vxlan, map->overlay))
166 return -EEXIST;
167
168 fan_map = kmalloc(sizeof(*fan_map), GFP_KERNEL);
169 fan_map->underlay = map->underlay;
170 fan_map->overlay = map->overlay;
171 fan_map->underlay_prefix = map->underlay_prefix;
172 fan_map->overlay_mask = ntohl(overlay_mask);
173 fan_map->overlay_prefix = map->overlay_prefix;
174
175 list_add_tail_rcu(&fan_map->list, &vxlan->fan.fan_maps);
176
177 return 0;
178}
179
180static int vxlan_parse_fan_map(struct nlattr *data[], struct vxlan_dev *vxlan)
181{
182 struct ifla_fan_map *map;
183 struct nlattr *attr;
184 int rem, rv;
185
186 nla_for_each_nested(attr, data[IFLA_IPTUN_FAN_MAP], rem) {
187 map = nla_data(attr);
188 rv = vxlan_fan_add_map(vxlan, map);
189 if (rv)
190 return rv;
191 }
192
193 return 0;
194}
195
196static int vxlan_fan_build_rdst(struct vxlan_dev *vxlan, struct sk_buff *skb,
197 struct vxlan_rdst *fan_rdst)
198{
199 struct ip_fan_map *f_map;
200 union vxlan_addr *va;
201 u32 daddr, underlay;
202 struct arphdr *arp;
203 void *arp_ptr;
204 struct ethhdr *eth;
205 struct iphdr *iph;
206
207 eth = eth_hdr(skb);
208 switch (eth->h_proto) {
209 case htons(ETH_P_IP):
210 iph = ip_hdr(skb);
211 if (!iph)
212 return -EINVAL;
213 daddr = iph->daddr;
214 break;
215 case htons(ETH_P_ARP):
216 arp = arp_hdr(skb);
217 if (!arp)
218 return -EINVAL;
219 arp_ptr = arp + 1;
220 netdev_dbg(vxlan->dev,
221 "vfbr: arp sha %pM sip %pI4 tha %pM tip %pI4\n",
222 arp_ptr, arp_ptr + skb->dev->addr_len,
223 arp_ptr + skb->dev->addr_len + 4,
224 arp_ptr + (skb->dev->addr_len * 2) + 4);
225 arp_ptr += (skb->dev->addr_len * 2) + 4;
226 memcpy(&daddr, arp_ptr, 4);
227 break;
228 default:
229 netdev_dbg(vxlan->dev, "vfbr: unknown eth p %x\n", eth->h_proto);
230 return -EINVAL;
231 }
232
233 f_map = vxlan_fan_find_map(vxlan, daddr);
234 if (!f_map)
235 return -EINVAL;
236
237 daddr = ntohl(daddr);
238 underlay = ntohl(f_map->underlay);
239 if (!underlay)
240 return -EINVAL;
241
242 memset(fan_rdst, 0, sizeof(*fan_rdst));
243 va = &fan_rdst->remote_ip;
244 va->sa.sa_family = AF_INET;
245 fan_rdst->remote_vni = vxlan->default_dst.remote_vni;
246 va->sin.sin_addr.s_addr = htonl(underlay |
247 ((daddr & ~f_map->overlay_mask) >>
248 (32 - f_map->overlay_prefix -
249 (32 - f_map->underlay_prefix))));
250 netdev_dbg(vxlan->dev, "vfbr: daddr %x ul %x dst %x\n",
251 daddr, underlay, va->sin.sin_addr.s_addr);
252
253 return 0;
254}
255
e4c7ed41
CW		 256#if IS_ENABLED(CONFIG_IPV6)
257static inline
258bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
259{
f0ef3126
JB		 260 if (a->sa.sa_family != b->sa.sa_family)
261 return false;
262 if (a->sa.sa_family == AF_INET6)
263 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
264 else
265 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 266}
267
268static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
269{
f0ef3126
JB		 270 if (ipa->sa.sa_family == AF_INET6)
271 return ipv6_addr_any(&ipa->sin6.sin6_addr);
272 else
273 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 274}
275
276static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
277{
f0ef3126
JB		 278 if (ipa->sa.sa_family == AF_INET6)
279 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
280 else
281 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 282}
283
284static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
285{
f0ef3126 286 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 287 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 288 ip->sa.sa_family = AF_INET6;
289 return 0;
290 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 291 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 292 ip->sa.sa_family = AF_INET;
293 return 0;
294 } else {
295 return -EAFNOSUPPORT;
296 }
e4c7ed41
CW		 297}
298
299static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 300 const union vxlan_addr *ip)
e4c7ed41 301{
f0ef3126 302 if (ip->sa.sa_family == AF_INET6)
930345ea 303 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 304 else
930345ea 305 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 306}
307
308#else /* !CONFIG_IPV6 */
309
310static inline
311bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
312{
f0ef3126 313 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 314}
315
316static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
317{
f0ef3126 318 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 319}
320
321static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
322{
f0ef3126 323 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 324}
325
326static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
327{
f0ef3126
JB		 328 if (nla_len(nla) >= sizeof(struct in6_addr)) {
329 return -EAFNOSUPPORT;
330 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 331 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 332 ip->sa.sa_family = AF_INET;
333 return 0;
334 } else {
335 return -EAFNOSUPPORT;
336 }
e4c7ed41
CW		 337}
338
339static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 340 const union vxlan_addr *ip)
e4c7ed41 341{
930345ea 342 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 343}
344#endif
345
553675fb 346/* Virtual Network hash table head */
54bfd872 347static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 348{
54bfd872 349 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 350}
351
352/* Socket hash table head */
353static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 354{
355 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
356
553675fb 357 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
358}
359
3e61aa8f
SH		 360/* First remote destination for a forwarding entry.
361 * Guaranteed to be non-NULL because remotes are never deleted.
362 */
5ca5461c 363static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 364{
5ca5461c 365 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
366}
367
368static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
369{
370 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 371}
372
ac5132d1
TG		 373/* Find VXLAN socket based on network namespace, address family and UDP port
374 * and enabled unshareable flags.
375 */
376static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
377 __be16 port, u32 flags)
553675fb 378{
379 struct vxlan_sock *vs;
af33c1ad
TH		 380
381 flags &= VXLAN_F_RCV_FLAGS;
553675fb 382
383 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 384 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 385 vxlan_get_sk_family(vs) == family &&
af33c1ad 386 vs->flags == flags)
553675fb 387 return vs;
388 }
389 return NULL;
d342894c 390}
391
49f810f0
MS		 392static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
393 __be32 vni)
d342894c 394{
69e76661 395 struct vxlan_dev_node *node;
d342894c 396
b9167b2e
JB		 397 /* For flow based devices, map all packets to VNI 0 */
398 if (vs->flags & VXLAN_F_COLLECT_METADATA)
399 vni = 0;
400
69e76661
JB		 401 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
402 if (node->vxlan->default_dst.remote_vni != vni)
49f810f0
MS		 403 continue;
404
405 if (IS_ENABLED(CONFIG_IPV6)) {
69e76661 406 const struct vxlan_config *cfg = &node->vxlan->cfg;
49f810f0
MS		 407
408 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
409 cfg->remote_ifindex != ifindex)
410 continue;
411 }
412
69e76661 413 return node->vxlan;
d342894c 414 }
415
416 return NULL;
417}
418
5cfccc5a 419/* Look up VNI in a per net namespace table */
49f810f0
MS		 420static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
421 __be32 vni, sa_family_t family,
422 __be16 port, u32 flags)
5cfccc5a
PS		 423{
424 struct vxlan_sock *vs;
425
ac5132d1 426 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 427 if (!vs)
428 return NULL;
429
49f810f0 430 return vxlan_vs_find_vni(vs, ifindex, vni);
5cfccc5a
PS		 431}
432
d342894c 433/* Fill in neighbour message in skbuff. */
434static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 435 const struct vxlan_fdb *fdb,
436 u32 portid, u32 seq, int type, unsigned int flags,
437 const struct vxlan_rdst *rdst)
d342894c 438{
439 unsigned long now = jiffies;
440 struct nda_cacheinfo ci;
441 struct nlmsghdr *nlh;
442 struct ndmsg *ndm;
e4f67add 443 bool send_ip, send_eth;
d342894c 444
445 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
446 if (nlh == NULL)
447 return -EMSGSIZE;
448
449 ndm = nlmsg_data(nlh);
450 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 451
452 send_eth = send_ip = true;
453
454 if (type == RTM_GETNEIGH) {
e4c7ed41 455 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add 456 send_eth = !is_zero_ether_addr(fdb->eth_addr);
8f48ba71 457 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
e4f67add
DS		 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;
545469f7 463 ndm->ndm_type = RTN_UNICAST;
d342894c 464
193523bf 465 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 466 nla_put_s32(skb, NDA_LINK_NETNSID,
38f507f1 467 peernet2id(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 468 goto nla_put_failure;
469
e4f67add 470 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 471 goto nla_put_failure;
472
e4c7ed41 473 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 474 goto nla_put_failure;
475
0dfbdf41 476 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 477 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
478 goto nla_put_failure;
c7995c43 479 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
54bfd872 480 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
6681712d 481 goto nla_put_failure;
dc5321d7 482 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
3ad7a4b1
RP		 483 nla_put_u32(skb, NDA_SRC_VNI,
484 be32_to_cpu(fdb->vni)))
485 goto nla_put_failure;
6681712d
DS		 486 if (rdst->remote_ifindex &&
487 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 488 goto nla_put_failure;
489
490 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
491 ci.ndm_confirmed = 0;
492 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
493 ci.ndm_refcnt = 0;
494
495 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
496 goto nla_put_failure;
497
053c095a
JB		 498 nlmsg_end(skb, nlh);
499 return 0;
d342894c 500
501nla_put_failure:
502 nlmsg_cancel(skb, nlh);
503 return -EMSGSIZE;
504}
505
506static inline size_t vxlan_nlmsg_size(void)
507{
508 return NLMSG_ALIGN(sizeof(struct ndmsg))
509 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 510 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 511 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 512 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
513 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 514 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 515 + nla_total_size(sizeof(struct nda_cacheinfo));
516}
517
9e4b93f9
ND		 518static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
519 struct vxlan_rdst *rd, int type)
d342894c 520{
521 struct net *net = dev_net(vxlan->dev);
522 struct sk_buff *skb;
523 int err = -ENOBUFS;
524
525 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
526 if (skb == NULL)
527 goto errout;
528
9e4b93f9 529 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 530 if (err < 0) {
531 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
532 WARN_ON(err == -EMSGSIZE);
533 kfree_skb(skb);
534 goto errout;
535 }
536
537 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
538 return;
539errout:
540 if (err < 0)
541 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
542}
543
e4c7ed41 544static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 545{
546 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 547 struct vxlan_fdb f = {
548 .state = NUD_STALE,
549 };
550 struct vxlan_rdst remote = {
e4c7ed41 551 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 552 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 553 };
3e61aa8f 554
9e4b93f9 555 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 556}
557
558static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
559{
bb3fd687
SH		 560 struct vxlan_fdb f = {
561 .state = NUD_STALE,
562 };
9e4b93f9 563 struct vxlan_rdst remote = { };
e4f67add 564
e4f67add
DS		 565 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
566
9e4b93f9 567 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 568}
569
d342894c 570/* Hash Ethernet address */
571static u32 eth_hash(const unsigned char *addr)
572{
573 u64 value = get_unaligned((u64 *)addr);
574
575 /* only want 6 bytes */
576#ifdef __BIG_ENDIAN
d342894c 577 value >>= 16;
321fb991 578#else
579 value <<= 16;
d342894c 580#endif
581 return hash_64(value, FDB_HASH_BITS);
582}
583
3ad7a4b1
RP		 584static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
585{
586 /* use 1 byte of OUI and 3 bytes of NIC */
587 u32 key = get_unaligned((u32 *)(addr + 2));
588
589 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
590}
591
d342894c 592/* Hash chain to use given mac address */
593static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
3ad7a4b1 594 const u8 *mac, __be32 vni)
d342894c 595{
dc5321d7 596 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
3ad7a4b1
RP		 597 return &vxlan->fdb_head[eth_vni_hash(mac, vni)];
598 else
599 return &vxlan->fdb_head[eth_hash(mac)];
d342894c 600}
601
602/* Look up Ethernet address in forwarding table */
014be2c8 603static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 604 const u8 *mac, __be32 vni)
d342894c 605{
3ad7a4b1 606 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
d342894c 607 struct vxlan_fdb *f;
d342894c 608
b67bfe0d 609 hlist_for_each_entry_rcu(f, head, hlist) {
3ad7a4b1 610 if (ether_addr_equal(mac, f->eth_addr)) {
dc5321d7 611 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 612 if (vni == f->vni)
613 return f;
614 } else {
615 return f;
616 }
617 }
d342894c 618 }
619
620 return NULL;
621}
622
014be2c8 623static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 624 const u8 *mac, __be32 vni)
014be2c8
SS		 625{
626 struct vxlan_fdb *f;
627
3ad7a4b1 628 f = __vxlan_find_mac(vxlan, mac, vni);
014be2c8
SS		 629 if (f)
630 f->used = jiffies;
631
632 return f;
633}
634
a5e7c10a
MR		 635/* caller should hold vxlan->hash_lock */
636static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 637 union vxlan_addr *ip, __be16 port,
54bfd872 638 __be32 vni, __u32 ifindex)
6681712d 639{
3e61aa8f 640 struct vxlan_rdst *rd;
6681712d 641
3e61aa8f 642 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 643 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 644 rd->remote_port == port &&
645 rd->remote_vni == vni &&
646 rd->remote_ifindex == ifindex)
a5e7c10a 647 return rd;
6681712d 648 }
3e61aa8f 649
a5e7c10a
MR		 650 return NULL;
651}
652
906dc186
TR		 653/* Replace destination of unicast mac */
654static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872
JB		 655 union vxlan_addr *ip, __be16 port, __be32 vni,
656 __u32 ifindex)
906dc186
TR		 657{
658 struct vxlan_rdst *rd;
659
660 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
661 if (rd)
662 return 0;
663
664 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
665 if (!rd)
666 return 0;
0c1d70af
PA		 667
668 dst_cache_reset(&rd->dst_cache);
e4c7ed41 669 rd->remote_ip = *ip;
906dc186
TR		 670 rd->remote_port = port;
671 rd->remote_vni = vni;
672 rd->remote_ifindex = ifindex;
673 return 1;
674}
675
a5e7c10a
MR		 676/* Add/update destinations for multicast */
677static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 678 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 679 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 680{
681 struct vxlan_rdst *rd;
682
683 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
684 if (rd)
685 return 0;
686
6681712d
DS		 687 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
688 if (rd == NULL)
689 return -ENOBUFS;
0c1d70af
PA		 690
691 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
692 kfree(rd);
693 return -ENOBUFS;
694 }
695
e4c7ed41 696 rd->remote_ip = *ip;
6681712d
DS		 697 rd->remote_port = port;
698 rd->remote_vni = vni;
699 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 700
701 list_add_tail_rcu(&rd->list, &f->remotes);
702
9e4b93f9 703 *rdp = rd;
6681712d
DS		 704 return 1;
705}
706
dfd8645e
TH		 707static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
708 unsigned int off,
709 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 710 __be32 vni_field,
711 struct gro_remcsum *grc,
0ace2ca8 712 bool nopartial)
dfd8645e 713{
b7fe10e5 714 size_t start, offset;
dfd8645e
TH		 715
716 if (skb->remcsum_offload)
b7fe10e5 717 return vh;
dfd8645e
TH		 718
719 if (!NAPI_GRO_CB(skb)->csum_valid)
720 return NULL;
721
54bfd872
JB		 722 start = vxlan_rco_start(vni_field);
723 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 724
b7fe10e5
TH		 725 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
726 start, offset, grc, nopartial);
dfd8645e
TH		 727
728 skb->remcsum_offload = 1;
729
730 return vh;
731}
732
5602c48c
TH		 733static struct sk_buff **vxlan_gro_receive(struct sock *sk,
734 struct sk_buff **head,
735 struct sk_buff *skb)
dc01e7d3
OG		 736{
737 struct sk_buff *p, **pp = NULL;
738 struct vxlanhdr *vh, *vh2;
9b174d88 739 unsigned int hlen, off_vx;
dc01e7d3 740 int flush = 1;
5602c48c 741 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
54bfd872 742 __be32 flags;
26c4f7da
TH		 743 struct gro_remcsum grc;
744
745 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 746
747 off_vx = skb_gro_offset(skb);
748 hlen = off_vx + sizeof(*vh);
749 vh = skb_gro_header_fast(skb, off_vx);
750 if (skb_gro_header_hard(skb, hlen)) {
751 vh = skb_gro_header_slow(skb, hlen, off_vx);
752 if (unlikely(!vh))
753 goto out;
754 }
dc01e7d3 755
dfd8645e
TH		 756 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
757
54bfd872 758 flags = vh->vx_flags;
dfd8645e
TH		 759
760 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
761 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 762 vh->vx_vni, &grc,
0ace2ca8
TH		 763 !!(vs->flags &
764 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 765
766 if (!vh)
767 goto out;
768 }
769
b7fe10e5
TH		 770 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
771
dc01e7d3
OG		 772 for (p = *head; p; p = p->next) {
773 if (!NAPI_GRO_CB(p)->same_flow)
774 continue;
775
776 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 777 if (vh->vx_flags != vh2->vx_flags ||
778 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 779 NAPI_GRO_CB(p)->same_flow = 0;
780 continue;
781 }
dc01e7d3
OG		 782 }
783
fcd91dd4 784 pp = call_gro_receive(eth_gro_receive, head, skb);
c194cf93 785 flush = 0;
dc01e7d3 786
dc01e7d3 787out:
12df8689 788 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
dc01e7d3
OG		 789
790 return pp;
791}
792
5602c48c 793static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
dc01e7d3 794{
229740c6
JR		 795 /* Sets 'skb->inner_mac_header' since we are always called with
796 * 'skb->encapsulation' set.
797 */
9b174d88 798 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 799}
800
ff252e19
RP		 801static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan,
802 const u8 *mac, __u16 state,
803 __be32 src_vni, __u8 ndm_flags)
804{
805 struct vxlan_fdb *f;
806
807 f = kmalloc(sizeof(*f), GFP_ATOMIC);
808 if (!f)
809 return NULL;
810 f->state = state;
811 f->flags = ndm_flags;
812 f->updated = f->used = jiffies;
813 f->vni = src_vni;
814 INIT_LIST_HEAD(&f->remotes);
815 memcpy(f->eth_addr, mac, ETH_ALEN);
816
817 return f;
818}
819
d342894c 820static int vxlan_fdb_create(struct vxlan_dev *vxlan,
ff252e19
RP		 821 const u8 *mac, union vxlan_addr *ip,
822 __u16 state, __be16 port, __be32 src_vni,
823 __be32 vni, __u32 ifindex, __u8 ndm_flags,
824 struct vxlan_fdb **fdb)
825{
826 struct vxlan_rdst *rd = NULL;
827 struct vxlan_fdb *f;
828 int rc;
829
830 if (vxlan->cfg.addrmax &&
831 vxlan->addrcnt >= vxlan->cfg.addrmax)
832 return -ENOSPC;
833
834 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
835 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
836 if (!f)
837 return -ENOMEM;
838
839 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
840 if (rc < 0) {
841 kfree(f);
842 return rc;
843 }
844
845 ++vxlan->addrcnt;
846 hlist_add_head_rcu(&f->hlist,
847 vxlan_fdb_head(vxlan, mac, src_vni));
848
849 *fdb = f;
850
851 return 0;
852}
853
854/* Add new entry to forwarding table -- assumes lock held */
855static int vxlan_fdb_update(struct vxlan_dev *vxlan,
e4c7ed41 856 const u8 *mac, union vxlan_addr *ip,
6681712d 857 __u16 state, __u16 flags,
3ad7a4b1
RP		 858 __be16 port, __be32 src_vni, __be32 vni,
859 __u32 ifindex, __u8 ndm_flags)
d342894c 860{
9e4b93f9 861 struct vxlan_rdst *rd = NULL;
d342894c 862 struct vxlan_fdb *f;
863 int notify = 0;
17b46365 864 int rc;
d342894c 865
3ad7a4b1 866 f = __vxlan_find_mac(vxlan, mac, src_vni);
d342894c 867 if (f) {
868 if (flags & NLM_F_EXCL) {
869 netdev_dbg(vxlan->dev,
870 "lost race to create %pM\n", mac);
871 return -EEXIST;
872 }
873 if (f->state != state) {
874 f->state = state;
875 f->updated = jiffies;
876 notify = 1;
877 }
ae884082
DS		 878 if (f->flags != ndm_flags) {
879 f->flags = ndm_flags;
880 f->updated = jiffies;
881 notify = 1;
882 }
906dc186
TR		 883 if ((flags & NLM_F_REPLACE)) {
884 /* Only change unicasts */
885 if (!(is_multicast_ether_addr(f->eth_addr) ||
886 is_zero_ether_addr(f->eth_addr))) {
60840429 887 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 888 ifindex);
906dc186
TR		 889 } else
890 return -EOPNOTSUPP;
891 }
6681712d 892 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 893 (is_multicast_ether_addr(f->eth_addr) ||
894 is_zero_ether_addr(f->eth_addr))) {
17b46365 895 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
6681712d
DS		 896
897 if (rc < 0)
898 return rc;
899 notify |= rc;
900 }
d342894c 901 } else {
902 if (!(flags & NLM_F_CREATE))
903 return -ENOENT;
904
906dc186
TR		 905 /* Disallow replace to add a multicast entry */
906 if ((flags & NLM_F_REPLACE) &&
907 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
908 return -EOPNOTSUPP;
909
e4c7ed41 910 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
ff252e19
RP		 911 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
912 vni, ifindex, ndm_flags, &f);
913 if (rc < 0)
17b46365 914 return rc;
ff252e19 915 notify = 1;
d342894c 916 }
917
9e4b93f9
ND		 918 if (notify) {
919 if (rd == NULL)
920 rd = first_remote_rtnl(f);
921 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
922 }
d342894c 923
924 return 0;
925}
926
6706c82e 927static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 928{
929 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 930 struct vxlan_rdst *rd, *nd;
6681712d 931
0c1d70af
PA		 932 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
933 dst_cache_destroy(&rd->dst_cache);
6681712d 934 kfree(rd);
0c1d70af 935 }
6681712d
DS		 936 kfree(f);
937}
938
d7e686dc
RP		 939static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
940 bool do_notify)
d342894c 941{
942 netdev_dbg(vxlan->dev,
943 "delete %pM\n", f->eth_addr);
944
945 --vxlan->addrcnt;
d7e686dc
RP		 946 if (do_notify)
947 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 948
949 hlist_del_rcu(&f->hlist);
6681712d 950 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 951}
952
35cf2845
LR		 953static void vxlan_dst_free(struct rcu_head *head)
954{
955 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
956
957 dst_cache_destroy(&rd->dst_cache);
958 kfree(rd);
959}
960
961static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
962 struct vxlan_rdst *rd)
963{
964 list_del_rcu(&rd->list);
965 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
966 call_rcu(&rd->rcu, vxlan_dst_free);
967}
968
f0b074be 969static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
3ad7a4b1
RP		 970 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
971 __be32 *vni, u32 *ifindex)
d342894c 972{
6681712d 973 struct net *net = dev_net(vxlan->dev);
e4c7ed41 974 int err;
d342894c 975
f0b074be 976 if (tb[NDA_DST]) {
e4c7ed41
CW		 977 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
978 if (err)
979 return err;
f0b074be 980 } else {
e4c7ed41
CW		 981 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
982 if (remote->sa.sa_family == AF_INET) {
983 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
984 ip->sa.sa_family = AF_INET;
985#if IS_ENABLED(CONFIG_IPV6)
986 } else {
987 ip->sin6.sin6_addr = in6addr_any;
988 ip->sa.sa_family = AF_INET6;
989#endif
990 }
f0b074be 991 }
d342894c 992
6681712d 993 if (tb[NDA_PORT]) {
73cf3317 994 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 995 return -EINVAL;
f0b074be
MR		 996 *port = nla_get_be16(tb[NDA_PORT]);
997 } else {
0dfbdf41 998 *port = vxlan->cfg.dst_port;
f0b074be 999 }
6681712d
DS		 1000
1001 if (tb[NDA_VNI]) {
1002 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
1003 return -EINVAL;
54bfd872 1004 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 1005 } else {
1006 *vni = vxlan->default_dst.remote_vni;
1007 }
6681712d 1008
3ad7a4b1
RP		 1009 if (tb[NDA_SRC_VNI]) {
1010 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
1011 return -EINVAL;
1012 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1013 } else {
1014 *src_vni = vxlan->default_dst.remote_vni;
1015 }
1016
6681712d 1017 if (tb[NDA_IFINDEX]) {
5abb0029 1018 struct net_device *tdev;
6681712d
DS		 1019
1020 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
1021 return -EINVAL;
f0b074be 1022 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 1023 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 1024 if (!tdev)
6681712d 1025 return -EADDRNOTAVAIL;
f0b074be
MR		 1026 } else {
1027 *ifindex = 0;
1028 }
1029
1030 return 0;
1031}
1032
1033/* Add static entry (via netlink) */
1034static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1035 struct net_device *dev,
f6f6424b 1036 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 1037{
1038 struct vxlan_dev *vxlan = netdev_priv(dev);
1039 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 1040 union vxlan_addr ip;
f0b074be 1041 __be16 port;
3ad7a4b1 1042 __be32 src_vni, vni;
54bfd872 1043 u32 ifindex;
f0b074be
MR		 1044 int err;
1045
1046 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1047 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1048 ndm->ndm_state);
1049 return -EINVAL;
1050 }
1051
1052 if (tb[NDA_DST] == NULL)
1053 return -EINVAL;
1054
3ad7a4b1 1055 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
f0b074be
MR		 1056 if (err)
1057 return err;
6681712d 1058
5933a7bb
MR		 1059 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1060 return -EAFNOSUPPORT;
1061
d342894c 1062 spin_lock_bh(&vxlan->hash_lock);
ff252e19 1063 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
3ad7a4b1 1064 port, src_vni, vni, ifindex, ndm->ndm_flags);
d342894c 1065 spin_unlock_bh(&vxlan->hash_lock);
1066
1067 return err;
1068}
1069
3ad7a4b1
RP		 1070static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1071 const unsigned char *addr, union vxlan_addr ip,
fc39c38b
XL		 1072 __be16 port, __be32 src_vni, __be32 vni,
1073 u32 ifindex, u16 vid)
d342894c 1074{
d342894c 1075 struct vxlan_fdb *f;
bc7892ba 1076 struct vxlan_rdst *rd = NULL;
3ad7a4b1 1077 int err = -ENOENT;
bc7892ba 1078
3ad7a4b1 1079 f = vxlan_find_mac(vxlan, addr, src_vni);
bc7892ba 1080 if (!f)
3ad7a4b1 1081 return err;
bc7892ba 1082
e4c7ed41
CW		 1083 if (!vxlan_addr_any(&ip)) {
1084 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 1085 if (!rd)
1086 goto out;
1087 }
1088
bc7892ba
MR		 1089 /* remove a destination if it's not the only one on the list,
1090 * otherwise destroy the fdb entry
1091 */
1092 if (rd && !list_is_singular(&f->remotes)) {
35cf2845 1093 vxlan_fdb_dst_destroy(vxlan, f, rd);
bc7892ba 1094 goto out;
d342894c 1095 }
bc7892ba 1096
d7e686dc 1097 vxlan_fdb_destroy(vxlan, f, true);
bc7892ba
MR		 1098
1099out:
3ad7a4b1
RP		 1100 return 0;
1101}
1102
1103/* Delete entry (via netlink) */
1104static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1105 struct net_device *dev,
1106 const unsigned char *addr, u16 vid)
1107{
1108 struct vxlan_dev *vxlan = netdev_priv(dev);
1109 union vxlan_addr ip;
1110 __be32 src_vni, vni;
1111 __be16 port;
1112 u32 ifindex;
1113 int err;
1114
1115 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
1116 if (err)
1117 return err;
1118
1119 spin_lock_bh(&vxlan->hash_lock);
1120 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1121 vid);
d342894c 1122 spin_unlock_bh(&vxlan->hash_lock);
1123
1124 return err;
1125}
1126
1127/* Dump forwarding table */
1128static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 1129 struct net_device *dev,
d297653d 1130 struct net_device *filter_dev, int *idx)
d342894c 1131{
1132 struct vxlan_dev *vxlan = netdev_priv(dev);
1133 unsigned int h;
d297653d 1134 int err = 0;
d342894c 1135
1136 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1137 struct vxlan_fdb *f;
d342894c 1138
b67bfe0d 1139 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 1140 struct vxlan_rdst *rd;
6681712d 1141
3e61aa8f 1142 list_for_each_entry_rcu(rd, &f->remotes, list) {
d297653d 1143 if (*idx < cb->args[2])
07a51cd3
AW		 1144 goto skip;
1145
6681712d
DS		 1146 err = vxlan_fdb_info(skb, vxlan, f,
1147 NETLINK_CB(cb->skb).portid,
1148 cb->nlh->nlmsg_seq,
1149 RTM_NEWNEIGH,
1150 NLM_F_MULTI, rd);
d297653d 1151 if (err < 0)
3e61aa8f 1152 goto out;
3e61aa8f 1153skip:
d297653d 1154 *idx += 1;
07a51cd3 1155 }
d342894c 1156 }
1157 }
3e61aa8f 1158out:
d297653d 1159 return err;
d342894c 1160}
1161
1162/* Watch incoming packets to learn mapping between Ethernet address
1163 * and Tunnel endpoint.
c4b49512 1164 * Return true if packet is bogus and should be dropped.
d342894c 1165 */
26a41ae6 1166static bool vxlan_snoop(struct net_device *dev,
3ad7a4b1 1167 union vxlan_addr *src_ip, const u8 *src_mac,
87613de9 1168 u32 src_ifindex, __be32 vni)
d342894c 1169{
1170 struct vxlan_dev *vxlan = netdev_priv(dev);
1171 struct vxlan_fdb *f;
87613de9
MS		 1172 u32 ifindex = 0;
1173
1174#if IS_ENABLED(CONFIG_IPV6)
1175 if (src_ip->sa.sa_family == AF_INET6 &&
1176 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1177 ifindex = src_ifindex;
1178#endif
d342894c 1179
3ad7a4b1 1180 f = vxlan_find_mac(vxlan, src_mac, vni);
d342894c 1181 if (likely(f)) {
5ca5461c 1182 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 1183
87613de9
MS		 1184 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1185 rdst->remote_ifindex == ifindex))
26a41ae6 1186 return false;
1187
1188 /* Don't migrate static entries, drop packets */
e0090a9e 1189 if (f->state & (NUD_PERMANENT | NUD_NOARP))
26a41ae6 1190 return true;
d342894c 1191
1192 if (net_ratelimit())
1193 netdev_info(dev,
e4c7ed41 1194 "%pM migrated from %pIS to %pIS\n",
a4870f79 1195 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 1196
e4c7ed41 1197 rdst->remote_ip = *src_ip;
d342894c 1198 f->updated = jiffies;
9e4b93f9 1199 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 1200 } else {
1201 /* learned new entry */
1202 spin_lock(&vxlan->hash_lock);
3bf74b1a 1203
1204 /* close off race between vxlan_flush and incoming packets */
1205 if (netif_running(dev))
ff252e19 1206 vxlan_fdb_update(vxlan, src_mac, src_ip,
3bf74b1a 1207 NUD_REACHABLE,
1208 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 1209 vxlan->cfg.dst_port,
3ad7a4b1 1210 vni,
3bf74b1a 1211 vxlan->default_dst.remote_vni,
87613de9 1212 ifindex, NTF_SELF);
d342894c 1213 spin_unlock(&vxlan->hash_lock);
1214 }
26a41ae6 1215
1216 return false;
d342894c 1217}
1218
d342894c 1219/* See if multicast group is already in use by other ID */
95ab0991 1220static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1221{
553675fb 1222 struct vxlan_dev *vxlan;
c6fcc4fc 1223 struct vxlan_sock *sock4;
4053ab1b
AB		 1224#if IS_ENABLED(CONFIG_IPV6)
1225 struct vxlan_sock *sock6;
1226#endif
b1be00a6 1227 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1228
c6fcc4fc 1229 sock4 = rtnl_dereference(dev->vn4_sock);
1230
95ab0991
G		 1231 /* The vxlan_sock is only used by dev, leaving group has
1232 * no effect on other vxlan devices.
1233 */
66af846f 1234 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
95ab0991 1235 return false;
b1be00a6 1236#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1237 sock6 = rtnl_dereference(dev->vn6_sock);
66af846f 1238 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
b1be00a6
JB		 1239 return false;
1240#endif
95ab0991 1241
553675fb 1242 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1243 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1244 continue;
d342894c 1245
c6fcc4fc 1246 if (family == AF_INET &&
1247 rtnl_dereference(vxlan->vn4_sock) != sock4)
95ab0991 1248 continue;
b1be00a6 1249#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1250 if (family == AF_INET6 &&
1251 rtnl_dereference(vxlan->vn6_sock) != sock6)
b1be00a6
JB		 1252 continue;
1253#endif
95ab0991
G		 1254
1255 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1256 &dev->default_dst.remote_ip))
1257 continue;
1258
1259 if (vxlan->default_dst.remote_ifindex !=
1260 dev->default_dst.remote_ifindex)
1261 continue;
1262
1263 return true;
553675fb 1264 }
d342894c 1265
1266 return false;
1267}
1268
544a773a 1269static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
7c47cedf 1270{
b1be00a6 1271 struct vxlan_net *vn;
012a5729 1272
b1be00a6 1273 if (!vs)
544a773a 1274 return false;
66af846f 1275 if (!refcount_dec_and_test(&vs->refcnt))
544a773a 1276 return false;
d342894c 1277
b1be00a6 1278 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1279 spin_lock(&vn->sock_lock);
7c47cedf 1280 hlist_del_rcu(&vs->hlist);
e7b3db5e 1281 udp_tunnel_notify_del_rx_port(vs->sock,
b9adcd69
AD		 1282 (vs->flags & VXLAN_F_GPE) ?
1283 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 1284 UDP_TUNNEL_TYPE_VXLAN);
1c51a915
SH		 1285 spin_unlock(&vn->sock_lock);
1286
544a773a 1287 return true;
d342894c 1288}
1289
b1be00a6
JB		 1290static void vxlan_sock_release(struct vxlan_dev *vxlan)
1291{
c6fcc4fc 1292 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
b1be00a6 1293#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1294 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1295
57d88182 1296 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
544a773a
HFS		 1297#endif
1298
57d88182 1299 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
544a773a
HFS		 1300 synchronize_net();
1301
a53cb29b
MB		 1302 vxlan_vs_del_dev(vxlan);
1303
c6fcc4fc 1304 if (__vxlan_sock_release_prep(sock4)) {
1305 udp_tunnel_sock_release(sock4->sock);
1306 kfree(sock4);
544a773a
HFS		 1307 }
1308
1309#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1310 if (__vxlan_sock_release_prep(sock6)) {
1311 udp_tunnel_sock_release(sock6->sock);
1312 kfree(sock6);
544a773a 1313 }
b1be00a6
JB		 1314#endif
1315}
1316
56ef9c90 1317/* Update multicast group membership when first VNI on
c4b49512 1318 * multicast address is brought up
7c47cedf 1319 */
56ef9c90 1320static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1321{
b1be00a6 1322 struct sock *sk;
e4c7ed41
CW		 1323 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1324 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1325 int ret = -EINVAL;
d342894c 1326
e4c7ed41 1327 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1328 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1329 struct ip_mreqn mreq = {
1330 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1331 .imr_ifindex = ifindex,
1332 };
1333
c6fcc4fc 1334 sk = sock4->sock->sk;
b1be00a6 1335 lock_sock(sk);
56ef9c90 1336 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1337 release_sock(sk);
e4c7ed41
CW		 1338#if IS_ENABLED(CONFIG_IPV6)
1339 } else {
c6fcc4fc 1340 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1341
1342 sk = sock6->sock->sk;
b1be00a6 1343 lock_sock(sk);
56ef9c90
MRL		 1344 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1345 &ip->sin6.sin6_addr);
b1be00a6 1346 release_sock(sk);
e4c7ed41
CW		 1347#endif
1348 }
3fc2de2f 1349
56ef9c90 1350 return ret;
3fc2de2f 1351}
1352
1353/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1354static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1355{
b1be00a6 1356 struct sock *sk;
e4c7ed41
CW		 1357 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1358 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1359 int ret = -EINVAL;
3fc2de2f 1360
e4c7ed41 1361 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1362 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1363 struct ip_mreqn mreq = {
1364 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1365 .imr_ifindex = ifindex,
1366 };
1367
c6fcc4fc 1368 sk = sock4->sock->sk;
b1be00a6 1369 lock_sock(sk);
56ef9c90 1370 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1371 release_sock(sk);
e4c7ed41
CW		 1372#if IS_ENABLED(CONFIG_IPV6)
1373 } else {
c6fcc4fc 1374 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1375
1376 sk = sock6->sock->sk;
b1be00a6 1377 lock_sock(sk);
56ef9c90
MRL		 1378 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1379 &ip->sin6.sin6_addr);
b1be00a6 1380 release_sock(sk);
e4c7ed41
CW		 1381#endif
1382 }
d342894c 1383
56ef9c90 1384 return ret;
d342894c 1385}
1386
f14ecebb
JB		 1387static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1388 struct sk_buff *skb, u32 vxflags)
dfd8645e 1389{
7d34fa75 1390 size_t start, offset;
dfd8645e 1391
f14ecebb
JB		 1392 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1393 goto out;
b7fe10e5 1394
f14ecebb
JB		 1395 start = vxlan_rco_start(unparsed->vx_vni);
1396 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1397
7d34fa75 1398 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1399 return false;
dfd8645e 1400
be5cfeab
JB		 1401 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1402 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1403out:
1404 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1405 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1406 return true;
dfd8645e
TH		 1407}
1408
f14ecebb 1409static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1410 struct sk_buff *skb, u32 vxflags,
10a5af23 1411 struct vxlan_metadata *md)
3288af08 1412{
f14ecebb 1413 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1414 struct metadata_dst *tun_dst;
f14ecebb
JB		 1415
1416 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1417 goto out;
3288af08 1418
3288af08
JB		 1419 md->gbp = ntohs(gbp->policy_id);
1420
10a5af23 1421 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1422 if (tun_dst) {
3288af08 1423 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1424 tun_dst->u.tun_info.options_len = sizeof(*md);
1425 }
3288af08
JB		 1426 if (gbp->dont_learn)
1427 md->gbp |= VXLAN_GBP_DONT_LEARN;
1428
1429 if (gbp->policy_applied)
1430 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1431
64f87d36
JB		 1432 /* In flow-based mode, GBP is carried in dst_metadata */
1433 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1434 skb->mark = md->gbp;
f14ecebb
JB		 1435out:
1436 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1437}
1438
e1e5314d 1439static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
61618eea 1440 __be16 *protocol,
e1e5314d
JB		 1441 struct sk_buff *skb, u32 vxflags)
1442{
1443 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1444
1445 /* Need to have Next Protocol set for interfaces in GPE mode. */
1446 if (!gpe->np_applied)
1447 return false;
1448 /* "The initial version is 0. If a receiver does not support the
1449 * version indicated it MUST drop the packet.
1450 */
1451 if (gpe->version != 0)
1452 return false;
1453 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1454 * processing MUST occur." However, we don't implement OAM
1455 * processing, thus drop the packet.
1456 */
1457 if (gpe->oam_flag)
1458 return false;
1459
fa20e0e3
JB		 1460 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1461 if (!*protocol)
e1e5314d 1462 return false;
e1e5314d
JB		 1463
1464 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1465 return true;
1466}
1467
1ab016e2
JB		 1468static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1469 struct vxlan_sock *vs,
3ad7a4b1 1470 struct sk_buff *skb, __be32 vni)
614732ea 1471{
614732ea 1472 union vxlan_addr saddr;
87613de9 1473 u32 ifindex = skb->dev->ifindex;
614732ea 1474
614732ea 1475 skb_reset_mac_header(skb);
614732ea
TG		 1476 skb->protocol = eth_type_trans(skb, vxlan->dev);
1477 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1478
1479 /* Ignore packet loops (and multicast echo) */
1480 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1481 return false;
614732ea 1482
760c6805 1483 /* Get address from the outer IP header */
ce212d0f 1484 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1485 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1486 saddr.sa.sa_family = AF_INET;
1487#if IS_ENABLED(CONFIG_IPV6)
1488 } else {
1ab016e2 1489 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1490 saddr.sa.sa_family = AF_INET6;
1491#endif
1492 }
1493
dc5321d7 1494 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
87613de9 1495 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1ab016e2
JB		 1496 return false;
1497
1498 return true;
1499}
1500
760c6805
JB		 1501static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1502 struct sk_buff *skb)
1503{
1504 int err = 0;
1505
1506 if (vxlan_get_sk_family(vs) == AF_INET)
1507 err = IP_ECN_decapsulate(oiph, skb);
1508#if IS_ENABLED(CONFIG_IPV6)
1509 else
1510 err = IP6_ECN_decapsulate(oiph, skb);
1511#endif
1512
1513 if (unlikely(err) && log_ecn_error) {
1514 if (vxlan_get_sk_family(vs) == AF_INET)
1515 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1516 &((struct iphdr *)oiph)->saddr,
1517 ((struct iphdr *)oiph)->tos);
1518 else
1519 net_info_ratelimited("non-ECT from %pI6\n",
1520 &((struct ipv6hdr *)oiph)->saddr);
1521 }
1522 return err <= 1;
1523}
1524
d342894c 1525/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1526static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1527{
f2d1968e 1528 struct pcpu_sw_netstats *stats;
c9e78efb 1529 struct vxlan_dev *vxlan;
5cfccc5a 1530 struct vxlan_sock *vs;
f14ecebb 1531 struct vxlanhdr unparsed;
ee122c79
TG		 1532 struct vxlan_metadata _md;
1533 struct vxlan_metadata *md = &_md;
61618eea 1534 __be16 protocol = htons(ETH_P_TEB);
e1e5314d 1535 bool raw_proto = false;
f2d1968e 1536 void *oiph;
3ad7a4b1 1537 __be32 vni = 0;
d342894c 1538
e1e5314d 1539 /* Need UDP and VXLAN header to be present */
7ce04758 1540 if (!pskb_may_pull(skb, VXLAN_HLEN))
e5aed006 1541 goto drop;
d342894c 1542
f14ecebb 1543 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1544 /* VNI flag always required to be set */
1545 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1546 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1547 ntohl(vxlan_hdr(skb)->vx_flags),
1548 ntohl(vxlan_hdr(skb)->vx_vni));
1549 /* Return non vxlan pkt */
e5aed006 1550 goto drop;
d342894c 1551 }
288b01c8
JB		 1552 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1553 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1554
559835ea 1555 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1556 if (!vs)
d342894c 1557 goto drop;
d342894c 1558
3ad7a4b1
RP		 1559 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1560
49f810f0 1561 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
c9e78efb
JB		 1562 if (!vxlan)
1563 goto drop;
1564
e1e5314d
JB		 1565 /* For backwards compatibility, only allow reserved fields to be
1566 * used by VXLAN extensions if explicitly requested.
1567 */
1568 if (vs->flags & VXLAN_F_GPE) {
1569 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1570 goto drop;
1571 raw_proto = true;
1572 }
1573
1574 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1575 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1576 goto drop;
c9e78efb 1577
ee122c79 1578 if (vxlan_collect_metadata(vs)) {
10a5af23 1579 struct metadata_dst *tun_dst;
07dabf20 1580
c29a70d2 1581 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
d817f432 1582 key32_to_tunnel_id(vni), sizeof(*md));
c29a70d2 1583
ee122c79
TG		 1584 if (!tun_dst)
1585 goto drop;
1586
0f1b7354 1587 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 1588
1589 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 1590 } else {
1591 memset(md, 0, sizeof(*md));
1592 }
1593
f14ecebb
JB		 1594 if (vs->flags & VXLAN_F_REMCSUM_RX)
1595 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1596 goto drop;
1597 if (vs->flags & VXLAN_F_GBP)
10a5af23 1598 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB		 1599 /* Note that GBP and GPE can never be active together. This is
1600 * ensured in vxlan_dev_configure.
1601 */
3511494c 1602
f14ecebb 1603 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1604 /* If there are any unprocessed flags remaining treat
1605 * this as a malformed packet. This behavior diverges from
1606 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1607 * in reserved fields are to be ignored. The approach here
c4b49512 1608 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1609 * is more robust and provides a little more security in
1610 * adding extensions to VXLAN.
1611 */
288b01c8 1612 goto drop;
3bf39475
TH		 1613 }
1614
e1e5314d 1615 if (!raw_proto) {
3ad7a4b1 1616 if (!vxlan_set_mac(vxlan, vs, skb, vni))
e1e5314d
JB		 1617 goto drop;
1618 } else {
8be0cfa4 1619 skb_reset_mac_header(skb);
e1e5314d
JB		 1620 skb->dev = vxlan->dev;
1621 skb->pkt_type = PACKET_HOST;
1622 }
f2d1968e 1623
f2d1968e
JB		 1624 oiph = skb_network_header(skb);
1625 skb_reset_network_header(skb);
1626
1627 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1628 ++vxlan->dev->stats.rx_frame_errors;
1629 ++vxlan->dev->stats.rx_errors;
1630 goto drop;
1631 }
1632
6d7f97e7
ED		 1633 rcu_read_lock();
1634
1635 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1636 rcu_read_unlock();
1637 atomic_long_inc(&vxlan->dev->rx_dropped);
1638 goto drop;
1639 }
1640
f2d1968e
JB		 1641 stats = this_cpu_ptr(vxlan->dev->tstats);
1642 u64_stats_update_begin(&stats->syncp);
1643 stats->rx_packets++;
1644 stats->rx_bytes += skb->len;
1645 u64_stats_update_end(&stats->syncp);
1646
1647 gro_cells_receive(&vxlan->gro_cells, skb);
6d7f97e7
ED		 1648
1649 rcu_read_unlock();
1650
5cfccc5a
PS		 1651 return 0;
1652
1653drop:
288b01c8
JB		 1654 /* Consume bad packet */
1655 kfree_skb(skb);
1656 return 0;
5cfccc5a
PS		 1657}
1658
3ad7a4b1 1659static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
e4f67add
DS		 1660{
1661 struct vxlan_dev *vxlan = netdev_priv(dev);
1662 struct arphdr *parp;
1663 u8 *arpptr, *sha;
1664 __be32 sip, tip;
1665 struct neighbour *n;
1666
1667 if (dev->flags & IFF_NOARP)
1668 goto out;
1669
1670 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1671 dev->stats.tx_dropped++;
1672 goto out;
1673 }
1674 parp = arp_hdr(skb);
1675
1676 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1677 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1678 parp->ar_pro != htons(ETH_P_IP) ||
1679 parp->ar_op != htons(ARPOP_REQUEST) ||
1680 parp->ar_hln != dev->addr_len ||
1681 parp->ar_pln != 4)
1682 goto out;
1683 arpptr = (u8 *)parp + sizeof(struct arphdr);
1684 sha = arpptr;
1685 arpptr += dev->addr_len; /* sha */
1686 memcpy(&sip, arpptr, sizeof(sip));
1687 arpptr += sizeof(sip);
1688 arpptr += dev->addr_len; /* tha */
1689 memcpy(&tip, arpptr, sizeof(tip));
1690
1691 if (ipv4_is_loopback(tip) ||
1692 ipv4_is_multicast(tip))
1693 goto out;
1694
1695 n = neigh_lookup(&arp_tbl, &tip, dev);
1696
1697 if (n) {
e4f67add
DS		 1698 struct vxlan_fdb *f;
1699 struct sk_buff *reply;
1700
1701 if (!(n->nud_state & NUD_CONNECTED)) {
1702 neigh_release(n);
1703 goto out;
1704 }
1705
3ad7a4b1 1706 f = vxlan_find_mac(vxlan, n->ha, vni);
e4c7ed41 1707 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1708 /* bridge-local neighbor */
1709 neigh_release(n);
1710 goto out;
1711 }
1712
1713 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1714 n->ha, sha);
1715
1716 neigh_release(n);
1717
7346135d
DS		 1718 if (reply == NULL)
1719 goto out;
1720
e4f67add
DS		 1721 skb_reset_mac_header(reply);
1722 __skb_pull(reply, skb_network_offset(reply));
1723 reply->ip_summed = CHECKSUM_UNNECESSARY;
1724 reply->pkt_type = PACKET_HOST;
1725
1726 if (netif_rx_ni(reply) == NET_RX_DROP)
1727 dev->stats.rx_dropped++;
dc5321d7 1728 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
e4c7ed41
CW		 1729 union vxlan_addr ipa = {
1730 .sin.sin_addr.s_addr = tip,
a45e92a5 1731 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1732 };
1733
1734 vxlan_ip_miss(dev, &ipa);
1735 }
e4f67add
DS		 1736out:
1737 consume_skb(skb);
1738 return NETDEV_TX_OK;
1739}
1740
f564f45c 1741#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1742static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1743 struct neighbour *n, bool isrouter)
1744{
1745 struct net_device *dev = request->dev;
1746 struct sk_buff *reply;
1747 struct nd_msg *ns, *na;
1748 struct ipv6hdr *pip6;
1749 u8 *daddr;
1750 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1751 int ns_olen;
1752 int i, len;
1753
f1fb08f6 1754 if (dev == NULL || !pskb_may_pull(request, request->len))
4b29dba9
DS		 1755 return NULL;
1756
1757 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1758 sizeof(*na) + na_olen + dev->needed_tailroom;
1759 reply = alloc_skb(len, GFP_ATOMIC);
1760 if (reply == NULL)
1761 return NULL;
1762
1763 reply->protocol = htons(ETH_P_IPV6);
1764 reply->dev = dev;
1765 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1766 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1767 skb_reset_mac_header(reply);
4b29dba9 1768
f1fb08f6 1769 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
4b29dba9
DS		 1770
1771 daddr = eth_hdr(request)->h_source;
f1fb08f6
VB		 1772 ns_olen = request->len - skb_network_offset(request) -
1773 sizeof(struct ipv6hdr) - sizeof(*ns);
4b29dba9
DS		 1774 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1775 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1776 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1777 break;
1778 }
1779 }
1780
1781 /* Ethernet header */
1782 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1783 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1784 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1785 reply->protocol = htons(ETH_P_IPV6);
1786
1787 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1788 skb_reset_network_header(reply);
4b29dba9
DS		 1789 skb_put(reply, sizeof(struct ipv6hdr));
1790
1791 /* IPv6 header */
1792
1793 pip6 = ipv6_hdr(reply);
1794 memset(pip6, 0, sizeof(struct ipv6hdr));
1795 pip6->version = 6;
1796 pip6->priority = ipv6_hdr(request)->priority;
1797 pip6->nexthdr = IPPROTO_ICMPV6;
1798 pip6->hop_limit = 255;
1799 pip6->daddr = ipv6_hdr(request)->saddr;
1800 pip6->saddr = *(struct in6_addr *)n->primary_key;
1801
1802 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1803 skb_reset_transport_header(reply);
4b29dba9 1804
4b29dba9 1805 /* Neighbor Advertisement */
b080db58 1806 na = skb_put_zero(reply, sizeof(*na) + na_olen);
4b29dba9
DS		 1807 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1808 na->icmph.icmp6_router = isrouter;
1809 na->icmph.icmp6_override = 1;
1810 na->icmph.icmp6_solicited = 1;
1811 na->target = ns->target;
1812 ether_addr_copy(&na->opt[2], n->ha);
1813 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1814 na->opt[1] = na_olen >> 3;
1815
1816 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1817 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1818 csum_partial(na, sizeof(*na)+na_olen, 0));
1819
1820 pip6->payload_len = htons(sizeof(*na)+na_olen);
1821
1822 skb_push(reply, sizeof(struct ipv6hdr));
1823
1824 reply->ip_summed = CHECKSUM_UNNECESSARY;
1825
1826 return reply;
1827}
1828
3ad7a4b1 1829static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
f564f45c
CW		 1830{
1831 struct vxlan_dev *vxlan = netdev_priv(dev);
8dcd81a9 1832 const struct in6_addr *daddr;
8bff3685 1833 const struct ipv6hdr *iphdr;
4b29dba9 1834 struct inet6_dev *in6_dev;
8bff3685
XL		 1835 struct neighbour *n;
1836 struct nd_msg *msg;
f564f45c
CW		 1837
1838 in6_dev = __in6_dev_get(dev);
1839 if (!in6_dev)
1840 goto out;
1841
f564f45c 1842 iphdr = ipv6_hdr(skb);
f564f45c 1843 daddr = &iphdr->daddr;
f1fb08f6 1844 msg = (struct nd_msg *)(iphdr + 1);
f564f45c 1845
4b29dba9
DS		 1846 if (ipv6_addr_loopback(daddr) ||
1847 ipv6_addr_is_multicast(&msg->target))
1848 goto out;
1849
1850 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1851
1852 if (n) {
1853 struct vxlan_fdb *f;
4b29dba9 1854 struct sk_buff *reply;
f564f45c
CW		 1855
1856 if (!(n->nud_state & NUD_CONNECTED)) {
1857 neigh_release(n);
1858 goto out;
1859 }
1860
3ad7a4b1 1861 f = vxlan_find_mac(vxlan, n->ha, vni);
f564f45c
CW		 1862 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1863 /* bridge-local neighbor */
1864 neigh_release(n);
1865 goto out;
1866 }
1867
4b29dba9
DS		 1868 reply = vxlan_na_create(skb, n,
1869 !!(f ? f->flags & NTF_ROUTER : 0));
1870
f564f45c 1871 neigh_release(n);
4b29dba9
DS		 1872
1873 if (reply == NULL)
1874 goto out;
1875
1876 if (netif_rx_ni(reply) == NET_RX_DROP)
1877 dev->stats.rx_dropped++;
1878
dc5321d7 1879 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1880 union vxlan_addr ipa = {
1881 .sin6.sin6_addr = msg->target,
a45e92a5 1882 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1883 };
1884
f564f45c
CW		 1885 vxlan_ip_miss(dev, &ipa);
1886 }
1887
1888out:
1889 consume_skb(skb);
1890 return NETDEV_TX_OK;
1891}
1892#endif
1893
e4f67add
DS		 1894static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1895{
1896 struct vxlan_dev *vxlan = netdev_priv(dev);
1897 struct neighbour *n;
e4f67add
DS		 1898
1899 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1900 return false;
1901
1902 n = NULL;
1903 switch (ntohs(eth_hdr(skb)->h_proto)) {
1904 case ETH_P_IP:
e15a00aa
CW		 1905 {
1906 struct iphdr *pip;
1907
e4f67add
DS		 1908 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1909 return false;
1910 pip = ip_hdr(skb);
1911 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
dc5321d7 1912 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e4c7ed41
CW		 1913 union vxlan_addr ipa = {
1914 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1915 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1916 };
1917
1918 vxlan_ip_miss(dev, &ipa);
1919 return false;
1920 }
1921
e4f67add 1922 break;
e15a00aa
CW		 1923 }
1924#if IS_ENABLED(CONFIG_IPV6)
1925 case ETH_P_IPV6:
1926 {
1927 struct ipv6hdr *pip6;
1928
1929 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1930 return false;
1931 pip6 = ipv6_hdr(skb);
1932 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
dc5321d7 1933 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e15a00aa
CW		 1934 union vxlan_addr ipa = {
1935 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1936 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1937 };
1938
1939 vxlan_ip_miss(dev, &ipa);
1940 return false;
1941 }
1942
1943 break;
1944 }
1945#endif
e4f67add
DS		 1946 default:
1947 return false;
1948 }
1949
1950 if (n) {
1951 bool diff;
1952
7367d0b5 1953 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1954 if (diff) {
1955 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1956 dev->addr_len);
1957 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1958 }
1959 neigh_release(n);
1960 return diff;
e4c7ed41
CW		 1961 }
1962
e4f67add
DS		 1963 return false;
1964}
1965
af33c1ad 1966static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1967 struct vxlan_metadata *md)
1968{
1969 struct vxlanhdr_gbp *gbp;
1970
db79a621
TG		 1971 if (!md->gbp)
1972 return;
1973
3511494c 1974 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1975 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 1976
1977 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1978 gbp->dont_learn = 1;
1979
1980 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1981 gbp->policy_applied = 1;
1982
1983 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1984}
1985
e1e5314d
JB		 1986static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1987 __be16 protocol)
1988{
1989 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1990
1991 gpe->np_applied = 1;
fa20e0e3
JB		 1992 gpe->next_protocol = tun_p_from_eth_p(protocol);
1993 if (!gpe->next_protocol)
1994 return -EPFNOSUPPORT;
1995 return 0;
e1e5314d
JB		 1996}
1997
f491e56d
JB		 1998static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1999 int iphdr_len, __be32 vni,
2000 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 2001 bool udp_sum)
e4c7ed41 2002{
e4c7ed41 2003 struct vxlanhdr *vxh;
e4c7ed41
CW		 2004 int min_headroom;
2005 int err;
dfd8645e 2006 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 2007 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 2008
af33c1ad 2009 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 2010 skb->ip_summed == CHECKSUM_PARTIAL) {
2011 int csum_start = skb_checksum_start_offset(skb);
2012
2013 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2014 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2015 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 2016 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 2017 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 2018 }
e4c7ed41 2019
e4c7ed41 2020 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
4a4f86cc 2021 + VXLAN_HLEN + iphdr_len;
649c5b8b
PS		 2022
2023 /* Need space for new headers (invalidates iph ptr) */
2024 err = skb_cow_head(skb, min_headroom);
e1e5314d 2025 if (unlikely(err))
c46b7897 2026 return err;
649c5b8b 2027
aed069df
AD		 2028 err = iptunnel_handle_offloads(skb, type);
2029 if (err)
c46b7897 2030 return err;
b736a623 2031
d58ff351 2032 vxh = __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 2033 vxh->vx_flags = VXLAN_HF_VNI;
2034 vxh->vx_vni = vxlan_vni_field(vni);
49560532 2035
dfd8645e 2036 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 2037 unsigned int start;
dfd8645e 2038
54bfd872
JB		 2039 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2040 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2041 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 2042
2043 if (!skb_is_gso(skb)) {
2044 skb->ip_summed = CHECKSUM_NONE;
2045 skb->encapsulation = 0;
2046 }
2047 }
2048
af33c1ad
TH		 2049 if (vxflags & VXLAN_F_GBP)
2050 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB		 2051 if (vxflags & VXLAN_F_GPE) {
2052 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2053 if (err < 0)
c46b7897 2054 return err;
e1e5314d
JB		 2055 inner_protocol = skb->protocol;
2056 }
3511494c 2057
e1e5314d 2058 skb_set_inner_protocol(skb, inner_protocol);
039f5062 2059 return 0;
49560532 2060}
49560532 2061
655c3de1 2062static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2063 struct vxlan_sock *sock4,
1a8496ba 2064 struct sk_buff *skb, int oif, u8 tos,
4ecb1d83 2065 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
0c1d70af 2066 struct dst_cache *dst_cache,
db3c6139 2067 const struct ip_tunnel_info *info)
1a8496ba 2068{
db3c6139 2069 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 2070 struct rtable *rt = NULL;
2071 struct flowi4 fl4;
2072
655c3de1 2073 if (!sock4)
2074 return ERR_PTR(-EIO);
2075
db3c6139
DB		 2076 if (tos && !info)
2077 use_cache = false;
2078 if (use_cache) {
0c1d70af
PA		 2079 rt = dst_cache_get_ip4(dst_cache, saddr);
2080 if (rt)
2081 return rt;
2082 }
2083
1a8496ba
JB		 2084 memset(&fl4, 0, sizeof(fl4));
2085 fl4.flowi4_oif = oif;
2086 fl4.flowi4_tos = RT_TOS(tos);
2087 fl4.flowi4_mark = skb->mark;
2088 fl4.flowi4_proto = IPPROTO_UDP;
2089 fl4.daddr = daddr;
272d96a5 2090 fl4.saddr = *saddr;
4ecb1d83
MP		 2091 fl4.fl4_dport = dport;
2092 fl4.fl4_sport = sport;
1a8496ba
JB		 2093
2094 rt = ip_route_output_key(vxlan->net, &fl4);
655c3de1 2095 if (likely(!IS_ERR(rt))) {
2096 if (rt->dst.dev == dev) {
2097 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2098 ip_rt_put(rt);
2099 return ERR_PTR(-ELOOP);
2100 }
2101
1a8496ba 2102 *saddr = fl4.saddr;
0c1d70af
PA		 2103 if (use_cache)
2104 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
655c3de1 2105 } else {
2106 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2107 return ERR_PTR(-ENETUNREACH);
0c1d70af 2108 }
1a8496ba
JB		 2109 return rt;
2110}
2111
e5d4b29f
JB		 2112#if IS_ENABLED(CONFIG_IPV6)
2113static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
655c3de1 2114 struct net_device *dev,
03dc52a8 2115 struct vxlan_sock *sock6,
1400615d 2116 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 2117 __be32 label,
e5d4b29f 2118 const struct in6_addr *daddr,
0c1d70af 2119 struct in6_addr *saddr,
4ecb1d83 2120 __be16 dport, __be16 sport,
db3c6139
DB		 2121 struct dst_cache *dst_cache,
2122 const struct ip_tunnel_info *info)
e5d4b29f 2123{
db3c6139 2124 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 2125 struct dst_entry *ndst;
2126 struct flowi6 fl6;
2127 int err;
2128
c6fcc4fc 2129 if (!sock6)
2130 return ERR_PTR(-EIO);
2131
1400615d
DB		 2132 if (tos && !info)
2133 use_cache = false;
db3c6139 2134 if (use_cache) {
0c1d70af
PA		 2135 ndst = dst_cache_get_ip6(dst_cache, saddr);
2136 if (ndst)
2137 return ndst;
2138 }
2139
e5d4b29f
JB		 2140 memset(&fl6, 0, sizeof(fl6));
2141 fl6.flowi6_oif = oif;
2142 fl6.daddr = *daddr;
272d96a5 2143 fl6.saddr = *saddr;
eaa93bf4 2144 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB		 2145 fl6.flowi6_mark = skb->mark;
2146 fl6.flowi6_proto = IPPROTO_UDP;
4ecb1d83
MP		 2147 fl6.fl6_dport = dport;
2148 fl6.fl6_sport = sport;
e5d4b29f
JB		 2149
2150 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
c6fcc4fc 2151 sock6->sock->sk,
e5d4b29f 2152 &ndst, &fl6);
655c3de1 2153 if (unlikely(err < 0)) {
2154 netdev_dbg(dev, "no route to %pI6\n", daddr);
2155 return ERR_PTR(-ENETUNREACH);
2156 }
2157
2158 if (unlikely(ndst->dev == dev)) {
2159 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2160 dst_release(ndst);
2161 return ERR_PTR(-ELOOP);
2162 }
e5d4b29f
JB		 2163
2164 *saddr = fl6.saddr;
db3c6139 2165 if (use_cache)
0c1d70af 2166 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 2167 return ndst;
2168}
2169#endif
2170
9dcc71e1
SS		 2171/* Bypass encapsulation if the destination is local */
2172static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
3ad7a4b1 2173 struct vxlan_dev *dst_vxlan, __be32 vni)
9dcc71e1 2174{
8f84985f 2175 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 2176 union vxlan_addr loopback;
2177 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
f059ea56 2178 struct net_device *dev;
ce6502a8 2179 int len = skb->len;
9dcc71e1 2180
8f84985f
LR		 2181 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2182 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 2183 skb->pkt_type = PACKET_HOST;
2184 skb->encapsulation = 0;
2185 skb->dev = dst_vxlan->dev;
2186 __skb_pull(skb, skb_network_offset(skb));
2187
e4c7ed41
CW		 2188 if (remote_ip->sa.sa_family == AF_INET) {
2189 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2190 loopback.sa.sa_family = AF_INET;
2191#if IS_ENABLED(CONFIG_IPV6)
2192 } else {
2193 loopback.sin6.sin6_addr = in6addr_loopback;
2194 loopback.sa.sa_family = AF_INET6;
2195#endif
2196 }
2197
f059ea56
ED		 2198 rcu_read_lock();
2199 dev = skb->dev;
2200 if (unlikely(!(dev->flags & IFF_UP))) {
2201 kfree_skb(skb);
2202 goto drop;
2203 }
2204
dc5321d7 2205 if (dst_vxlan->cfg.flags & VXLAN_F_LEARN)
f059ea56 2206 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
9dcc71e1
SS		 2207
2208 u64_stats_update_begin(&tx_stats->syncp);
2209 tx_stats->tx_packets++;
ce6502a8 2210 tx_stats->tx_bytes += len;
9dcc71e1
SS		 2211 u64_stats_update_end(&tx_stats->syncp);
2212
2213 if (netif_rx(skb) == NET_RX_SUCCESS) {
2214 u64_stats_update_begin(&rx_stats->syncp);
2215 rx_stats->rx_packets++;
ce6502a8 2216 rx_stats->rx_bytes += len;
9dcc71e1
SS		 2217 u64_stats_update_end(&rx_stats->syncp);
2218 } else {
f059ea56 2219drop:
ce6502a8 2220 dev->stats.rx_dropped++;
9dcc71e1 2221 }
f059ea56 2222 rcu_read_unlock();
9dcc71e1
SS		 2223}
2224
fee1fad7 2225static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
49f810f0
MS		 2226 struct vxlan_dev *vxlan,
2227 union vxlan_addr *daddr,
2228 __be16 dst_port, int dst_ifindex, __be32 vni,
2229 struct dst_entry *dst,
fee1fad7 2230 u32 rt_flags)
2231{
2232#if IS_ENABLED(CONFIG_IPV6)
2233 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2234 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2235 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2236 */
2237 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2238#endif
2239 /* Bypass encapsulation if the destination is local */
2240 if (rt_flags & RTCF_LOCAL &&
2241 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2242 struct vxlan_dev *dst_vxlan;
2243
2244 dst_release(dst);
49f810f0 2245 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
fee1fad7 2246 daddr->sa.sa_family, dst_port,
dc5321d7 2247 vxlan->cfg.flags);
fee1fad7 2248 if (!dst_vxlan) {
2249 dev->stats.tx_errors++;
2250 kfree_skb(skb);
2251
2252 return -ENOENT;
2253 }
3ad7a4b1 2254 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
fee1fad7 2255 return 1;
2256 }
2257
2258 return 0;
2259}
2260
4ad16930 2261static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
3ad7a4b1
RP		 2262 __be32 default_vni, struct vxlan_rdst *rdst,
2263 bool did_rsc)
d342894c 2264{
d71785ff 2265 struct dst_cache *dst_cache;
3093fbe7 2266 struct ip_tunnel_info *info;
d342894c 2267 struct vxlan_dev *vxlan = netdev_priv(dev);
0770b53b 2268 const struct iphdr *old_iph = ip_hdr(skb);
e4c7ed41 2269 union vxlan_addr *dst;
272d96a5 2270 union vxlan_addr remote_ip, local_ip;
ee122c79
TG		 2271 struct vxlan_metadata _md;
2272 struct vxlan_metadata *md = &_md;
e4c7ed41 2273 __be16 src_port = 0, dst_port;
655c3de1 2274 struct dst_entry *ndst = NULL;
e7f70af1 2275 __be32 vni, label;
d342894c 2276 __u8 tos, ttl;
49f810f0 2277 int ifindex;
0e6fbc5b 2278 int err;
dc5321d7 2279 u32 flags = vxlan->cfg.flags;
b4ed5cad 2280 bool udp_sum = false;
f491e56d 2281 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 2282
61adedf3 2283 info = skb_tunnel_info(skb);
3093fbe7 2284
ee122c79 2285 if (rdst) {
0770b53b 2286 dst = &rdst->remote_ip;
2287 if (vxlan_addr_any(dst)) {
2288 if (did_rsc) {
2289 /* short-circuited back to local bridge */
3ad7a4b1 2290 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
0770b53b 2291 return;
2292 }
2293 goto drop;
2294 }
2295
0dfbdf41 2296 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
3ad7a4b1 2297 vni = (rdst->remote_vni) ? : default_vni;
49f810f0 2298 ifindex = rdst->remote_ifindex;
1158632b 2299 local_ip = vxlan->cfg.saddr;
d71785ff 2300 dst_cache = &rdst->dst_cache;
0770b53b 2301 md->gbp = skb->mark;
2302 ttl = vxlan->cfg.ttl;
2303 if (!ttl && vxlan_addr_multicast(dst))
2304 ttl = 1;
2305
2306 tos = vxlan->cfg.tos;
2307 if (tos == 1)
2308 tos = ip_tunnel_get_dsfield(old_iph, skb);
2309
2310 if (dst->sa.sa_family == AF_INET)
2311 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2312 else
2313 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2314 label = vxlan->cfg.label;
ee122c79
TG		 2315 } else {
2316 if (!info) {
2317 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2318 dev->name);
2319 goto drop;
2320 }
b1be00a6 2321 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
272d96a5 2322 if (remote_ip.sa.sa_family == AF_INET) {
a725e514 2323 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
272d96a5 2324 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2325 } else {
a725e514 2326 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
272d96a5 2327 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2328 }
ee122c79 2329 dst = &remote_ip;
0770b53b 2330 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2331 vni = tunnel_id_to_key32(info->key.tun_id);
49f810f0 2332 ifindex = 0;
d71785ff 2333 dst_cache = &info->dst_cache;
eb19850e
XL		 2334 if (info->options_len) {
2335 if (info->options_len < sizeof(*md))
2336 goto drop;
0770b53b 2337 md = ip_tunnel_info_opts(info);
eb19850e 2338 }
a725e514
JB		 2339 ttl = info->key.ttl;
2340 tos = info->key.tos;
e7f70af1 2341 label = info->key.label;
b4ed5cad 2342 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514 2343 }
0770b53b 2344 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2345 vxlan->cfg.port_max, true);
a725e514 2346
56de859e 2347 rcu_read_lock();
e4c7ed41 2348 if (dst->sa.sa_family == AF_INET) {
c6fcc4fc 2349 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
c46b7897 2350 struct rtable *rt;
0770b53b 2351 __be16 df = 0;
c6fcc4fc 2352
49f810f0 2353 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
272d96a5 2354 dst->sin.sin_addr.s_addr,
1158632b 2355 &local_ip.sin.sin_addr.s_addr,
4ecb1d83 2356 dst_port, src_port,
d71785ff 2357 dst_cache, info);
8ebd115b
DM		 2358 if (IS_ERR(rt)) {
2359 err = PTR_ERR(rt);
c46b7897 2360 goto tx_error;
8ebd115b 2361 }
e4c7ed41 2362
d57420a1
JV		 2363 if (fan_has_map(&vxlan->fan) && rt->rt_flags & RTCF_LOCAL) {
2364 netdev_dbg(dev, "discard fan to localhost %pI4\n",
2365 &dst->sin.sin_addr.s_addr);
2366 ip_rt_put(rt);
2367 goto tx_free;
2368 }
2369
e4c7ed41 2370 /* Bypass encapsulation if the destination is local */
fee1fad7 2371 if (!info) {
2372 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2373 dst_port, ifindex, vni,
2374 &rt->dst, rt->rt_flags);
fee1fad7 2375 if (err)
56de859e 2376 goto out_unlock;
fee1fad7 2377 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
6ceb31ca 2378 df = htons(IP_DF);
fee1fad7 2379 }
6ceb31ca 2380
c46b7897 2381 ndst = &rt->dst;
a93bf0ff
XL		 2382 if (skb_dst(skb)) {
2383 int mtu = dst_mtu(ndst) - VXLAN_HEADROOM;
2384
f15ca723 2385 skb_dst_update_pmtu(skb, mtu);
a93bf0ff
XL		 2386 }
2387
3e3b50a3 2388 tos = ip_tunnel_ecn_encap(RT_TOS(tos), old_iph, skb);
e4c7ed41 2389 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
c46b7897 2390 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
54bfd872 2391 vni, md, flags, udp_sum);
f491e56d 2392 if (err < 0)
c46b7897 2393 goto tx_error;
f491e56d 2394
1158632b 2395 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
f491e56d
JB		 2396 dst->sin.sin_addr.s_addr, tos, ttl, df,
2397 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2398#if IS_ENABLED(CONFIG_IPV6)
2399 } else {
c6fcc4fc 2400 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
e4c7ed41 2401
49f810f0 2402 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
272d96a5 2403 label, &dst->sin6.sin6_addr,
1158632b 2404 &local_ip.sin6.sin6_addr,
4ecb1d83 2405 dst_port, src_port,
db3c6139 2406 dst_cache, info);
e5d4b29f 2407 if (IS_ERR(ndst)) {
8ebd115b 2408 err = PTR_ERR(ndst);
c46b7897 2409 ndst = NULL;
9dcc71e1 2410 goto tx_error;
e4c7ed41 2411 }
655c3de1 2412
fee1fad7 2413 if (!info) {
2414 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
49560532 2415
fee1fad7 2416 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2417 dst_port, ifindex, vni,
2418 ndst, rt6i_flags);
fee1fad7 2419 if (err)
56de859e 2420 goto out_unlock;
fee1fad7 2421 }
35e2d115 2422
a93bf0ff
XL		 2423 if (skb_dst(skb)) {
2424 int mtu = dst_mtu(ndst) - VXLAN6_HEADROOM;
2425
f15ca723 2426 skb_dst_update_pmtu(skb, mtu);
a93bf0ff
XL		 2427 }
2428
3e3b50a3 2429 tos = ip_tunnel_ecn_encap(RT_TOS(tos), old_iph, skb);
e4c7ed41 2430 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2431 skb_scrub_packet(skb, xnet);
2432 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2433 vni, md, flags, udp_sum);
c46b7897 2434 if (err < 0)
2435 goto tx_error;
2436
0770b53b 2437 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
1158632b 2438 &local_ip.sin6.sin6_addr,
272d96a5 2439 &dst->sin6.sin6_addr, tos, ttl,
e7f70af1 2440 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2441#endif
2442 }
56de859e
JK		 2443out_unlock:
2444 rcu_read_unlock();
4ad16930 2445 return;
d342894c 2446
2447drop:
2448 dev->stats.tx_dropped++;
c46b7897 2449 dev_kfree_skb(skb);
2450 return;
d342894c 2451
2452tx_error:
56de859e 2453 rcu_read_unlock();
655c3de1 2454 if (err == -ELOOP)
2455 dev->stats.collisions++;
2456 else if (err == -ENETUNREACH)
2457 dev->stats.tx_carrier_errors++;
c46b7897 2458 dst_release(ndst);
d342894c 2459 dev->stats.tx_errors++;
d57420a1 2460tx_free:
c46b7897 2461 kfree_skb(skb);
d342894c 2462}
2463
6681712d
DS		 2464/* Transmit local packets over Vxlan
2465 *
2466 * Outer IP header inherits ECN and DF from inner header.
2467 * Outer UDP destination is the VXLAN assigned port.
2468 * source port is based on hash of flow
2469 */
2470static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2471{
2472 struct vxlan_dev *vxlan = netdev_priv(dev);
8bff3685 2473 struct vxlan_rdst *rdst, *fdst = NULL;
3093fbe7 2474 const struct ip_tunnel_info *info;
6681712d 2475 bool did_rsc = false;
6681712d 2476 struct vxlan_fdb *f;
8bff3685 2477 struct ethhdr *eth;
3ad7a4b1 2478 __be32 vni = 0;
6681712d 2479
61adedf3 2480 info = skb_tunnel_info(skb);
3093fbe7 2481
6681712d 2482 skb_reset_mac_header(skb);
6681712d 2483
dc5321d7 2484 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 2485 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2486 info->mode & IP_TUNNEL_INFO_TX) {
2487 vni = tunnel_id_to_key32(info->key.tun_id);
2488 } else {
2489 if (info && info->mode & IP_TUNNEL_INFO_TX)
2490 vxlan_xmit_one(skb, dev, vni, NULL, false);
2491 else
2492 kfree_skb(skb);
2493 return NETDEV_TX_OK;
2494 }
47e5d1b0
JB		 2495 }
2496
dc5321d7 2497 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
47e5d1b0 2498 eth = eth_hdr(skb);
f564f45c 2499 if (ntohs(eth->h_proto) == ETH_P_ARP)
3ad7a4b1 2500 return arp_reduce(dev, skb, vni);
f564f45c 2501#if IS_ENABLED(CONFIG_IPV6)
8bff3685
XL		 2502 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2503 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2504 sizeof(struct nd_msg)) &&
2505 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2506 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2507
2508 if (m->icmph.icmp6_code == 0 &&
2509 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
f1fb08f6 2510 return neigh_reduce(dev, skb, vni);
f564f45c
CW		 2511 }
2512#endif
2513 }
6681712d 2514
d57420a1
JV		 2515 if (fan_has_map(&vxlan->fan)) {
2516 struct vxlan_rdst fan_rdst;
2517
2518 netdev_dbg(vxlan->dev, "vxlan_xmit p %x d %pM\n",
2519 eth->h_proto, eth->h_dest);
2520 if (vxlan_fan_build_rdst(vxlan, skb, &fan_rdst)) {
2521 dev->stats.tx_dropped++;
2522 kfree_skb(skb);
2523 return NETDEV_TX_OK;
2524 }
2525 vxlan_xmit_one(skb, dev, vni, &fan_rdst, 0);
2526 return NETDEV_TX_OK;
2527 }
2528
47e5d1b0 2529 eth = eth_hdr(skb);
3ad7a4b1 2530 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2531 did_rsc = false;
2532
dc5321d7 2533 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2534 (ntohs(eth->h_proto) == ETH_P_IP ||
2535 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2536 did_rsc = route_shortcircuit(dev, skb);
2537 if (did_rsc)
3ad7a4b1 2538 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2539 }
2540
6681712d 2541 if (f == NULL) {
3ad7a4b1 2542 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 2543 if (f == NULL) {
dc5321d7 2544 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
afbd8bae
MR		 2545 !is_multicast_ether_addr(eth->h_dest))
2546 vxlan_fdb_miss(vxlan, eth->h_dest);
2547
2548 dev->stats.tx_dropped++;
8f646c92 2549 kfree_skb(skb);
afbd8bae
MR		 2550 return NETDEV_TX_OK;
2551 }
2552 }
6681712d 2553
afbd8bae
MR		 2554 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2555 struct sk_buff *skb1;
6681712d 2556
8f646c92
ED		 2557 if (!fdst) {
2558 fdst = rdst;
2559 continue;
2560 }
afbd8bae
MR		 2561 skb1 = skb_clone(skb, GFP_ATOMIC);
2562 if (skb1)
3ad7a4b1 2563 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
6681712d
DS		 2564 }
2565
8f646c92 2566 if (fdst)
3ad7a4b1 2567 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
8f646c92
ED		 2568 else
2569 kfree_skb(skb);
4ad16930 2570 return NETDEV_TX_OK;
6681712d
DS		 2571}
2572
d342894c 2573/* Walk the forwarding table and purge stale entries */
df7e828c 2574static void vxlan_cleanup(struct timer_list *t)
d342894c 2575{
df7e828c 2576 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
d342894c 2577 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2578 unsigned int h;
2579
2580 if (!netif_running(vxlan->dev))
2581 return;
2582
d342894c 2583 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2584 struct hlist_node *p, *n;
14e1d0fa
SD		 2585
2586 spin_lock_bh(&vxlan->hash_lock);
d342894c 2587 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2588 struct vxlan_fdb *f
2589 = container_of(p, struct vxlan_fdb, hlist);
2590 unsigned long timeout;
2591
efb5f68f 2592 if (f->state & (NUD_PERMANENT | NUD_NOARP))
d342894c 2593 continue;
2594
def499c9
RP		 2595 if (f->flags & NTF_EXT_LEARNED)
2596 continue;
2597
0dfbdf41 2598 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2599 if (time_before_eq(timeout, jiffies)) {
2600 netdev_dbg(vxlan->dev,
2601 "garbage collect %pM\n",
2602 f->eth_addr);
2603 f->state = NUD_STALE;
d7e686dc 2604 vxlan_fdb_destroy(vxlan, f, true);
d342894c 2605 } else if (time_before(timeout, next_timer))
2606 next_timer = timeout;
2607 }
14e1d0fa 2608 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2609 }
d342894c 2610
2611 mod_timer(&vxlan->age_timer, next_timer);
2612}
2613
a53cb29b
MB		 2614static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2615{
2616 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2617
2618 spin_lock(&vn->sock_lock);
69e76661
JB		 2619 hlist_del_init_rcu(&vxlan->hlist4.hlist);
2620#if IS_ENABLED(CONFIG_IPV6)
2621 hlist_del_init_rcu(&vxlan->hlist6.hlist);
2622#endif
a53cb29b
MB		 2623 spin_unlock(&vn->sock_lock);
2624}
2625
69e76661
JB		 2626static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2627 struct vxlan_dev_node *node)
9c2e24e1 2628{
56ef9c90 2629 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2630 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2631
69e76661 2632 node->vxlan = vxlan;
56ef9c90 2633 spin_lock(&vn->sock_lock);
69e76661 2634 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
56ef9c90 2635 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2636}
2637
d342894c 2638/* Setup stats when device is created */
2639static int vxlan_init(struct net_device *dev)
2640{
1c213bd2 2641 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2642 if (!dev->tstats)
d342894c 2643 return -ENOMEM;
2644
2645 return 0;
2646}
2647
3ad7a4b1 2648static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
afbd8bae
MR		 2649{
2650 struct vxlan_fdb *f;
2651
2652 spin_lock_bh(&vxlan->hash_lock);
3ad7a4b1 2653 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 2654 if (f)
d7e686dc 2655 vxlan_fdb_destroy(vxlan, f, true);
afbd8bae
MR		 2656 spin_unlock_bh(&vxlan->hash_lock);
2657}
2658
ebf4063e
SH		 2659static void vxlan_uninit(struct net_device *dev)
2660{
2661 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2662
aebd39ca
SB		 2663 gro_cells_destroy(&vxlan->gro_cells);
2664
3ad7a4b1 2665 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
afbd8bae 2666
ebf4063e
SH		 2667 free_percpu(dev->tstats);
2668}
2669
d342894c 2670/* Start ageing timer and join group when device is brought up */
2671static int vxlan_open(struct net_device *dev)
2672{
2673 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2674 int ret;
56ef9c90 2675
205f356d
JB		 2676 ret = vxlan_sock_add(vxlan);
2677 if (ret < 0)
2678 return ret;
d342894c 2679
79d4a94f 2680 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2681 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2682 if (ret == -EADDRINUSE)
2683 ret = 0;
56ef9c90 2684 if (ret) {
205f356d 2685 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2686 return ret;
2687 }
d342894c 2688 }
2689
0dfbdf41 2690 if (vxlan->cfg.age_interval)
d342894c 2691 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2692
56ef9c90 2693 return ret;
d342894c 2694}
2695
2696/* Purge the forwarding table */
8b3f9337 2697static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
d342894c 2698{
31fec5aa 2699 unsigned int h;
d342894c 2700
2701 spin_lock_bh(&vxlan->hash_lock);
2702 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2703 struct hlist_node *p, *n;
2704 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2705 struct vxlan_fdb *f
2706 = container_of(p, struct vxlan_fdb, hlist);
8b3f9337
RP		 2707 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2708 continue;
afbd8bae
MR		 2709 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2710 if (!is_zero_ether_addr(f->eth_addr))
d7e686dc 2711 vxlan_fdb_destroy(vxlan, f, true);
d342894c 2712 }
2713 }
2714 spin_unlock_bh(&vxlan->hash_lock);
2715}
2716
2717/* Cleanup timer and forwarding table on shutdown */
2718static int vxlan_stop(struct net_device *dev)
2719{
2720 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2721 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2722 int ret = 0;
d342894c 2723
24c0e683 2724 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2725 !vxlan_group_used(vn, vxlan))
56ef9c90 2726 ret = vxlan_igmp_leave(vxlan);
d342894c 2727
2728 del_timer_sync(&vxlan->age_timer);
2729
8b3f9337 2730 vxlan_flush(vxlan, false);
205f356d 2731 vxlan_sock_release(vxlan);
d342894c 2732
56ef9c90 2733 return ret;
d342894c 2734}
2735
d342894c 2736/* Stub, nothing needs to be done. */
2737static void vxlan_set_multicast_list(struct net_device *dev)
2738{
2739}
2740
91572088 2741static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
345010b5 2742{
91572088
JW		 2743 struct vxlan_dev *vxlan = netdev_priv(dev);
2744 struct vxlan_rdst *dst = &vxlan->default_dst;
2745 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2746 dst->remote_ifindex);
ce44a4ae 2747 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
345010b5 2748
91572088
JW		 2749 /* This check is different than dev->max_mtu, because it looks at
2750 * the lowerdev->mtu, rather than the static dev->max_mtu
2751 */
2752 if (lowerdev) {
2753 int max_mtu = lowerdev->mtu -
2754 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2755 if (new_mtu > max_mtu)
72564b59 2756 return -EINVAL;
72564b59
DW		 2757 }
2758
345010b5
DB		 2759 dev->mtu = new_mtu;
2760 return 0;
2761}
2762
fc4099f1
PS		 2763static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2764{
2765 struct vxlan_dev *vxlan = netdev_priv(dev);
2766 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2767 __be16 sport, dport;
2768
2769 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2770 vxlan->cfg.port_max, true);
2771 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2772
239e944f 2773 if (ip_tunnel_info_af(info) == AF_INET) {
c6fcc4fc 2774 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
1a8496ba
JB		 2775 struct rtable *rt;
2776
655c3de1 2777 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
1a8496ba 2778 info->key.u.ipv4.dst,
22f0708a
PA		 2779 &info->key.u.ipv4.src, dport, sport,
2780 &info->dst_cache, info);
1a8496ba
JB		 2781 if (IS_ERR(rt))
2782 return PTR_ERR(rt);
2783 ip_rt_put(rt);
239e944f
JB		 2784 } else {
2785#if IS_ENABLED(CONFIG_IPV6)
03dc52a8 2786 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
239e944f
JB		 2787 struct dst_entry *ndst;
2788
655c3de1 2789 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
e7f70af1 2790 info->key.label, &info->key.u.ipv6.dst,
22f0708a
PA		 2791 &info->key.u.ipv6.src, dport, sport,
2792 &info->dst_cache, info);
239e944f
JB		 2793 if (IS_ERR(ndst))
2794 return PTR_ERR(ndst);
2795 dst_release(ndst);
239e944f
JB		 2796#else /* !CONFIG_IPV6 */
2797 return -EPFNOSUPPORT;
2798#endif
2799 }
1a8496ba
JB		 2800 info->key.tp_src = sport;
2801 info->key.tp_dst = dport;
239e944f 2802 return 0;
fc4099f1
PS		 2803}
2804
0c867c9b 2805static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 2806 .ndo_init = vxlan_init,
ebf4063e 2807 .ndo_uninit = vxlan_uninit,
d342894c 2808 .ndo_open = vxlan_open,
2809 .ndo_stop = vxlan_stop,
2810 .ndo_start_xmit = vxlan_xmit,
e8171045 2811 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2812 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2813 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2814 .ndo_validate_addr = eth_validate_addr,
2815 .ndo_set_mac_address = eth_mac_addr,
2816 .ndo_fdb_add = vxlan_fdb_add,
2817 .ndo_fdb_del = vxlan_fdb_delete,
2818 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2819 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2820};
2821
e1e5314d
JB		 2822static const struct net_device_ops vxlan_netdev_raw_ops = {
2823 .ndo_init = vxlan_init,
2824 .ndo_uninit = vxlan_uninit,
2825 .ndo_open = vxlan_open,
2826 .ndo_stop = vxlan_stop,
2827 .ndo_start_xmit = vxlan_xmit,
2828 .ndo_get_stats64 = ip_tunnel_get_stats64,
2829 .ndo_change_mtu = vxlan_change_mtu,
2830 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2831};
2832
d342894c 2833/* Info for udev, that this is a virtual tunnel endpoint */
2834static struct device_type vxlan_type = {
2835 .name = "vxlan",
2836};
2837
e5de25dc 2838/* Calls the ndo_udp_tunnel_add of the caller in order to
35e42379 2839 * supply the listening VXLAN udp ports. Callers are expected
e5de25dc 2840 * to implement the ndo_udp_tunnel_add.
53cf5275 2841 */
2d2b13fc 2842static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
53cf5275
JG		 2843{
2844 struct vxlan_sock *vs;
2845 struct net *net = dev_net(dev);
2846 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
35e42379 2847 unsigned int i;
53cf5275
JG		 2848
2849 spin_lock(&vn->sock_lock);
2850 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2d2b13fc
SD		 2851 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2852 unsigned short type;
2853
2854 if (vs->flags & VXLAN_F_GPE)
2855 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
2856 else
2857 type = UDP_TUNNEL_TYPE_VXLAN;
2858
2859 if (push)
2860 udp_tunnel_push_rx_port(dev, vs->sock, type);
2861 else
2862 udp_tunnel_drop_rx_port(dev, vs->sock, type);
2863 }
53cf5275
JG		 2864 }
2865 spin_unlock(&vn->sock_lock);
2866}
53cf5275 2867
d342894c 2868/* Initialize the device structure. */
2869static void vxlan_setup(struct net_device *dev)
2870{
2871 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2872 unsigned int h;
d342894c 2873
65226ef8
JB		 2874 eth_hw_addr_random(dev);
2875 ether_setup(dev);
2876
cf124db5 2877 dev->needs_free_netdev = true;
d342894c 2878 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2879
d342894c 2880 dev->features |= NETIF_F_LLTX;
d6727fe3 2881 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2882 dev->features |= NETIF_F_RXCSUM;
05c0db08 2883 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2884
1eaa8178 2885 dev->vlan_features = dev->features;
0afb1666 2886 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2887 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
02875878 2888 netif_keep_dst(dev);
0c867c9b 2889 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 2890
a985343b
MS		 2891 /* MTU range: 68 - 65535 */
2892 dev->min_mtu = ETH_MIN_MTU;
2893 dev->max_mtu = ETH_MAX_MTU;
2894
553675fb 2895 INIT_LIST_HEAD(&vxlan->next);
d342894c 2896 spin_lock_init(&vxlan->hash_lock);
2897
df7e828c 2898 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
d342894c 2899
2900 vxlan->dev = dev;
2901
58ce31cc
TH		 2902 gro_cells_init(&vxlan->gro_cells, dev);
2903
d342894c 2904 for (h = 0; h < FDB_HASH_SIZE; ++h)
2905 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
d57420a1
JV		 2906
2907 INIT_LIST_HEAD(&vxlan->fan.fan_maps);
d342894c 2908}
2909
0c867c9b
JB		 2910static void vxlan_ether_setup(struct net_device *dev)
2911{
0c867c9b
JB		 2912 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2913 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2914 dev->netdev_ops = &vxlan_netdev_ether_ops;
2915}
2916
e1e5314d
JB		 2917static void vxlan_raw_setup(struct net_device *dev)
2918{
65226ef8 2919 dev->header_ops = NULL;
e1e5314d
JB		 2920 dev->type = ARPHRD_NONE;
2921 dev->hard_header_len = 0;
2922 dev->addr_len = 0;
e1e5314d
JB		 2923 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2924 dev->netdev_ops = &vxlan_netdev_raw_ops;
2925}
2926
d342894c 2927static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2928 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2929 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2930 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2931 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2932 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2933 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2934 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2935 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 2936 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 2937 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2938 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2939 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2940 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2941 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2942 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2943 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2944 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2945 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2946 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2947 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2948 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2949 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2950 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2951 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2952 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 2953 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 2954 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2955};
2956
a8b8a889
MS		 2957static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
2958 struct netlink_ext_ack *extack)
d342894c 2959{
2960 if (tb[IFLA_ADDRESS]) {
2961 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
653ef6a3
GM		 2962 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2963 "Provided link layer address is not Ethernet");
d342894c 2964 return -EINVAL;
2965 }
2966
2967 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
653ef6a3
GM		 2968 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2969 "Provided Ethernet address is not unicast");
d342894c 2970 return -EADDRNOTAVAIL;
2971 }
2972 }
2973
a985343b 2974 if (tb[IFLA_MTU]) {
019b13ae 2975 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
a985343b 2976
653ef6a3
GM		 2977 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
2978 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
2979 "MTU must be between 68 and 65535");
a985343b 2980 return -EINVAL;
653ef6a3 2981 }
a985343b
MS		 2982 }
2983
653ef6a3
GM		 2984 if (!data) {
2985 NL_SET_ERR_MSG(extack,
2986 "Required attributes not provided to perform the operation");
d342894c 2987 return -EINVAL;
653ef6a3 2988 }
d342894c 2989
2990 if (data[IFLA_VXLAN_ID]) {
a985343b
MS		 2991 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2992
653ef6a3
GM		 2993 if (id >= VXLAN_N_VID) {
2994 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID],
2995 "VXLAN ID must be lower than 16777216");
d342894c 2996 return -ERANGE;
653ef6a3 2997 }
d342894c 2998 }
2999
05f47d69 3000 if (data[IFLA_VXLAN_PORT_RANGE]) {
3001 const struct ifla_vxlan_port_range *p
3002 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3003
3004 if (ntohs(p->high) < ntohs(p->low)) {
653ef6a3
GM		 3005 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
3006 "Invalid source port range");
05f47d69 3007 return -EINVAL;
3008 }
3009 }
3010
d342894c 3011 return 0;
3012}
3013
1b13c97f
YB		 3014static void vxlan_get_drvinfo(struct net_device *netdev,
3015 struct ethtool_drvinfo *drvinfo)
3016{
3017 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3018 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3019}
3020
3021static const struct ethtool_ops vxlan_ethtool_ops = {
3022 .get_drvinfo = vxlan_get_drvinfo,
3023 .get_link = ethtool_op_get_link,
3024};
3025
3ee64f39
TH		 3026static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3027 __be16 port, u32 flags)
553675fb 3028{
e4c7ed41 3029 struct socket *sock;
3ee64f39
TH		 3030 struct udp_port_cfg udp_conf;
3031 int err;
e4c7ed41 3032
3ee64f39 3033 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 3034
3ee64f39
TH		 3035 if (ipv6) {
3036 udp_conf.family = AF_INET6;
3ee64f39 3037 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 3038 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 3039 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 3040 } else {
3041 udp_conf.family = AF_INET;
e4c7ed41 3042 }
e4c7ed41 3043
3ee64f39 3044 udp_conf.local_udp_port = port;
553675fb 3045
3ee64f39
TH		 3046 /* Open UDP socket */
3047 err = udp_sock_create(net, &udp_conf, &sock);
3048 if (err < 0)
3049 return ERR_PTR(err);
e4c7ed41 3050
39deb2c7 3051 return sock;
e4c7ed41
CW		 3052}
3053
3054/* Create new listen socket if needed */
b1be00a6
JB		 3055static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3056 __be16 port, u32 flags)
e4c7ed41
CW		 3057{
3058 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3059 struct vxlan_sock *vs;
3060 struct socket *sock;
e4c7ed41 3061 unsigned int h;
acbf74a7 3062 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 3063
dc01e7d3 3064 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 3065 if (!vs)
3066 return ERR_PTR(-ENOMEM);
3067
3068 for (h = 0; h < VNI_HASH_SIZE; ++h)
3069 INIT_HLIST_HEAD(&vs->vni_list[h]);
3070
3ee64f39 3071 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 3072 if (IS_ERR(sock)) {
553675fb 3073 kfree(vs);
e50fddc8 3074 return ERR_CAST(sock);
553675fb 3075 }
e4c7ed41
CW		 3076
3077 vs->sock = sock;
66af846f 3078 refcount_set(&vs->refcnt, 1);
af33c1ad 3079 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 3080
9c2e24e1
PS		 3081 spin_lock(&vn->sock_lock);
3082 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
e7b3db5e 3083 udp_tunnel_notify_add_rx_port(sock,
b9adcd69
AD		 3084 (vs->flags & VXLAN_F_GPE) ?
3085 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 3086 UDP_TUNNEL_TYPE_VXLAN);
9c2e24e1 3087 spin_unlock(&vn->sock_lock);
553675fb 3088
3089 /* Mark socket as an encapsulation socket. */
5602c48c 3090 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ		 3091 tunnel_cfg.sk_user_data = vs;
3092 tunnel_cfg.encap_type = 1;
f2d1968e 3093 tunnel_cfg.encap_rcv = vxlan_rcv;
acbf74a7 3094 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH		 3095 tunnel_cfg.gro_receive = vxlan_gro_receive;
3096 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ		 3097
3098 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 3099
9c2e24e1
PS		 3100 return vs;
3101}
3102
b1be00a6 3103static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 3104{
205f356d
JB		 3105 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3106 struct vxlan_sock *vs = NULL;
69e76661 3107 struct vxlan_dev_node *node;
9c2e24e1 3108
205f356d 3109 if (!vxlan->cfg.no_share) {
56ef9c90 3110 spin_lock(&vn->sock_lock);
205f356d 3111 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
dc5321d7 3112 vxlan->cfg.dst_port, vxlan->cfg.flags);
66af846f 3113 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
56ef9c90 3114 spin_unlock(&vn->sock_lock);
205f356d 3115 return -EBUSY;
56ef9c90
MRL		 3116 }
3117 spin_unlock(&vn->sock_lock);
3118 }
205f356d 3119 if (!vs)
b1be00a6 3120 vs = vxlan_socket_create(vxlan->net, ipv6,
dc5321d7 3121 vxlan->cfg.dst_port, vxlan->cfg.flags);
205f356d
JB		 3122 if (IS_ERR(vs))
3123 return PTR_ERR(vs);
b1be00a6 3124#if IS_ENABLED(CONFIG_IPV6)
69e76661 3125 if (ipv6) {
c6fcc4fc 3126 rcu_assign_pointer(vxlan->vn6_sock, vs);
69e76661
JB		 3127 node = &vxlan->hlist6;
3128 } else
b1be00a6 3129#endif
69e76661 3130 {
c6fcc4fc 3131 rcu_assign_pointer(vxlan->vn4_sock, vs);
69e76661
JB		 3132 node = &vxlan->hlist4;
3133 }
3134 vxlan_vs_add_dev(vs, vxlan, node);
205f356d 3135 return 0;
553675fb 3136}
3137
b1be00a6
JB		 3138static int vxlan_sock_add(struct vxlan_dev *vxlan)
3139{
dc5321d7
MS		 3140 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3141 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
d074bf96 3142 bool ipv4 = !ipv6 || metadata;
b1be00a6
JB		 3143 int ret = 0;
3144
c6fcc4fc 3145 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
b1be00a6 3146#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 3147 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
d074bf96 3148 if (ipv6) {
b1be00a6 3149 ret = __vxlan_sock_add(vxlan, true);
d074bf96
JB		 3150 if (ret < 0 && ret != -EAFNOSUPPORT)
3151 ipv4 = false;
3152 }
b1be00a6 3153#endif
d074bf96 3154 if (ipv4)
b1be00a6
JB		 3155 ret = __vxlan_sock_add(vxlan, false);
3156 if (ret < 0)
3157 vxlan_sock_release(vxlan);
3158 return ret;
3159}
3160
a985343b
MS		 3161static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3162 struct net_device **lower,
653ef6a3
GM		 3163 struct vxlan_dev *old,
3164 struct netlink_ext_ack *extack)
d342894c 3165{
33564bbb 3166 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
a985343b 3167 struct vxlan_dev *tmp;
e4c7ed41 3168 bool use_ipv6 = false;
d342894c 3169
a985343b
MS		 3170 if (conf->flags & VXLAN_F_GPE) {
3171 /* For now, allow GPE only together with
3172 * COLLECT_METADATA. This can be relaxed later; in such
3173 * case, the other side of the PtP link will have to be
3174 * provided.
3175 */
3176 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3177 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
653ef6a3
GM		 3178 NL_SET_ERR_MSG(extack,
3179 "VXLAN GPE does not support this combination of attributes");
a985343b 3180 return -EINVAL;
3555621d 3181 }
e1e5314d 3182 }
0c867c9b 3183
ce44a4ae
MS		 3184 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3185 /* Unless IPv6 is explicitly requested, assume IPv4 */
a985343b 3186 conf->remote_ip.sa.sa_family = AF_INET;
ce44a4ae
MS		 3187 conf->saddr.sa.sa_family = AF_INET;
3188 } else if (!conf->remote_ip.sa.sa_family) {
3189 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3190 } else if (!conf->saddr.sa.sa_family) {
3191 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3192 }
3193
653ef6a3
GM		 3194 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3195 NL_SET_ERR_MSG(extack,
3196 "Local and remote address must be from the same family");
ce44a4ae 3197 return -EINVAL;
653ef6a3 3198 }
e4c7ed41 3199
653ef6a3
GM		 3200 if (vxlan_addr_multicast(&conf->saddr)) {
3201 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
0f22a3c6 3202 return -EINVAL;
653ef6a3 3203 }
0f22a3c6 3204
ce44a4ae 3205 if (conf->saddr.sa.sa_family == AF_INET6) {
653ef6a3
GM		 3206 if (!IS_ENABLED(CONFIG_IPV6)) {
3207 NL_SET_ERR_MSG(extack,
3208 "IPv6 support not enabled in the kernel");
057ba29b 3209 return -EPFNOSUPPORT;
653ef6a3 3210 }
e4c7ed41 3211 use_ipv6 = true;
a985343b 3212 conf->flags |= VXLAN_F_IPV6;
0f22a3c6
MS		 3213
3214 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3215 int local_type =
3216 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3217 int remote_type =
3218 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3219
3220 if (local_type & IPV6_ADDR_LINKLOCAL) {
3221 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
653ef6a3
GM		 3222 (remote_type != IPV6_ADDR_ANY)) {
3223 NL_SET_ERR_MSG(extack,
3224 "Invalid combination of local and remote address scopes");
0f22a3c6 3225 return -EINVAL;
653ef6a3 3226 }
0f22a3c6
MS		 3227
3228 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3229 } else {
3230 if (remote_type ==
653ef6a3
GM		 3231 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3232 NL_SET_ERR_MSG(extack,
3233 "Invalid combination of local and remote address scopes");
0f22a3c6 3234 return -EINVAL;
653ef6a3 3235 }
0f22a3c6
MS		 3236
3237 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3238 }
3239 }
057ba29b 3240 }
d342894c 3241
653ef6a3
GM		 3242 if (conf->label && !use_ipv6) {
3243 NL_SET_ERR_MSG(extack,
3244 "Label attribute only applies to IPv6 VXLAN devices");
e7f70af1 3245 return -EINVAL;
653ef6a3 3246 }
e7f70af1 3247
a985343b
MS		 3248 if (conf->remote_ifindex) {
3249 struct net_device *lowerdev;
34e02aa1 3250
a985343b 3251 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
653ef6a3
GM		 3252 if (!lowerdev) {
3253 NL_SET_ERR_MSG(extack,
3254 "Invalid local interface, device not found");
34e02aa1 3255 return -ENODEV;
653ef6a3 3256 }
d342894c 3257
e4c7ed41
CW		 3258#if IS_ENABLED(CONFIG_IPV6)
3259 if (use_ipv6) {
3260 struct inet6_dev *idev = __in6_dev_get(lowerdev);
653ef6a3
GM		 3261 if (idev && idev->cnf.disable_ipv6) {
3262 NL_SET_ERR_MSG(extack,
3263 "IPv6 support disabled by administrator");
e4c7ed41 3264 return -EPERM;
653ef6a3 3265 }
e4c7ed41
CW		 3266 }
3267#endif
3268
a985343b
MS		 3269 *lower = lowerdev;
3270 } else {
653ef6a3
GM		 3271 if (vxlan_addr_multicast(&conf->remote_ip)) {
3272 NL_SET_ERR_MSG(extack,
3273 "Local interface required for multicast remote destination");
3274
a985343b 3275 return -EINVAL;
653ef6a3 3276 }
1ba56fb4 3277
0f22a3c6 3278#if IS_ENABLED(CONFIG_IPV6)
653ef6a3
GM		 3279 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3280 NL_SET_ERR_MSG(extack,
3281 "Local interface required for link-local local/remote addresses");
0f22a3c6 3282 return -EINVAL;
653ef6a3 3283 }
0f22a3c6
MS		 3284#endif
3285
a985343b 3286 *lower = NULL;
9dc2ad10 3287 }
d342894c 3288
a985343b
MS		 3289 if (!conf->dst_port) {
3290 if (conf->flags & VXLAN_F_GPE)
3291 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3292 else
3293 conf->dst_port = htons(vxlan_port);
d6acfeb1
FM		 3294 }
3295
a985343b
MS		 3296 if (!conf->age_interval)
3297 conf->age_interval = FDB_AGE_DEFAULT;
91572088 3298
a985343b
MS		 3299 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3300 if (tmp == old)
3301 continue;
91572088 3302
49f810f0
MS		 3303 if (tmp->cfg.vni != conf->vni)
3304 continue;
3305 if (tmp->cfg.dst_port != conf->dst_port)
3306 continue;
3307 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
ce44a4ae 3308 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
49f810f0
MS		 3309 continue;
3310
3311 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3312 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3313 continue;
3314
653ef6a3
GM		 3315 NL_SET_ERR_MSG(extack,
3316 "A VXLAN device with the specified VNI already exists");
49f810f0 3317 return -EEXIST;
a985343b 3318 }
91572088 3319
a985343b
MS		 3320 return 0;
3321}
3322
3323static void vxlan_config_apply(struct net_device *dev,
3324 struct vxlan_config *conf,
889ce937
SD		 3325 struct net_device *lowerdev,
3326 struct net *src_net,
3327 bool changelink)
a985343b
MS		 3328{
3329 struct vxlan_dev *vxlan = netdev_priv(dev);
3330 struct vxlan_rdst *dst = &vxlan->default_dst;
3331 unsigned short needed_headroom = ETH_HLEN;
3332 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3333 int max_mtu = ETH_MAX_MTU;
3334
3335 if (!changelink) {
3336 if (conf->flags & VXLAN_F_GPE)
3337 vxlan_raw_setup(dev);
3338 else
3339 vxlan_ether_setup(dev);
3340
3341 if (conf->mtu)
3342 dev->mtu = conf->mtu;
889ce937
SD		 3343
3344 vxlan->net = src_net;
a985343b
MS		 3345 }
3346
3347 dst->remote_vni = conf->vni;
3348
3349 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3350
3351 if (lowerdev) {
3352 dst->remote_ifindex = conf->remote_ifindex;
3353
3354 dev->gso_max_size = lowerdev->gso_max_size;
3355 dev->gso_max_segs = lowerdev->gso_max_segs;
91572088 3356
a985343b 3357 needed_headroom = lowerdev->hard_header_len;
91572088 3358
a985343b
MS		 3359 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3360 VXLAN_HEADROOM);
f870c1ff
AK		 3361 if (max_mtu < ETH_MIN_MTU)
3362 max_mtu = ETH_MIN_MTU;
3363
3364 if (!changelink && !conf->mtu)
3365 dev->mtu = max_mtu;
7e059158
DW		 3366 }
3367
a985343b
MS		 3368 if (dev->mtu > max_mtu)
3369 dev->mtu = max_mtu;
3370
b1be00a6
JB		 3371 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3372 needed_headroom += VXLAN6_HEADROOM;
3373 else
3374 needed_headroom += VXLAN_HEADROOM;
3375 dev->needed_headroom = needed_headroom;
3376
0dfbdf41 3377 memcpy(&vxlan->cfg, conf, sizeof(*conf));
a985343b 3378}
0dfbdf41 3379
a985343b 3380static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
653ef6a3
GM		 3381 struct vxlan_config *conf, bool changelink,
3382 struct netlink_ext_ack *extack)
a985343b
MS		 3383{
3384 struct vxlan_dev *vxlan = netdev_priv(dev);
3385 struct net_device *lowerdev;
3386 int ret;
0dfbdf41 3387
653ef6a3 3388 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
a985343b
MS		 3389 if (ret)
3390 return ret;
8bcdc4f3 3391
889ce937 3392 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
0dfbdf41 3393
0dfbdf41
TG		 3394 return 0;
3395}
3396
c80498e3 3397static int __vxlan_dev_create(struct net *net, struct net_device *dev,
653ef6a3
GM		 3398 struct vxlan_config *conf,
3399 struct netlink_ext_ack *extack)
c80498e3
ND		 3400{
3401 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3402 struct vxlan_dev *vxlan = netdev_priv(dev);
d56a6568 3403 struct vxlan_fdb *f = NULL;
775d142f 3404 bool unregister = false;
c80498e3
ND		 3405 int err;
3406
653ef6a3 3407 err = vxlan_dev_configure(net, dev, conf, false, extack);
c80498e3
ND		 3408 if (err)
3409 return err;
3410
3411 dev->ethtool_ops = &vxlan_ethtool_ops;
3412
3413 /* create an fdb entry for a valid default destination */
3414 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
d56a6568 3415 err = vxlan_fdb_create(vxlan, all_zeros_mac,
c80498e3
ND		 3416 &vxlan->default_dst.remote_ip,
3417 NUD_REACHABLE | NUD_PERMANENT,
c80498e3
ND		 3418 vxlan->cfg.dst_port,
3419 vxlan->default_dst.remote_vni,
3420 vxlan->default_dst.remote_vni,
3421 vxlan->default_dst.remote_ifindex,
d56a6568 3422 NTF_SELF, &f);
c80498e3
ND		 3423 if (err)
3424 return err;
3425 }
3426
3427 err = register_netdevice(dev);
d56a6568
RP		 3428 if (err)
3429 goto errout;
775d142f 3430 unregister = true;
d56a6568
RP		 3431
3432 err = rtnl_configure_link(dev, NULL);
775d142f 3433 if (err)
d56a6568 3434 goto errout;
c80498e3 3435
d56a6568
RP		 3436 /* notify default fdb entry */
3437 if (f)
3438 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH);
3439
c80498e3
ND		 3440 list_add(&vxlan->next, &vn->vxlan_list);
3441 return 0;
775d142f 3442
d56a6568 3443errout:
775d142f
PM		 3444 /* unregister_netdevice() destroys the default FDB entry with deletion
3445 * notification. But the addition notification was not sent yet, so
3446 * destroy the entry by hand here.
3447 */
d56a6568
RP		 3448 if (f)
3449 vxlan_fdb_destroy(vxlan, f, false);
775d142f
PM		 3450 if (unregister)
3451 unregister_netdevice(dev);
d56a6568 3452 return err;
c80498e3
ND		 3453}
3454
8bcdc4f3
RP		 3455static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3456 struct net_device *dev, struct vxlan_config *conf,
3457 bool changelink)
0dfbdf41 3458{
8bcdc4f3 3459 struct vxlan_dev *vxlan = netdev_priv(dev);
d57420a1 3460 int err;
0dfbdf41 3461
8bcdc4f3 3462 memset(conf, 0, sizeof(*conf));
e277de5f 3463
8bcdc4f3
RP		 3464 /* if changelink operation, start with old existing cfg */
3465 if (changelink)
3466 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3467
3468 if (data[IFLA_VXLAN_ID]) {
3469 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3470
3471 if (changelink && (vni != conf->vni))
3472 return -EOPNOTSUPP;
3473 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3474 }
0dfbdf41
TG		 3475
3476 if (data[IFLA_VXLAN_GROUP]) {
ce44a4ae
MS		 3477 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET))
3478 return -EOPNOTSUPP;
3479
8bcdc4f3 3480 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
ce44a4ae 3481 conf->remote_ip.sa.sa_family = AF_INET;
0dfbdf41
TG		 3482 } else if (data[IFLA_VXLAN_GROUP6]) {
3483 if (!IS_ENABLED(CONFIG_IPV6))
3484 return -EPFNOSUPPORT;
3485
ce44a4ae
MS		 3486 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6))
3487 return -EOPNOTSUPP;
3488
8bcdc4f3
RP		 3489 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3490 conf->remote_ip.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3491 }
3492
d57420a1
JV		 3493 if (data[IFLA_VXLAN_FAN_MAP]) {
3494 err = vxlan_parse_fan_map(data, vxlan);
3495 if (err)
3496 return err;
3497 }
3498
0dfbdf41 3499 if (data[IFLA_VXLAN_LOCAL]) {
ce44a4ae
MS		 3500 if (changelink && (conf->saddr.sa.sa_family != AF_INET))
3501 return -EOPNOTSUPP;
3502
8bcdc4f3
RP		 3503 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3504 conf->saddr.sa.sa_family = AF_INET;
0dfbdf41
TG		 3505 } else if (data[IFLA_VXLAN_LOCAL6]) {
3506 if (!IS_ENABLED(CONFIG_IPV6))
3507 return -EPFNOSUPPORT;
3508
ce44a4ae
MS		 3509 if (changelink && (conf->saddr.sa.sa_family != AF_INET6))
3510 return -EOPNOTSUPP;
3511
0dfbdf41 3512 /* TODO: respect scope id */
8bcdc4f3
RP		 3513 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3514 conf->saddr.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3515 }
3516
3517 if (data[IFLA_VXLAN_LINK])
8bcdc4f3 3518 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
0dfbdf41 3519
d342894c 3520 if (data[IFLA_VXLAN_TOS])
8bcdc4f3 3521 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 3522
afb97186 3523 if (data[IFLA_VXLAN_TTL])
8bcdc4f3 3524 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 3525
e7f70af1 3526 if (data[IFLA_VXLAN_LABEL])
8bcdc4f3 3527 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
e7f70af1
DB		 3528 IPV6_FLOWLABEL_MASK;
3529
8bcdc4f3 3530 if (data[IFLA_VXLAN_LEARNING]) {
dc5321d7 3531 if (nla_get_u8(data[IFLA_VXLAN_LEARNING]))
8bcdc4f3 3532 conf->flags |= VXLAN_F_LEARN;
dc5321d7 3533 else
8bcdc4f3 3534 conf->flags &= ~VXLAN_F_LEARN;
8bcdc4f3
RP		 3535 } else if (!changelink) {
3536 /* default to learn on a new device */
3537 conf->flags |= VXLAN_F_LEARN;
3538 }
d342894c 3539
8bcdc4f3
RP		 3540 if (data[IFLA_VXLAN_AGEING]) {
3541 if (changelink)
3542 return -EOPNOTSUPP;
3543 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3544 }
d342894c 3545
8bcdc4f3
RP		 3546 if (data[IFLA_VXLAN_PROXY]) {
3547 if (changelink)
3548 return -EOPNOTSUPP;
3549 if (nla_get_u8(data[IFLA_VXLAN_PROXY]))
3550 conf->flags |= VXLAN_F_PROXY;
3551 }
e4f67add 3552
8bcdc4f3
RP		 3553 if (data[IFLA_VXLAN_RSC]) {
3554 if (changelink)
3555 return -EOPNOTSUPP;
3556 if (nla_get_u8(data[IFLA_VXLAN_RSC]))
3557 conf->flags |= VXLAN_F_RSC;
3558 }
e4f67add 3559
8bcdc4f3
RP		 3560 if (data[IFLA_VXLAN_L2MISS]) {
3561 if (changelink)
3562 return -EOPNOTSUPP;
3563 if (nla_get_u8(data[IFLA_VXLAN_L2MISS]))
3564 conf->flags |= VXLAN_F_L2MISS;
3565 }
e4f67add 3566
8bcdc4f3
RP		 3567 if (data[IFLA_VXLAN_L3MISS]) {
3568 if (changelink)
3569 return -EOPNOTSUPP;
3570 if (nla_get_u8(data[IFLA_VXLAN_L3MISS]))
3571 conf->flags |= VXLAN_F_L3MISS;
3572 }
e4f67add 3573
8bcdc4f3
RP		 3574 if (data[IFLA_VXLAN_LIMIT]) {
3575 if (changelink)
3576 return -EOPNOTSUPP;
3577 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3578 }
d342894c 3579
8bcdc4f3
RP		 3580 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3581 if (changelink)
3582 return -EOPNOTSUPP;
3583 if (nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3584 conf->flags |= VXLAN_F_COLLECT_METADATA;
3585 }
f8a9b1bc 3586
05f47d69 3587 if (data[IFLA_VXLAN_PORT_RANGE]) {
8bcdc4f3
RP		 3588 if (!changelink) {
3589 const struct ifla_vxlan_port_range *p
3590 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3591 conf->port_min = ntohs(p->low);
3592 conf->port_max = ntohs(p->high);
3593 } else {
3594 return -EOPNOTSUPP;
3595 }
05f47d69 3596 }
3597
8bcdc4f3
RP		 3598 if (data[IFLA_VXLAN_PORT]) {
3599 if (changelink)
3600 return -EOPNOTSUPP;
3601 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3602 }
823aa873 3603
8bcdc4f3
RP		 3604 if (data[IFLA_VXLAN_UDP_CSUM]) {
3605 if (changelink)
3606 return -EOPNOTSUPP;
3607 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3608 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3609 }
359a0ea9 3610
8bcdc4f3
RP		 3611 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3612 if (changelink)
3613 return -EOPNOTSUPP;
3614 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
3615 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
3616 }
359a0ea9 3617
8bcdc4f3
RP		 3618 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3619 if (changelink)
3620 return -EOPNOTSUPP;
3621 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
3622 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
3623 }
359a0ea9 3624
8bcdc4f3
RP		 3625 if (data[IFLA_VXLAN_REMCSUM_TX]) {
3626 if (changelink)
3627 return -EOPNOTSUPP;
3628 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
3629 conf->flags |= VXLAN_F_REMCSUM_TX;
3630 }
dfd8645e 3631
8bcdc4f3
RP		 3632 if (data[IFLA_VXLAN_REMCSUM_RX]) {
3633 if (changelink)
3634 return -EOPNOTSUPP;
3635 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3636 conf->flags |= VXLAN_F_REMCSUM_RX;
3637 }
3638
3639 if (data[IFLA_VXLAN_GBP]) {
3640 if (changelink)
3641 return -EOPNOTSUPP;
3642 conf->flags |= VXLAN_F_GBP;
3643 }
3644
3645 if (data[IFLA_VXLAN_GPE]) {
3646 if (changelink)
3647 return -EOPNOTSUPP;
3648 conf->flags |= VXLAN_F_GPE;
3649 }
3650
3651 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3652 if (changelink)
3653 return -EOPNOTSUPP;
3654 conf->flags |= VXLAN_F_REMCSUM_NOPARTIAL;
3655 }
3656
3657 if (tb[IFLA_MTU]) {
3658 if (changelink)
3659 return -EOPNOTSUPP;
3660 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3661 }
3662
3663 return 0;
3664}
3665
3666static int vxlan_newlink(struct net *src_net, struct net_device *dev,
7a3f4a18
MS		 3667 struct nlattr *tb[], struct nlattr *data[],
3668 struct netlink_ext_ack *extack)
8bcdc4f3 3669{
8bcdc4f3
RP		 3670 struct vxlan_config conf;
3671 int err;
3672
3673 err = vxlan_nl2conf(tb, data, dev, &conf, false);
3674 if (err)
3675 return err;
3676
653ef6a3 3677 return __vxlan_dev_create(src_net, dev, &conf, extack);
8bcdc4f3 3678}
dfd8645e 3679
8bcdc4f3 3680static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
ad744b22
MS		 3681 struct nlattr *data[],
3682 struct netlink_ext_ack *extack)
8bcdc4f3
RP		 3683{
3684 struct vxlan_dev *vxlan = netdev_priv(dev);
3685 struct vxlan_rdst *dst = &vxlan->default_dst;
3686 struct vxlan_rdst old_dst;
3687 struct vxlan_config conf;
3688 int err;
3689
3690 err = vxlan_nl2conf(tb, data,
3691 dev, &conf, true);
3692 if (err)
3693 return err;
3511494c 3694
8bcdc4f3 3695 memcpy(&old_dst, dst, sizeof(struct vxlan_rdst));
e1e5314d 3696
653ef6a3 3697 err = vxlan_dev_configure(vxlan->net, dev, &conf, true, extack);
8bcdc4f3
RP		 3698 if (err)
3699 return err;
0ace2ca8 3700
8bcdc4f3
RP		 3701 /* handle default dst entry */
3702 if (!vxlan_addr_equal(&dst->remote_ip, &old_dst.remote_ip)) {
3703 spin_lock_bh(&vxlan->hash_lock);
3704 if (!vxlan_addr_any(&old_dst.remote_ip))
3705 __vxlan_fdb_delete(vxlan, all_zeros_mac,
3706 old_dst.remote_ip,
3707 vxlan->cfg.dst_port,
3708 old_dst.remote_vni,
3709 old_dst.remote_vni,
3710 old_dst.remote_ifindex, 0);
3711
3712 if (!vxlan_addr_any(&dst->remote_ip)) {
3833df52 3713 err = vxlan_fdb_update(vxlan, all_zeros_mac,
8bcdc4f3
RP		 3714 &dst->remote_ip,
3715 NUD_REACHABLE | NUD_PERMANENT,
3833df52 3716 NLM_F_APPEND | NLM_F_CREATE,
8bcdc4f3
RP		 3717 vxlan->cfg.dst_port,
3718 dst->remote_vni,
3719 dst->remote_vni,
3720 dst->remote_ifindex,
3833df52 3721 NTF_SELF);
8bcdc4f3
RP		 3722 if (err) {
3723 spin_unlock_bh(&vxlan->hash_lock);
3724 return err;
3725 }
3726 }
3727 spin_unlock_bh(&vxlan->hash_lock);
3728 }
ce577668 3729
8bcdc4f3 3730 return 0;
d342894c 3731}
3732
3733static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3734{
3735 struct vxlan_dev *vxlan = netdev_priv(dev);
3736
8b3f9337
RP		 3737 vxlan_flush(vxlan, true);
3738
553675fb 3739 list_del(&vxlan->next);
d342894c 3740 unregister_netdevice_queue(dev, head);
3741}
3742
3743static size_t vxlan_get_size(const struct net_device *dev)
3744{
3745
3746 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3747 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3748 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3749 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3750 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3751 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
e7f70af1 3752 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 3753 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 3754 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3755 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3756 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3757 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3758 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3759 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3760 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3761 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3762 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3763 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3764 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3765 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3766 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3767 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d57420a1 3768 nla_total_size(sizeof(struct ip_fan_map) * 256) +
d342894c 3769 0;
3770}
3771
3772static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3773{
3774 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3775 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3776 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3777 .low = htons(vxlan->cfg.port_min),
3778 .high = htons(vxlan->cfg.port_max),
05f47d69 3779 };
d342894c 3780
54bfd872 3781 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3782 goto nla_put_failure;
3783
e4c7ed41
CW		 3784 if (!vxlan_addr_any(&dst->remote_ip)) {
3785 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3786 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3787 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3788 goto nla_put_failure;
3789#if IS_ENABLED(CONFIG_IPV6)
3790 } else {
930345ea
JB		 3791 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3792 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3793 goto nla_put_failure;
3794#endif
3795 }
3796 }
d342894c 3797
c7995c43 3798 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3799 goto nla_put_failure;
3800
0dfbdf41
TG		 3801 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3802 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3803 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3804 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3805 goto nla_put_failure;
3806#if IS_ENABLED(CONFIG_IPV6)
3807 } else {
930345ea 3808 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3809 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3810 goto nla_put_failure;
3811#endif
3812 }
3813 }
d342894c 3814
d57420a1
JV		 3815 if (fan_has_map(&vxlan->fan)) {
3816 struct nlattr *fan_nest;
3817 struct ip_fan_map *fan_map;
3818
3819 fan_nest = nla_nest_start(skb, IFLA_VXLAN_FAN_MAP);
3820 if (!fan_nest)
3821 goto nla_put_failure;
3822 list_for_each_entry_rcu(fan_map, &vxlan->fan.fan_maps, list) {
3823 struct ifla_fan_map map;
3824
3825 map.underlay = fan_map->underlay;
3826 map.underlay_prefix = fan_map->underlay_prefix;
3827 map.overlay = fan_map->overlay;
3828 map.overlay_prefix = fan_map->overlay_prefix;
3829 if (nla_put(skb, IFLA_FAN_MAPPING, sizeof(map), &map))
3830 goto nla_put_failure;
3831 }
3832 nla_nest_end(skb, fan_nest);
3833 }
3834
0dfbdf41
TG		 3835 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3836 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e7f70af1 3837 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add 3838 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
dc5321d7 3839 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
e4f67add 3840 nla_put_u8(skb, IFLA_VXLAN_PROXY,
dc5321d7
MS		 3841 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
3842 nla_put_u8(skb, IFLA_VXLAN_RSC,
3843 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
e4f67add 3844 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
dc5321d7 3845 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
e4f67add 3846 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
dc5321d7 3847 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
da8b43c0 3848 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
dc5321d7 3849 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3850 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3851 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3852 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3853 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
dc5321d7 3854 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9 3855 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
dc5321d7 3856 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
359a0ea9 3857 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dc5321d7 3858 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
dfd8645e 3859 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
dc5321d7 3860 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
dfd8645e 3861 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
dc5321d7 3862 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
d342894c 3863 goto nla_put_failure;
3864
05f47d69 3865 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3866 goto nla_put_failure;
3867
dc5321d7 3868 if (vxlan->cfg.flags & VXLAN_F_GBP &&
3511494c
TG		 3869 nla_put_flag(skb, IFLA_VXLAN_GBP))
3870 goto nla_put_failure;
3871
dc5321d7 3872 if (vxlan->cfg.flags & VXLAN_F_GPE &&
e1e5314d
JB		 3873 nla_put_flag(skb, IFLA_VXLAN_GPE))
3874 goto nla_put_failure;
3875
dc5321d7 3876 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
0ace2ca8
TH		 3877 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3878 goto nla_put_failure;
3879
d342894c 3880 return 0;
3881
3882nla_put_failure:
3883 return -EMSGSIZE;
3884}
3885
1728d4fa
ND		 3886static struct net *vxlan_get_link_net(const struct net_device *dev)
3887{
3888 struct vxlan_dev *vxlan = netdev_priv(dev);
3889
3890 return vxlan->net;
3891}
3892
d342894c 3893static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3894 .kind = "vxlan",
3895 .maxtype = IFLA_VXLAN_MAX,
3896 .policy = vxlan_policy,
3897 .priv_size = sizeof(struct vxlan_dev),
3898 .setup = vxlan_setup,
3899 .validate = vxlan_validate,
3900 .newlink = vxlan_newlink,
8bcdc4f3 3901 .changelink = vxlan_changelink,
d342894c 3902 .dellink = vxlan_dellink,
3903 .get_size = vxlan_get_size,
3904 .fill_info = vxlan_fill_info,
1728d4fa 3905 .get_link_net = vxlan_get_link_net,
d342894c 3906};
3907
cf5da330
ND		 3908struct net_device *vxlan_dev_create(struct net *net, const char *name,
3909 u8 name_assign_type,
3910 struct vxlan_config *conf)
3911{
3912 struct nlattr *tb[IFLA_MAX + 1];
3913 struct net_device *dev;
3914 int err;
3915
3916 memset(&tb, 0, sizeof(tb));
3917
3918 dev = rtnl_create_link(net, name, name_assign_type,
3919 &vxlan_link_ops, tb);
3920 if (IS_ERR(dev))
3921 return dev;
3922
653ef6a3 3923 err = __vxlan_dev_create(net, dev, conf, NULL);
cf5da330
ND		 3924 if (err < 0) {
3925 free_netdev(dev);
3926 return ERR_PTR(err);
3927 }
3928
3929 err = rtnl_configure_link(dev, NULL);
3930 if (err < 0) {
3931 LIST_HEAD(list_kill);
3932
3933 vxlan_dellink(dev, &list_kill);
3934 unregister_netdevice_many(&list_kill);
3935 return ERR_PTR(err);
3936 }
3937
3938 return dev;
3939}
3940EXPORT_SYMBOL_GPL(vxlan_dev_create);
3941
acaf4e70
DB		 3942static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3943 struct net_device *dev)
3944{
3945 struct vxlan_dev *vxlan, *next;
3946 LIST_HEAD(list_kill);
3947
3948 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3949 struct vxlan_rdst *dst = &vxlan->default_dst;
3950
3951 /* In case we created vxlan device with carrier
3952 * and we loose the carrier due to module unload
3953 * we also need to remove vxlan device. In other
3954 * cases, it's not necessary and remote_ifindex
3955 * is 0 here, so no matches.
3956 */
3957 if (dst->remote_ifindex == dev->ifindex)
3958 vxlan_dellink(vxlan->dev, &list_kill);
3959 }
3960
3961 unregister_netdevice_many(&list_kill);
3962}
3963
b7aade15
HFS		 3964static int vxlan_netdevice_event(struct notifier_block *unused,
3965 unsigned long event, void *ptr)
acaf4e70
DB		 3966{
3967 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3968 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3969
04584957
SD		 3970 if (event == NETDEV_UNREGISTER) {
3971 vxlan_offload_rx_ports(dev, false);
acaf4e70 3972 vxlan_handle_lowerdev_unregister(vn, dev);
04584957
SD		 3973 } else if (event == NETDEV_REGISTER) {
3974 vxlan_offload_rx_ports(dev, true);
3975 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
3976 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
2d2b13fc 3977 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
04584957 3978 }
acaf4e70
DB		 3979
3980 return NOTIFY_DONE;
3981}
3982
3983static struct notifier_block vxlan_notifier_block __read_mostly = {
b7aade15 3984 .notifier_call = vxlan_netdevice_event,
acaf4e70
DB		 3985};
3986
d342894c 3987static __net_init int vxlan_init_net(struct net *net)
3988{
3989 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3990 unsigned int h;
d342894c 3991
553675fb 3992 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3993 spin_lock_init(&vn->sock_lock);
d342894c 3994
553675fb 3995 for (h = 0; h < PORT_HASH_SIZE; ++h)
3996 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3997
3998 return 0;
3999}
4000
d57420a1
JV		 4001#ifdef CONFIG_SYSCTL
4002static struct ctl_table_header *vxlan_fan_header;
4003static unsigned int vxlan_fan_version = 4;
4004
4005static struct ctl_table vxlan_fan_sysctls[] = {
4006 {
4007 .procname = "vxlan",
4008 .data = &vxlan_fan_version,
4009 .maxlen = sizeof(vxlan_fan_version),
4010 .mode = 0444,
4011 .proc_handler = proc_dointvec,
4012 },
4013 {},
4014};
4015#endif /* CONFIG_SYSCTL */
4016
f01ec1c0
ND		 4017static void __net_exit vxlan_exit_net(struct net *net)
4018{
4019 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4020 struct vxlan_dev *vxlan, *next;
4021 struct net_device *dev, *aux;
0e4ec5ac 4022 unsigned int h;
f01ec1c0
ND		 4023 LIST_HEAD(list);
4024
4025 rtnl_lock();
4026 for_each_netdev_safe(net, dev, aux)
4027 if (dev->rtnl_link_ops == &vxlan_link_ops)
4028 unregister_netdevice_queue(dev, &list);
4029
4030 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4031 /* If vxlan->dev is in the same netns, it has already been added
4032 * to the list by the previous loop.
4033 */
7c3dbea2 4034 if (!net_eq(dev_net(vxlan->dev), net))
13c3ed6a 4035 unregister_netdevice_queue(vxlan->dev, &list);
f01ec1c0
ND		 4036 }
4037
4038 unregister_netdevice_many(&list);
4039 rtnl_unlock();
0e4ec5ac
VA		 4040
4041 for (h = 0; h < PORT_HASH_SIZE; ++h)
4042 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
f01ec1c0
ND		 4043}
4044
d342894c 4045static struct pernet_operations vxlan_net_ops = {
4046 .init = vxlan_init_net,
f01ec1c0 4047 .exit = vxlan_exit_net,
d342894c 4048 .id = &vxlan_net_id,
4049 .size = sizeof(struct vxlan_net),
4050};
4051
4052static int __init vxlan_init_module(void)
4053{
4054 int rc;
4055
4056 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4057
783c1463 4058 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 4059 if (rc)
4060 goto out1;
4061
acaf4e70 4062 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 4063 if (rc)
4064 goto out2;
4065
acaf4e70
DB		 4066 rc = rtnl_link_register(&vxlan_link_ops);
4067 if (rc)
4068 goto out3;
d342894c 4069
d57420a1
JV		 4070#ifdef CONFIG_SYSCTL
4071 vxlan_fan_header = register_net_sysctl(&init_net, "net/fan",
4072 vxlan_fan_sysctls);
4073 if (!vxlan_fan_header) {
4074 rc = -ENOMEM;
4075 goto sysctl_failed;
4076 }
4077#endif /* CONFIG_SYSCTL */
4078
acaf4e70 4079 return 0;
d57420a1
JV		 4080#ifdef CONFIG_SYSCTL
4081sysctl_failed:
4082 rtnl_link_unregister(&vxlan_link_ops);
4083#endif /* CONFIG_SYSCTL */
acaf4e70
DB		 4084out3:
4085 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 4086out2:
783c1463 4087 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 4088out1:
4089 return rc;
4090}
7332a13b 4091late_initcall(vxlan_init_module);
d342894c 4092
4093static void __exit vxlan_cleanup_module(void)
4094{
d57420a1
JV		 4095#ifdef CONFIG_SYSCTL
4096 unregister_net_sysctl_table(vxlan_fan_header);
4097#endif /* CONFIG_SYSCTL */
b7153984 4098 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 4099 unregister_netdevice_notifier(&vxlan_notifier_block);
783c1463
DB		 4100 unregister_pernet_subsys(&vxlan_net_ops);
4101 /* rcu_barrier() is called by netns */
d342894c 4102}
4103module_exit(vxlan_cleanup_module);
4104
4105MODULE_LICENSE("GPL");
4106MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 4107MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 4108MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 4109MODULE_ALIAS_RTNL_LINK("vxlan");
ubuntu-bionic-kernel mirror