]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - drivers/net/vxlan.c
Merge tag 'mmc-v4.15-2' of git://git.kernel.org/pub/scm/linux/kernel/git/ulfh/mmc
[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>
d342894c 19#include <linux/if_ether.h>
1b13c97f 20#include <linux/ethtool.h>
e4f67add
DS
21#include <net/arp.h>
22#include <net/ndisc.h>
d342894c 23#include <net/ip.h>
24#include <net/icmp.h>
d342894c 25#include <net/rtnetlink.h>
d342894c 26#include <net/inet_ecn.h>
27#include <net/net_namespace.h>
28#include <net/netns/generic.h>
fa20e0e3 29#include <net/tun_proto.h>
012a5729 30#include <net/vxlan.h>
40d29af0 31
e4c7ed41 32#if IS_ENABLED(CONFIG_IPV6)
e4c7ed41 33#include <net/ip6_tunnel.h>
660d98ca 34#include <net/ip6_checksum.h>
e4c7ed41 35#endif
d342894c 36
37#define VXLAN_VERSION "0.1"
38
553675fb 39#define PORT_HASH_BITS 8
40#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 41#define FDB_AGE_DEFAULT 300 /* 5 min */
42#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
43
23c578bf 44/* UDP port for VXLAN traffic.
45 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 46 * for compatibility with early adopters.
23c578bf 47 */
9daaa397
SH
48static unsigned short vxlan_port __read_mostly = 8472;
49module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 50MODULE_PARM_DESC(udp_port, "Destination UDP port");
51
52static bool log_ecn_error = true;
53module_param(log_ecn_error, bool, 0644);
54MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
55
c7d03a00 56static unsigned int vxlan_net_id;
0dfbdf41 57static struct rtnl_link_ops vxlan_link_ops;
553675fb 58
7256eac1 59static const u8 all_zeros_mac[ETH_ALEN + 2];
afbd8bae 60
205f356d 61static int vxlan_sock_add(struct vxlan_dev *vxlan);
614732ea 62
a53cb29b
MB
63static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
64
553675fb 65/* per-network namespace private data for this module */
66struct vxlan_net {
67 struct list_head vxlan_list;
68 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 69 spinlock_t sock_lock;
553675fb 70};
71
d342894c 72/* Forwarding table entry */
73struct vxlan_fdb {
74 struct hlist_node hlist; /* linked list of entries */
75 struct rcu_head rcu;
76 unsigned long updated; /* jiffies */
77 unsigned long used;
3e61aa8f 78 struct list_head remotes;
7177a3b0 79 u8 eth_addr[ETH_ALEN];
d342894c 80 u16 state; /* see ndm_state */
3ad7a4b1 81 __be32 vni;
ae884082 82 u8 flags; /* see ndm_flags */
d342894c 83};
84
d342894c 85/* salt for hash table */
86static u32 vxlan_salt __read_mostly;
87
ee122c79
TG
88static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
89{
e7030878
TG
90 return vs->flags & VXLAN_F_COLLECT_METADATA ||
91 ip_tunnel_collect_metadata();
ee122c79
TG
92}
93
e4c7ed41
CW
94#if IS_ENABLED(CONFIG_IPV6)
95static inline
96bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
97{
f0ef3126
JB
98 if (a->sa.sa_family != b->sa.sa_family)
99 return false;
100 if (a->sa.sa_family == AF_INET6)
101 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
102 else
103 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW
104}
105
106static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
107{
f0ef3126
JB
108 if (ipa->sa.sa_family == AF_INET6)
109 return ipv6_addr_any(&ipa->sin6.sin6_addr);
110 else
111 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW
112}
113
114static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
115{
f0ef3126
JB
116 if (ipa->sa.sa_family == AF_INET6)
117 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
118 else
119 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW
120}
121
122static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
123{
f0ef3126 124 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 125 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB
126 ip->sa.sa_family = AF_INET6;
127 return 0;
128 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 129 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB
130 ip->sa.sa_family = AF_INET;
131 return 0;
132 } else {
133 return -EAFNOSUPPORT;
134 }
e4c7ed41
CW
135}
136
137static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 138 const union vxlan_addr *ip)
e4c7ed41 139{
f0ef3126 140 if (ip->sa.sa_family == AF_INET6)
930345ea 141 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 142 else
930345ea 143 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW
144}
145
146#else /* !CONFIG_IPV6 */
147
148static inline
149bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
150{
f0ef3126 151 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW
152}
153
154static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
155{
f0ef3126 156 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW
157}
158
159static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
160{
f0ef3126 161 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW
162}
163
164static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
165{
f0ef3126
JB
166 if (nla_len(nla) >= sizeof(struct in6_addr)) {
167 return -EAFNOSUPPORT;
168 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 169 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB
170 ip->sa.sa_family = AF_INET;
171 return 0;
172 } else {
173 return -EAFNOSUPPORT;
174 }
e4c7ed41
CW
175}
176
177static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 178 const union vxlan_addr *ip)
e4c7ed41 179{
930345ea 180 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW
181}
182#endif
183
553675fb 184/* Virtual Network hash table head */
54bfd872 185static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 186{
54bfd872 187 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 188}
189
190/* Socket hash table head */
191static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 192{
193 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
194
553675fb 195 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
196}
197
3e61aa8f
SH
198/* First remote destination for a forwarding entry.
199 * Guaranteed to be non-NULL because remotes are never deleted.
200 */
5ca5461c 201static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 202{
5ca5461c 203 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
204}
205
206static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
207{
208 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH
209}
210
ac5132d1
TG
211/* Find VXLAN socket based on network namespace, address family and UDP port
212 * and enabled unshareable flags.
213 */
214static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
215 __be16 port, u32 flags)
553675fb 216{
217 struct vxlan_sock *vs;
af33c1ad
TH
218
219 flags &= VXLAN_F_RCV_FLAGS;
553675fb 220
221 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 222 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 223 vxlan_get_sk_family(vs) == family &&
af33c1ad 224 vs->flags == flags)
553675fb 225 return vs;
226 }
227 return NULL;
d342894c 228}
229
49f810f0
MS
230static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
231 __be32 vni)
d342894c 232{
69e76661 233 struct vxlan_dev_node *node;
d342894c 234
b9167b2e
JB
235 /* For flow based devices, map all packets to VNI 0 */
236 if (vs->flags & VXLAN_F_COLLECT_METADATA)
237 vni = 0;
238
69e76661
JB
239 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
240 if (node->vxlan->default_dst.remote_vni != vni)
49f810f0
MS
241 continue;
242
243 if (IS_ENABLED(CONFIG_IPV6)) {
69e76661 244 const struct vxlan_config *cfg = &node->vxlan->cfg;
49f810f0
MS
245
246 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
247 cfg->remote_ifindex != ifindex)
248 continue;
249 }
250
69e76661 251 return node->vxlan;
d342894c 252 }
253
254 return NULL;
255}
256
5cfccc5a 257/* Look up VNI in a per net namespace table */
49f810f0
MS
258static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
259 __be32 vni, sa_family_t family,
260 __be16 port, u32 flags)
5cfccc5a
PS
261{
262 struct vxlan_sock *vs;
263
ac5132d1 264 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS
265 if (!vs)
266 return NULL;
267
49f810f0 268 return vxlan_vs_find_vni(vs, ifindex, vni);
5cfccc5a
PS
269}
270
d342894c 271/* Fill in neighbour message in skbuff. */
272static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH
273 const struct vxlan_fdb *fdb,
274 u32 portid, u32 seq, int type, unsigned int flags,
275 const struct vxlan_rdst *rdst)
d342894c 276{
277 unsigned long now = jiffies;
278 struct nda_cacheinfo ci;
279 struct nlmsghdr *nlh;
280 struct ndmsg *ndm;
e4f67add 281 bool send_ip, send_eth;
d342894c 282
283 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
284 if (nlh == NULL)
285 return -EMSGSIZE;
286
287 ndm = nlmsg_data(nlh);
288 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS
289
290 send_eth = send_ip = true;
291
292 if (type == RTM_GETNEIGH) {
e4c7ed41 293 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add 294 send_eth = !is_zero_ether_addr(fdb->eth_addr);
8f48ba71 295 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
e4f67add
DS
296 } else
297 ndm->ndm_family = AF_BRIDGE;
d342894c 298 ndm->ndm_state = fdb->state;
299 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 300 ndm->ndm_flags = fdb->flags;
545469f7 301 ndm->ndm_type = RTN_UNICAST;
d342894c 302
193523bf 303 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 304 nla_put_s32(skb, NDA_LINK_NETNSID,
38f507f1 305 peernet2id(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND
306 goto nla_put_failure;
307
e4f67add 308 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 309 goto nla_put_failure;
310
e4c7ed41 311 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS
312 goto nla_put_failure;
313
0dfbdf41 314 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS
315 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
316 goto nla_put_failure;
c7995c43 317 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
54bfd872 318 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
6681712d 319 goto nla_put_failure;
dc5321d7 320 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
3ad7a4b1
RP
321 nla_put_u32(skb, NDA_SRC_VNI,
322 be32_to_cpu(fdb->vni)))
323 goto nla_put_failure;
6681712d
DS
324 if (rdst->remote_ifindex &&
325 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 326 goto nla_put_failure;
327
328 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
329 ci.ndm_confirmed = 0;
330 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
331 ci.ndm_refcnt = 0;
332
333 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
334 goto nla_put_failure;
335
053c095a
JB
336 nlmsg_end(skb, nlh);
337 return 0;
d342894c 338
339nla_put_failure:
340 nlmsg_cancel(skb, nlh);
341 return -EMSGSIZE;
342}
343
344static inline size_t vxlan_nlmsg_size(void)
345{
346 return NLMSG_ALIGN(sizeof(struct ndmsg))
347 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 348 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 349 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS
350 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
351 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 352 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 353 + nla_total_size(sizeof(struct nda_cacheinfo));
354}
355
9e4b93f9
ND
356static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
357 struct vxlan_rdst *rd, int type)
d342894c 358{
359 struct net *net = dev_net(vxlan->dev);
360 struct sk_buff *skb;
361 int err = -ENOBUFS;
362
363 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
364 if (skb == NULL)
365 goto errout;
366
9e4b93f9 367 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 368 if (err < 0) {
369 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
370 WARN_ON(err == -EMSGSIZE);
371 kfree_skb(skb);
372 goto errout;
373 }
374
375 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
376 return;
377errout:
378 if (err < 0)
379 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
380}
381
e4c7ed41 382static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS
383{
384 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH
385 struct vxlan_fdb f = {
386 .state = NUD_STALE,
387 };
388 struct vxlan_rdst remote = {
e4c7ed41 389 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 390 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 391 };
3e61aa8f 392
9e4b93f9 393 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS
394}
395
396static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
397{
bb3fd687
SH
398 struct vxlan_fdb f = {
399 .state = NUD_STALE,
400 };
9e4b93f9 401 struct vxlan_rdst remote = { };
e4f67add 402
e4f67add
DS
403 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
404
9e4b93f9 405 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS
406}
407
d342894c 408/* Hash Ethernet address */
409static u32 eth_hash(const unsigned char *addr)
410{
411 u64 value = get_unaligned((u64 *)addr);
412
413 /* only want 6 bytes */
414#ifdef __BIG_ENDIAN
d342894c 415 value >>= 16;
321fb991 416#else
417 value <<= 16;
d342894c 418#endif
419 return hash_64(value, FDB_HASH_BITS);
420}
421
3ad7a4b1
RP
422static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
423{
424 /* use 1 byte of OUI and 3 bytes of NIC */
425 u32 key = get_unaligned((u32 *)(addr + 2));
426
427 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
428}
429
d342894c 430/* Hash chain to use given mac address */
431static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
3ad7a4b1 432 const u8 *mac, __be32 vni)
d342894c 433{
dc5321d7 434 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
3ad7a4b1
RP
435 return &vxlan->fdb_head[eth_vni_hash(mac, vni)];
436 else
437 return &vxlan->fdb_head[eth_hash(mac)];
d342894c 438}
439
440/* Look up Ethernet address in forwarding table */
014be2c8 441static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 442 const u8 *mac, __be32 vni)
d342894c 443{
3ad7a4b1 444 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
d342894c 445 struct vxlan_fdb *f;
d342894c 446
b67bfe0d 447 hlist_for_each_entry_rcu(f, head, hlist) {
3ad7a4b1 448 if (ether_addr_equal(mac, f->eth_addr)) {
dc5321d7 449 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP
450 if (vni == f->vni)
451 return f;
452 } else {
453 return f;
454 }
455 }
d342894c 456 }
457
458 return NULL;
459}
460
014be2c8 461static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 462 const u8 *mac, __be32 vni)
014be2c8
SS
463{
464 struct vxlan_fdb *f;
465
3ad7a4b1 466 f = __vxlan_find_mac(vxlan, mac, vni);
014be2c8
SS
467 if (f)
468 f->used = jiffies;
469
470 return f;
471}
472
a5e7c10a
MR
473/* caller should hold vxlan->hash_lock */
474static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 475 union vxlan_addr *ip, __be16 port,
54bfd872 476 __be32 vni, __u32 ifindex)
6681712d 477{
3e61aa8f 478 struct vxlan_rdst *rd;
6681712d 479
3e61aa8f 480 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 481 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS
482 rd->remote_port == port &&
483 rd->remote_vni == vni &&
484 rd->remote_ifindex == ifindex)
a5e7c10a 485 return rd;
6681712d 486 }
3e61aa8f 487
a5e7c10a
MR
488 return NULL;
489}
490
906dc186
TR
491/* Replace destination of unicast mac */
492static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872
JB
493 union vxlan_addr *ip, __be16 port, __be32 vni,
494 __u32 ifindex)
906dc186
TR
495{
496 struct vxlan_rdst *rd;
497
498 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
499 if (rd)
500 return 0;
501
502 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
503 if (!rd)
504 return 0;
0c1d70af
PA
505
506 dst_cache_reset(&rd->dst_cache);
e4c7ed41 507 rd->remote_ip = *ip;
906dc186
TR
508 rd->remote_port = port;
509 rd->remote_vni = vni;
510 rd->remote_ifindex = ifindex;
511 return 1;
512}
513
a5e7c10a
MR
514/* Add/update destinations for multicast */
515static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 516 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 517 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR
518{
519 struct vxlan_rdst *rd;
520
521 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
522 if (rd)
523 return 0;
524
6681712d
DS
525 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
526 if (rd == NULL)
527 return -ENOBUFS;
0c1d70af
PA
528
529 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
530 kfree(rd);
531 return -ENOBUFS;
532 }
533
e4c7ed41 534 rd->remote_ip = *ip;
6681712d
DS
535 rd->remote_port = port;
536 rd->remote_vni = vni;
537 rd->remote_ifindex = ifindex;
3e61aa8f
SH
538
539 list_add_tail_rcu(&rd->list, &f->remotes);
540
9e4b93f9 541 *rdp = rd;
6681712d
DS
542 return 1;
543}
544
dfd8645e
TH
545static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
546 unsigned int off,
547 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB
548 __be32 vni_field,
549 struct gro_remcsum *grc,
0ace2ca8 550 bool nopartial)
dfd8645e 551{
b7fe10e5 552 size_t start, offset;
dfd8645e
TH
553
554 if (skb->remcsum_offload)
b7fe10e5 555 return vh;
dfd8645e
TH
556
557 if (!NAPI_GRO_CB(skb)->csum_valid)
558 return NULL;
559
54bfd872
JB
560 start = vxlan_rco_start(vni_field);
561 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 562
b7fe10e5
TH
563 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
564 start, offset, grc, nopartial);
dfd8645e
TH
565
566 skb->remcsum_offload = 1;
567
568 return vh;
569}
570
5602c48c
TH
571static struct sk_buff **vxlan_gro_receive(struct sock *sk,
572 struct sk_buff **head,
573 struct sk_buff *skb)
dc01e7d3
OG
574{
575 struct sk_buff *p, **pp = NULL;
576 struct vxlanhdr *vh, *vh2;
9b174d88 577 unsigned int hlen, off_vx;
dc01e7d3 578 int flush = 1;
5602c48c 579 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
54bfd872 580 __be32 flags;
26c4f7da
TH
581 struct gro_remcsum grc;
582
583 skb_gro_remcsum_init(&grc);
dc01e7d3
OG
584
585 off_vx = skb_gro_offset(skb);
586 hlen = off_vx + sizeof(*vh);
587 vh = skb_gro_header_fast(skb, off_vx);
588 if (skb_gro_header_hard(skb, hlen)) {
589 vh = skb_gro_header_slow(skb, hlen, off_vx);
590 if (unlikely(!vh))
591 goto out;
592 }
dc01e7d3 593
dfd8645e
TH
594 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
595
54bfd872 596 flags = vh->vx_flags;
dfd8645e
TH
597
598 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
599 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 600 vh->vx_vni, &grc,
0ace2ca8
TH
601 !!(vs->flags &
602 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH
603
604 if (!vh)
605 goto out;
606 }
607
b7fe10e5
TH
608 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
609
dc01e7d3
OG
610 for (p = *head; p; p = p->next) {
611 if (!NAPI_GRO_CB(p)->same_flow)
612 continue;
613
614 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG
615 if (vh->vx_flags != vh2->vx_flags ||
616 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG
617 NAPI_GRO_CB(p)->same_flow = 0;
618 continue;
619 }
dc01e7d3
OG
620 }
621
fcd91dd4 622 pp = call_gro_receive(eth_gro_receive, head, skb);
c194cf93 623 flush = 0;
dc01e7d3 624
dc01e7d3 625out:
26c4f7da 626 skb_gro_remcsum_cleanup(skb, &grc);
be73b304 627 skb->remcsum_offload = 0;
dc01e7d3
OG
628 NAPI_GRO_CB(skb)->flush |= flush;
629
630 return pp;
631}
632
5602c48c 633static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
dc01e7d3 634{
229740c6
JR
635 /* Sets 'skb->inner_mac_header' since we are always called with
636 * 'skb->encapsulation' set.
637 */
9b174d88 638 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG
639}
640
d342894c 641/* Add new entry to forwarding table -- assumes lock held */
642static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 643 const u8 *mac, union vxlan_addr *ip,
6681712d 644 __u16 state, __u16 flags,
3ad7a4b1
RP
645 __be16 port, __be32 src_vni, __be32 vni,
646 __u32 ifindex, __u8 ndm_flags)
d342894c 647{
9e4b93f9 648 struct vxlan_rdst *rd = NULL;
d342894c 649 struct vxlan_fdb *f;
650 int notify = 0;
17b46365 651 int rc;
d342894c 652
3ad7a4b1 653 f = __vxlan_find_mac(vxlan, mac, src_vni);
d342894c 654 if (f) {
655 if (flags & NLM_F_EXCL) {
656 netdev_dbg(vxlan->dev,
657 "lost race to create %pM\n", mac);
658 return -EEXIST;
659 }
660 if (f->state != state) {
661 f->state = state;
662 f->updated = jiffies;
663 notify = 1;
664 }
ae884082
DS
665 if (f->flags != ndm_flags) {
666 f->flags = ndm_flags;
667 f->updated = jiffies;
668 notify = 1;
669 }
906dc186
TR
670 if ((flags & NLM_F_REPLACE)) {
671 /* Only change unicasts */
672 if (!(is_multicast_ether_addr(f->eth_addr) ||
673 is_zero_ether_addr(f->eth_addr))) {
60840429 674 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 675 ifindex);
906dc186
TR
676 } else
677 return -EOPNOTSUPP;
678 }
6681712d 679 if ((flags & NLM_F_APPEND) &&
58e4c767
MR
680 (is_multicast_ether_addr(f->eth_addr) ||
681 is_zero_ether_addr(f->eth_addr))) {
17b46365 682 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
6681712d
DS
683
684 if (rc < 0)
685 return rc;
686 notify |= rc;
687 }
d342894c 688 } else {
689 if (!(flags & NLM_F_CREATE))
690 return -ENOENT;
691
0dfbdf41
TG
692 if (vxlan->cfg.addrmax &&
693 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 694 return -ENOSPC;
695
906dc186
TR
696 /* Disallow replace to add a multicast entry */
697 if ((flags & NLM_F_REPLACE) &&
698 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
699 return -EOPNOTSUPP;
700
e4c7ed41 701 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 702 f = kmalloc(sizeof(*f), GFP_ATOMIC);
703 if (!f)
704 return -ENOMEM;
705
706 notify = 1;
d342894c 707 f->state = state;
ae884082 708 f->flags = ndm_flags;
d342894c 709 f->updated = f->used = jiffies;
3ad7a4b1 710 f->vni = src_vni;
3e61aa8f 711 INIT_LIST_HEAD(&f->remotes);
d342894c 712 memcpy(f->eth_addr, mac, ETH_ALEN);
713
17b46365
HY
714 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
715 if (rc < 0) {
716 kfree(f);
717 return rc;
718 }
3e61aa8f 719
d342894c 720 ++vxlan->addrcnt;
721 hlist_add_head_rcu(&f->hlist,
3ad7a4b1 722 vxlan_fdb_head(vxlan, mac, src_vni));
d342894c 723 }
724
9e4b93f9
ND
725 if (notify) {
726 if (rd == NULL)
727 rd = first_remote_rtnl(f);
728 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
729 }
d342894c 730
731 return 0;
732}
733
6706c82e 734static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS
735{
736 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 737 struct vxlan_rdst *rd, *nd;
6681712d 738
0c1d70af
PA
739 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
740 dst_cache_destroy(&rd->dst_cache);
6681712d 741 kfree(rd);
0c1d70af 742 }
6681712d
DS
743 kfree(f);
744}
745
d342894c 746static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
747{
748 netdev_dbg(vxlan->dev,
749 "delete %pM\n", f->eth_addr);
750
751 --vxlan->addrcnt;
9e4b93f9 752 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 753
754 hlist_del_rcu(&f->hlist);
6681712d 755 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 756}
757
35cf2845
LR
758static void vxlan_dst_free(struct rcu_head *head)
759{
760 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
761
762 dst_cache_destroy(&rd->dst_cache);
763 kfree(rd);
764}
765
766static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
767 struct vxlan_rdst *rd)
768{
769 list_del_rcu(&rd->list);
770 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
771 call_rcu(&rd->rcu, vxlan_dst_free);
772}
773
f0b074be 774static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
3ad7a4b1
RP
775 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
776 __be32 *vni, u32 *ifindex)
d342894c 777{
6681712d 778 struct net *net = dev_net(vxlan->dev);
e4c7ed41 779 int err;
d342894c 780
f0b074be 781 if (tb[NDA_DST]) {
e4c7ed41
CW
782 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
783 if (err)
784 return err;
f0b074be 785 } else {
e4c7ed41
CW
786 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
787 if (remote->sa.sa_family == AF_INET) {
788 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
789 ip->sa.sa_family = AF_INET;
790#if IS_ENABLED(CONFIG_IPV6)
791 } else {
792 ip->sin6.sin6_addr = in6addr_any;
793 ip->sa.sa_family = AF_INET6;
794#endif
795 }
f0b074be 796 }
d342894c 797
6681712d 798 if (tb[NDA_PORT]) {
73cf3317 799 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 800 return -EINVAL;
f0b074be
MR
801 *port = nla_get_be16(tb[NDA_PORT]);
802 } else {
0dfbdf41 803 *port = vxlan->cfg.dst_port;
f0b074be 804 }
6681712d
DS
805
806 if (tb[NDA_VNI]) {
807 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
808 return -EINVAL;
54bfd872 809 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR
810 } else {
811 *vni = vxlan->default_dst.remote_vni;
812 }
6681712d 813
3ad7a4b1
RP
814 if (tb[NDA_SRC_VNI]) {
815 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
816 return -EINVAL;
817 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
818 } else {
819 *src_vni = vxlan->default_dst.remote_vni;
820 }
821
6681712d 822 if (tb[NDA_IFINDEX]) {
5abb0029 823 struct net_device *tdev;
6681712d
DS
824
825 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
826 return -EINVAL;
f0b074be 827 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 828 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 829 if (!tdev)
6681712d 830 return -EADDRNOTAVAIL;
f0b074be
MR
831 } else {
832 *ifindex = 0;
833 }
834
835 return 0;
836}
837
838/* Add static entry (via netlink) */
839static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
840 struct net_device *dev,
f6f6424b 841 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR
842{
843 struct vxlan_dev *vxlan = netdev_priv(dev);
844 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 845 union vxlan_addr ip;
f0b074be 846 __be16 port;
3ad7a4b1 847 __be32 src_vni, vni;
54bfd872 848 u32 ifindex;
f0b074be
MR
849 int err;
850
851 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
852 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
853 ndm->ndm_state);
854 return -EINVAL;
855 }
856
857 if (tb[NDA_DST] == NULL)
858 return -EINVAL;
859
3ad7a4b1 860 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
f0b074be
MR
861 if (err)
862 return err;
6681712d 863
5933a7bb
MR
864 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
865 return -EAFNOSUPPORT;
866
d342894c 867 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 868 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
3ad7a4b1 869 port, src_vni, vni, ifindex, ndm->ndm_flags);
d342894c 870 spin_unlock_bh(&vxlan->hash_lock);
871
872 return err;
873}
874
3ad7a4b1
RP
875static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
876 const unsigned char *addr, union vxlan_addr ip,
fc39c38b
XL
877 __be16 port, __be32 src_vni, __be32 vni,
878 u32 ifindex, u16 vid)
d342894c 879{
d342894c 880 struct vxlan_fdb *f;
bc7892ba 881 struct vxlan_rdst *rd = NULL;
3ad7a4b1 882 int err = -ENOENT;
bc7892ba 883
3ad7a4b1 884 f = vxlan_find_mac(vxlan, addr, src_vni);
bc7892ba 885 if (!f)
3ad7a4b1 886 return err;
bc7892ba 887
e4c7ed41
CW
888 if (!vxlan_addr_any(&ip)) {
889 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR
890 if (!rd)
891 goto out;
892 }
893
bc7892ba
MR
894 /* remove a destination if it's not the only one on the list,
895 * otherwise destroy the fdb entry
896 */
897 if (rd && !list_is_singular(&f->remotes)) {
35cf2845 898 vxlan_fdb_dst_destroy(vxlan, f, rd);
bc7892ba 899 goto out;
d342894c 900 }
bc7892ba
MR
901
902 vxlan_fdb_destroy(vxlan, f);
903
904out:
3ad7a4b1
RP
905 return 0;
906}
907
908/* Delete entry (via netlink) */
909static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
910 struct net_device *dev,
911 const unsigned char *addr, u16 vid)
912{
913 struct vxlan_dev *vxlan = netdev_priv(dev);
914 union vxlan_addr ip;
915 __be32 src_vni, vni;
916 __be16 port;
917 u32 ifindex;
918 int err;
919
920 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
921 if (err)
922 return err;
923
924 spin_lock_bh(&vxlan->hash_lock);
925 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
926 vid);
d342894c 927 spin_unlock_bh(&vxlan->hash_lock);
928
929 return err;
930}
931
932/* Dump forwarding table */
933static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 934 struct net_device *dev,
d297653d 935 struct net_device *filter_dev, int *idx)
d342894c 936{
937 struct vxlan_dev *vxlan = netdev_priv(dev);
938 unsigned int h;
d297653d 939 int err = 0;
d342894c 940
941 for (h = 0; h < FDB_HASH_SIZE; ++h) {
942 struct vxlan_fdb *f;
d342894c 943
b67bfe0d 944 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 945 struct vxlan_rdst *rd;
6681712d 946
3e61aa8f 947 list_for_each_entry_rcu(rd, &f->remotes, list) {
d297653d 948 if (*idx < cb->args[2])
07a51cd3
AW
949 goto skip;
950
6681712d
DS
951 err = vxlan_fdb_info(skb, vxlan, f,
952 NETLINK_CB(cb->skb).portid,
953 cb->nlh->nlmsg_seq,
954 RTM_NEWNEIGH,
955 NLM_F_MULTI, rd);
d297653d 956 if (err < 0)
3e61aa8f 957 goto out;
3e61aa8f 958skip:
d297653d 959 *idx += 1;
07a51cd3 960 }
d342894c 961 }
962 }
3e61aa8f 963out:
d297653d 964 return err;
d342894c 965}
966
967/* Watch incoming packets to learn mapping between Ethernet address
968 * and Tunnel endpoint.
c4b49512 969 * Return true if packet is bogus and should be dropped.
d342894c 970 */
26a41ae6 971static bool vxlan_snoop(struct net_device *dev,
3ad7a4b1 972 union vxlan_addr *src_ip, const u8 *src_mac,
87613de9 973 u32 src_ifindex, __be32 vni)
d342894c 974{
975 struct vxlan_dev *vxlan = netdev_priv(dev);
976 struct vxlan_fdb *f;
87613de9
MS
977 u32 ifindex = 0;
978
979#if IS_ENABLED(CONFIG_IPV6)
980 if (src_ip->sa.sa_family == AF_INET6 &&
981 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
982 ifindex = src_ifindex;
983#endif
d342894c 984
3ad7a4b1 985 f = vxlan_find_mac(vxlan, src_mac, vni);
d342894c 986 if (likely(f)) {
5ca5461c 987 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 988
87613de9
MS
989 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
990 rdst->remote_ifindex == ifindex))
26a41ae6 991 return false;
992
993 /* Don't migrate static entries, drop packets */
e0090a9e 994 if (f->state & (NUD_PERMANENT | NUD_NOARP))
26a41ae6 995 return true;
d342894c 996
997 if (net_ratelimit())
998 netdev_info(dev,
e4c7ed41 999 "%pM migrated from %pIS to %pIS\n",
a4870f79 1000 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 1001
e4c7ed41 1002 rdst->remote_ip = *src_ip;
d342894c 1003 f->updated = jiffies;
9e4b93f9 1004 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 1005 } else {
1006 /* learned new entry */
1007 spin_lock(&vxlan->hash_lock);
3bf74b1a 1008
1009 /* close off race between vxlan_flush and incoming packets */
1010 if (netif_running(dev))
1011 vxlan_fdb_create(vxlan, src_mac, src_ip,
1012 NUD_REACHABLE,
1013 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 1014 vxlan->cfg.dst_port,
3ad7a4b1 1015 vni,
3bf74b1a 1016 vxlan->default_dst.remote_vni,
87613de9 1017 ifindex, NTF_SELF);
d342894c 1018 spin_unlock(&vxlan->hash_lock);
1019 }
26a41ae6 1020
1021 return false;
d342894c 1022}
1023
d342894c 1024/* See if multicast group is already in use by other ID */
95ab0991 1025static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1026{
553675fb 1027 struct vxlan_dev *vxlan;
c6fcc4fc 1028 struct vxlan_sock *sock4;
4053ab1b
AB
1029#if IS_ENABLED(CONFIG_IPV6)
1030 struct vxlan_sock *sock6;
1031#endif
b1be00a6 1032 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1033
c6fcc4fc 1034 sock4 = rtnl_dereference(dev->vn4_sock);
1035
95ab0991
G
1036 /* The vxlan_sock is only used by dev, leaving group has
1037 * no effect on other vxlan devices.
1038 */
66af846f 1039 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
95ab0991 1040 return false;
b1be00a6 1041#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1042 sock6 = rtnl_dereference(dev->vn6_sock);
66af846f 1043 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
b1be00a6
JB
1044 return false;
1045#endif
95ab0991 1046
553675fb 1047 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1048 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1049 continue;
d342894c 1050
c6fcc4fc 1051 if (family == AF_INET &&
1052 rtnl_dereference(vxlan->vn4_sock) != sock4)
95ab0991 1053 continue;
b1be00a6 1054#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1055 if (family == AF_INET6 &&
1056 rtnl_dereference(vxlan->vn6_sock) != sock6)
b1be00a6
JB
1057 continue;
1058#endif
95ab0991
G
1059
1060 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1061 &dev->default_dst.remote_ip))
1062 continue;
1063
1064 if (vxlan->default_dst.remote_ifindex !=
1065 dev->default_dst.remote_ifindex)
1066 continue;
1067
1068 return true;
553675fb 1069 }
d342894c 1070
1071 return false;
1072}
1073
544a773a 1074static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
7c47cedf 1075{
b1be00a6 1076 struct vxlan_net *vn;
012a5729 1077
b1be00a6 1078 if (!vs)
544a773a 1079 return false;
66af846f 1080 if (!refcount_dec_and_test(&vs->refcnt))
544a773a 1081 return false;
d342894c 1082
b1be00a6 1083 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1084 spin_lock(&vn->sock_lock);
7c47cedf 1085 hlist_del_rcu(&vs->hlist);
e7b3db5e 1086 udp_tunnel_notify_del_rx_port(vs->sock,
b9adcd69
AD
1087 (vs->flags & VXLAN_F_GPE) ?
1088 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 1089 UDP_TUNNEL_TYPE_VXLAN);
1c51a915
SH
1090 spin_unlock(&vn->sock_lock);
1091
544a773a 1092 return true;
d342894c 1093}
1094
b1be00a6
JB
1095static void vxlan_sock_release(struct vxlan_dev *vxlan)
1096{
c6fcc4fc 1097 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
b1be00a6 1098#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1099 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1100
57d88182 1101 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
544a773a
HFS
1102#endif
1103
57d88182 1104 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
544a773a
HFS
1105 synchronize_net();
1106
a53cb29b
MB
1107 vxlan_vs_del_dev(vxlan);
1108
c6fcc4fc 1109 if (__vxlan_sock_release_prep(sock4)) {
1110 udp_tunnel_sock_release(sock4->sock);
1111 kfree(sock4);
544a773a
HFS
1112 }
1113
1114#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1115 if (__vxlan_sock_release_prep(sock6)) {
1116 udp_tunnel_sock_release(sock6->sock);
1117 kfree(sock6);
544a773a 1118 }
b1be00a6
JB
1119#endif
1120}
1121
56ef9c90 1122/* Update multicast group membership when first VNI on
c4b49512 1123 * multicast address is brought up
7c47cedf 1124 */
56ef9c90 1125static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1126{
b1be00a6 1127 struct sock *sk;
e4c7ed41
CW
1128 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1129 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1130 int ret = -EINVAL;
d342894c 1131
e4c7ed41 1132 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1133 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW
1134 struct ip_mreqn mreq = {
1135 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1136 .imr_ifindex = ifindex,
1137 };
1138
c6fcc4fc 1139 sk = sock4->sock->sk;
b1be00a6 1140 lock_sock(sk);
56ef9c90 1141 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1142 release_sock(sk);
e4c7ed41
CW
1143#if IS_ENABLED(CONFIG_IPV6)
1144 } else {
c6fcc4fc 1145 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1146
1147 sk = sock6->sock->sk;
b1be00a6 1148 lock_sock(sk);
56ef9c90
MRL
1149 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1150 &ip->sin6.sin6_addr);
b1be00a6 1151 release_sock(sk);
e4c7ed41
CW
1152#endif
1153 }
3fc2de2f 1154
56ef9c90 1155 return ret;
3fc2de2f 1156}
1157
1158/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1159static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1160{
b1be00a6 1161 struct sock *sk;
e4c7ed41
CW
1162 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1163 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1164 int ret = -EINVAL;
3fc2de2f 1165
e4c7ed41 1166 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1167 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW
1168 struct ip_mreqn mreq = {
1169 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1170 .imr_ifindex = ifindex,
1171 };
1172
c6fcc4fc 1173 sk = sock4->sock->sk;
b1be00a6 1174 lock_sock(sk);
56ef9c90 1175 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1176 release_sock(sk);
e4c7ed41
CW
1177#if IS_ENABLED(CONFIG_IPV6)
1178 } else {
c6fcc4fc 1179 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1180
1181 sk = sock6->sock->sk;
b1be00a6 1182 lock_sock(sk);
56ef9c90
MRL
1183 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1184 &ip->sin6.sin6_addr);
b1be00a6 1185 release_sock(sk);
e4c7ed41
CW
1186#endif
1187 }
d342894c 1188
56ef9c90 1189 return ret;
d342894c 1190}
1191
f14ecebb
JB
1192static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1193 struct sk_buff *skb, u32 vxflags)
dfd8645e 1194{
7d34fa75 1195 size_t start, offset;
dfd8645e 1196
f14ecebb
JB
1197 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1198 goto out;
b7fe10e5 1199
f14ecebb
JB
1200 start = vxlan_rco_start(unparsed->vx_vni);
1201 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1202
7d34fa75 1203 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1204 return false;
dfd8645e 1205
be5cfeab
JB
1206 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1207 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB
1208out:
1209 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1210 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1211 return true;
dfd8645e
TH
1212}
1213
f14ecebb 1214static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1215 struct sk_buff *skb, u32 vxflags,
10a5af23 1216 struct vxlan_metadata *md)
3288af08 1217{
f14ecebb 1218 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1219 struct metadata_dst *tun_dst;
f14ecebb
JB
1220
1221 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1222 goto out;
3288af08 1223
3288af08
JB
1224 md->gbp = ntohs(gbp->policy_id);
1225
10a5af23 1226 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1227 if (tun_dst) {
3288af08 1228 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM
1229 tun_dst->u.tun_info.options_len = sizeof(*md);
1230 }
3288af08
JB
1231 if (gbp->dont_learn)
1232 md->gbp |= VXLAN_GBP_DONT_LEARN;
1233
1234 if (gbp->policy_applied)
1235 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1236
64f87d36
JB
1237 /* In flow-based mode, GBP is carried in dst_metadata */
1238 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1239 skb->mark = md->gbp;
f14ecebb
JB
1240out:
1241 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB
1242}
1243
e1e5314d 1244static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
61618eea 1245 __be16 *protocol,
e1e5314d
JB
1246 struct sk_buff *skb, u32 vxflags)
1247{
1248 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1249
1250 /* Need to have Next Protocol set for interfaces in GPE mode. */
1251 if (!gpe->np_applied)
1252 return false;
1253 /* "The initial version is 0. If a receiver does not support the
1254 * version indicated it MUST drop the packet.
1255 */
1256 if (gpe->version != 0)
1257 return false;
1258 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1259 * processing MUST occur." However, we don't implement OAM
1260 * processing, thus drop the packet.
1261 */
1262 if (gpe->oam_flag)
1263 return false;
1264
fa20e0e3
JB
1265 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1266 if (!*protocol)
e1e5314d 1267 return false;
e1e5314d
JB
1268
1269 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1270 return true;
1271}
1272
1ab016e2
JB
1273static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1274 struct vxlan_sock *vs,
3ad7a4b1 1275 struct sk_buff *skb, __be32 vni)
614732ea 1276{
614732ea 1277 union vxlan_addr saddr;
87613de9 1278 u32 ifindex = skb->dev->ifindex;
614732ea 1279
614732ea 1280 skb_reset_mac_header(skb);
614732ea
TG
1281 skb->protocol = eth_type_trans(skb, vxlan->dev);
1282 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1283
1284 /* Ignore packet loops (and multicast echo) */
1285 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1286 return false;
614732ea 1287
760c6805 1288 /* Get address from the outer IP header */
ce212d0f 1289 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1290 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG
1291 saddr.sa.sa_family = AF_INET;
1292#if IS_ENABLED(CONFIG_IPV6)
1293 } else {
1ab016e2 1294 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG
1295 saddr.sa.sa_family = AF_INET6;
1296#endif
1297 }
1298
dc5321d7 1299 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
87613de9 1300 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1ab016e2
JB
1301 return false;
1302
1303 return true;
1304}
1305
760c6805
JB
1306static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1307 struct sk_buff *skb)
1308{
1309 int err = 0;
1310
1311 if (vxlan_get_sk_family(vs) == AF_INET)
1312 err = IP_ECN_decapsulate(oiph, skb);
1313#if IS_ENABLED(CONFIG_IPV6)
1314 else
1315 err = IP6_ECN_decapsulate(oiph, skb);
1316#endif
1317
1318 if (unlikely(err) && log_ecn_error) {
1319 if (vxlan_get_sk_family(vs) == AF_INET)
1320 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1321 &((struct iphdr *)oiph)->saddr,
1322 ((struct iphdr *)oiph)->tos);
1323 else
1324 net_info_ratelimited("non-ECT from %pI6\n",
1325 &((struct ipv6hdr *)oiph)->saddr);
1326 }
1327 return err <= 1;
1328}
1329
d342894c 1330/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1331static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1332{
f2d1968e 1333 struct pcpu_sw_netstats *stats;
c9e78efb 1334 struct vxlan_dev *vxlan;
5cfccc5a 1335 struct vxlan_sock *vs;
f14ecebb 1336 struct vxlanhdr unparsed;
ee122c79
TG
1337 struct vxlan_metadata _md;
1338 struct vxlan_metadata *md = &_md;
61618eea 1339 __be16 protocol = htons(ETH_P_TEB);
e1e5314d 1340 bool raw_proto = false;
f2d1968e 1341 void *oiph;
3ad7a4b1 1342 __be32 vni = 0;
d342894c 1343
e1e5314d 1344 /* Need UDP and VXLAN header to be present */
7ce04758 1345 if (!pskb_may_pull(skb, VXLAN_HLEN))
e5aed006 1346 goto drop;
d342894c 1347
f14ecebb 1348 unparsed = *vxlan_hdr(skb);
288b01c8
JB
1349 /* VNI flag always required to be set */
1350 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1351 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1352 ntohl(vxlan_hdr(skb)->vx_flags),
1353 ntohl(vxlan_hdr(skb)->vx_vni));
1354 /* Return non vxlan pkt */
e5aed006 1355 goto drop;
d342894c 1356 }
288b01c8
JB
1357 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1358 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1359
559835ea 1360 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1361 if (!vs)
d342894c 1362 goto drop;
d342894c 1363
3ad7a4b1
RP
1364 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1365
49f810f0 1366 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
c9e78efb
JB
1367 if (!vxlan)
1368 goto drop;
1369
e1e5314d
JB
1370 /* For backwards compatibility, only allow reserved fields to be
1371 * used by VXLAN extensions if explicitly requested.
1372 */
1373 if (vs->flags & VXLAN_F_GPE) {
1374 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1375 goto drop;
1376 raw_proto = true;
1377 }
1378
1379 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1380 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1381 goto drop;
c9e78efb 1382
ee122c79 1383 if (vxlan_collect_metadata(vs)) {
10a5af23 1384 struct metadata_dst *tun_dst;
07dabf20 1385
c29a70d2 1386 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
d817f432 1387 key32_to_tunnel_id(vni), sizeof(*md));
c29a70d2 1388
ee122c79
TG
1389 if (!tun_dst)
1390 goto drop;
1391
0f1b7354 1392 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB
1393
1394 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG
1395 } else {
1396 memset(md, 0, sizeof(*md));
1397 }
1398
f14ecebb
JB
1399 if (vs->flags & VXLAN_F_REMCSUM_RX)
1400 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1401 goto drop;
1402 if (vs->flags & VXLAN_F_GBP)
10a5af23 1403 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB
1404 /* Note that GBP and GPE can never be active together. This is
1405 * ensured in vxlan_dev_configure.
1406 */
3511494c 1407
f14ecebb 1408 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH
1409 /* If there are any unprocessed flags remaining treat
1410 * this as a malformed packet. This behavior diverges from
1411 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1412 * in reserved fields are to be ignored. The approach here
c4b49512 1413 * maintains compatibility with previous stack code, and also
3bf39475
TH
1414 * is more robust and provides a little more security in
1415 * adding extensions to VXLAN.
1416 */
288b01c8 1417 goto drop;
3bf39475
TH
1418 }
1419
e1e5314d 1420 if (!raw_proto) {
3ad7a4b1 1421 if (!vxlan_set_mac(vxlan, vs, skb, vni))
e1e5314d
JB
1422 goto drop;
1423 } else {
8be0cfa4 1424 skb_reset_mac_header(skb);
e1e5314d
JB
1425 skb->dev = vxlan->dev;
1426 skb->pkt_type = PACKET_HOST;
1427 }
f2d1968e 1428
f2d1968e
JB
1429 oiph = skb_network_header(skb);
1430 skb_reset_network_header(skb);
1431
1432 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1433 ++vxlan->dev->stats.rx_frame_errors;
1434 ++vxlan->dev->stats.rx_errors;
1435 goto drop;
1436 }
1437
1438 stats = this_cpu_ptr(vxlan->dev->tstats);
1439 u64_stats_update_begin(&stats->syncp);
1440 stats->rx_packets++;
1441 stats->rx_bytes += skb->len;
1442 u64_stats_update_end(&stats->syncp);
1443
1444 gro_cells_receive(&vxlan->gro_cells, skb);
5cfccc5a
PS
1445 return 0;
1446
1447drop:
288b01c8
JB
1448 /* Consume bad packet */
1449 kfree_skb(skb);
1450 return 0;
5cfccc5a
PS
1451}
1452
3ad7a4b1 1453static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
e4f67add
DS
1454{
1455 struct vxlan_dev *vxlan = netdev_priv(dev);
1456 struct arphdr *parp;
1457 u8 *arpptr, *sha;
1458 __be32 sip, tip;
1459 struct neighbour *n;
1460
1461 if (dev->flags & IFF_NOARP)
1462 goto out;
1463
1464 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1465 dev->stats.tx_dropped++;
1466 goto out;
1467 }
1468 parp = arp_hdr(skb);
1469
1470 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1471 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1472 parp->ar_pro != htons(ETH_P_IP) ||
1473 parp->ar_op != htons(ARPOP_REQUEST) ||
1474 parp->ar_hln != dev->addr_len ||
1475 parp->ar_pln != 4)
1476 goto out;
1477 arpptr = (u8 *)parp + sizeof(struct arphdr);
1478 sha = arpptr;
1479 arpptr += dev->addr_len; /* sha */
1480 memcpy(&sip, arpptr, sizeof(sip));
1481 arpptr += sizeof(sip);
1482 arpptr += dev->addr_len; /* tha */
1483 memcpy(&tip, arpptr, sizeof(tip));
1484
1485 if (ipv4_is_loopback(tip) ||
1486 ipv4_is_multicast(tip))
1487 goto out;
1488
1489 n = neigh_lookup(&arp_tbl, &tip, dev);
1490
1491 if (n) {
e4f67add
DS
1492 struct vxlan_fdb *f;
1493 struct sk_buff *reply;
1494
1495 if (!(n->nud_state & NUD_CONNECTED)) {
1496 neigh_release(n);
1497 goto out;
1498 }
1499
3ad7a4b1 1500 f = vxlan_find_mac(vxlan, n->ha, vni);
e4c7ed41 1501 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS
1502 /* bridge-local neighbor */
1503 neigh_release(n);
1504 goto out;
1505 }
1506
1507 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1508 n->ha, sha);
1509
1510 neigh_release(n);
1511
7346135d
DS
1512 if (reply == NULL)
1513 goto out;
1514
e4f67add
DS
1515 skb_reset_mac_header(reply);
1516 __skb_pull(reply, skb_network_offset(reply));
1517 reply->ip_summed = CHECKSUM_UNNECESSARY;
1518 reply->pkt_type = PACKET_HOST;
1519
1520 if (netif_rx_ni(reply) == NET_RX_DROP)
1521 dev->stats.rx_dropped++;
dc5321d7 1522 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
e4c7ed41
CW
1523 union vxlan_addr ipa = {
1524 .sin.sin_addr.s_addr = tip,
a45e92a5 1525 .sin.sin_family = AF_INET,
e4c7ed41
CW
1526 };
1527
1528 vxlan_ip_miss(dev, &ipa);
1529 }
e4f67add
DS
1530out:
1531 consume_skb(skb);
1532 return NETDEV_TX_OK;
1533}
1534
f564f45c 1535#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS
1536static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1537 struct neighbour *n, bool isrouter)
1538{
1539 struct net_device *dev = request->dev;
1540 struct sk_buff *reply;
1541 struct nd_msg *ns, *na;
1542 struct ipv6hdr *pip6;
1543 u8 *daddr;
1544 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1545 int ns_olen;
1546 int i, len;
1547
f1fb08f6 1548 if (dev == NULL || !pskb_may_pull(request, request->len))
4b29dba9
DS
1549 return NULL;
1550
1551 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1552 sizeof(*na) + na_olen + dev->needed_tailroom;
1553 reply = alloc_skb(len, GFP_ATOMIC);
1554 if (reply == NULL)
1555 return NULL;
1556
1557 reply->protocol = htons(ETH_P_IPV6);
1558 reply->dev = dev;
1559 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1560 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1561 skb_reset_mac_header(reply);
4b29dba9 1562
f1fb08f6 1563 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
4b29dba9
DS
1564
1565 daddr = eth_hdr(request)->h_source;
f1fb08f6
VB
1566 ns_olen = request->len - skb_network_offset(request) -
1567 sizeof(struct ipv6hdr) - sizeof(*ns);
4b29dba9
DS
1568 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1569 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1570 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1571 break;
1572 }
1573 }
1574
1575 /* Ethernet header */
1576 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1577 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1578 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1579 reply->protocol = htons(ETH_P_IPV6);
1580
1581 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1582 skb_reset_network_header(reply);
4b29dba9
DS
1583 skb_put(reply, sizeof(struct ipv6hdr));
1584
1585 /* IPv6 header */
1586
1587 pip6 = ipv6_hdr(reply);
1588 memset(pip6, 0, sizeof(struct ipv6hdr));
1589 pip6->version = 6;
1590 pip6->priority = ipv6_hdr(request)->priority;
1591 pip6->nexthdr = IPPROTO_ICMPV6;
1592 pip6->hop_limit = 255;
1593 pip6->daddr = ipv6_hdr(request)->saddr;
1594 pip6->saddr = *(struct in6_addr *)n->primary_key;
1595
1596 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1597 skb_reset_transport_header(reply);
4b29dba9 1598
4b29dba9 1599 /* Neighbor Advertisement */
b080db58 1600 na = skb_put_zero(reply, sizeof(*na) + na_olen);
4b29dba9
DS
1601 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1602 na->icmph.icmp6_router = isrouter;
1603 na->icmph.icmp6_override = 1;
1604 na->icmph.icmp6_solicited = 1;
1605 na->target = ns->target;
1606 ether_addr_copy(&na->opt[2], n->ha);
1607 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1608 na->opt[1] = na_olen >> 3;
1609
1610 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1611 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1612 csum_partial(na, sizeof(*na)+na_olen, 0));
1613
1614 pip6->payload_len = htons(sizeof(*na)+na_olen);
1615
1616 skb_push(reply, sizeof(struct ipv6hdr));
1617
1618 reply->ip_summed = CHECKSUM_UNNECESSARY;
1619
1620 return reply;
1621}
1622
3ad7a4b1 1623static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
f564f45c
CW
1624{
1625 struct vxlan_dev *vxlan = netdev_priv(dev);
8dcd81a9 1626 const struct in6_addr *daddr;
8bff3685 1627 const struct ipv6hdr *iphdr;
4b29dba9 1628 struct inet6_dev *in6_dev;
8bff3685
XL
1629 struct neighbour *n;
1630 struct nd_msg *msg;
f564f45c
CW
1631
1632 in6_dev = __in6_dev_get(dev);
1633 if (!in6_dev)
1634 goto out;
1635
f564f45c 1636 iphdr = ipv6_hdr(skb);
f564f45c 1637 daddr = &iphdr->daddr;
f1fb08f6 1638 msg = (struct nd_msg *)(iphdr + 1);
f564f45c 1639
4b29dba9
DS
1640 if (ipv6_addr_loopback(daddr) ||
1641 ipv6_addr_is_multicast(&msg->target))
1642 goto out;
1643
1644 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW
1645
1646 if (n) {
1647 struct vxlan_fdb *f;
4b29dba9 1648 struct sk_buff *reply;
f564f45c
CW
1649
1650 if (!(n->nud_state & NUD_CONNECTED)) {
1651 neigh_release(n);
1652 goto out;
1653 }
1654
3ad7a4b1 1655 f = vxlan_find_mac(vxlan, n->ha, vni);
f564f45c
CW
1656 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1657 /* bridge-local neighbor */
1658 neigh_release(n);
1659 goto out;
1660 }
1661
4b29dba9
DS
1662 reply = vxlan_na_create(skb, n,
1663 !!(f ? f->flags & NTF_ROUTER : 0));
1664
f564f45c 1665 neigh_release(n);
4b29dba9
DS
1666
1667 if (reply == NULL)
1668 goto out;
1669
1670 if (netif_rx_ni(reply) == NET_RX_DROP)
1671 dev->stats.rx_dropped++;
1672
dc5321d7 1673 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
4b29dba9
DS
1674 union vxlan_addr ipa = {
1675 .sin6.sin6_addr = msg->target,
a45e92a5 1676 .sin6.sin6_family = AF_INET6,
4b29dba9
DS
1677 };
1678
f564f45c
CW
1679 vxlan_ip_miss(dev, &ipa);
1680 }
1681
1682out:
1683 consume_skb(skb);
1684 return NETDEV_TX_OK;
1685}
1686#endif
1687
e4f67add
DS
1688static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1689{
1690 struct vxlan_dev *vxlan = netdev_priv(dev);
1691 struct neighbour *n;
e4f67add
DS
1692
1693 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1694 return false;
1695
1696 n = NULL;
1697 switch (ntohs(eth_hdr(skb)->h_proto)) {
1698 case ETH_P_IP:
e15a00aa
CW
1699 {
1700 struct iphdr *pip;
1701
e4f67add
DS
1702 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1703 return false;
1704 pip = ip_hdr(skb);
1705 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
dc5321d7 1706 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e4c7ed41
CW
1707 union vxlan_addr ipa = {
1708 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1709 .sin.sin_family = AF_INET,
e4c7ed41
CW
1710 };
1711
1712 vxlan_ip_miss(dev, &ipa);
1713 return false;
1714 }
1715
e4f67add 1716 break;
e15a00aa
CW
1717 }
1718#if IS_ENABLED(CONFIG_IPV6)
1719 case ETH_P_IPV6:
1720 {
1721 struct ipv6hdr *pip6;
1722
1723 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1724 return false;
1725 pip6 = ipv6_hdr(skb);
1726 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
dc5321d7 1727 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e15a00aa
CW
1728 union vxlan_addr ipa = {
1729 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1730 .sin6.sin6_family = AF_INET6,
e15a00aa
CW
1731 };
1732
1733 vxlan_ip_miss(dev, &ipa);
1734 return false;
1735 }
1736
1737 break;
1738 }
1739#endif
e4f67add
DS
1740 default:
1741 return false;
1742 }
1743
1744 if (n) {
1745 bool diff;
1746
7367d0b5 1747 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS
1748 if (diff) {
1749 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1750 dev->addr_len);
1751 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1752 }
1753 neigh_release(n);
1754 return diff;
e4c7ed41
CW
1755 }
1756
e4f67add
DS
1757 return false;
1758}
1759
af33c1ad 1760static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG
1761 struct vxlan_metadata *md)
1762{
1763 struct vxlanhdr_gbp *gbp;
1764
db79a621
TG
1765 if (!md->gbp)
1766 return;
1767
3511494c 1768 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1769 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG
1770
1771 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1772 gbp->dont_learn = 1;
1773
1774 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1775 gbp->policy_applied = 1;
1776
1777 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1778}
1779
e1e5314d
JB
1780static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1781 __be16 protocol)
1782{
1783 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1784
1785 gpe->np_applied = 1;
fa20e0e3
JB
1786 gpe->next_protocol = tun_p_from_eth_p(protocol);
1787 if (!gpe->next_protocol)
1788 return -EPFNOSUPPORT;
1789 return 0;
e1e5314d
JB
1790}
1791
f491e56d
JB
1792static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1793 int iphdr_len, __be32 vni,
1794 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1795 bool udp_sum)
e4c7ed41 1796{
e4c7ed41 1797 struct vxlanhdr *vxh;
e4c7ed41
CW
1798 int min_headroom;
1799 int err;
dfd8645e 1800 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 1801 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 1802
af33c1ad 1803 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH
1804 skb->ip_summed == CHECKSUM_PARTIAL) {
1805 int csum_start = skb_checksum_start_offset(skb);
1806
1807 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1808 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1809 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1810 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1811 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1812 }
e4c7ed41 1813
e4c7ed41 1814 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
4a4f86cc 1815 + VXLAN_HLEN + iphdr_len;
649c5b8b
PS
1816
1817 /* Need space for new headers (invalidates iph ptr) */
1818 err = skb_cow_head(skb, min_headroom);
e1e5314d 1819 if (unlikely(err))
c46b7897 1820 return err;
649c5b8b 1821
aed069df
AD
1822 err = iptunnel_handle_offloads(skb, type);
1823 if (err)
c46b7897 1824 return err;
b736a623 1825
d58ff351 1826 vxh = __skb_push(skb, sizeof(*vxh));
54bfd872
JB
1827 vxh->vx_flags = VXLAN_HF_VNI;
1828 vxh->vx_vni = vxlan_vni_field(vni);
49560532 1829
dfd8645e 1830 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 1831 unsigned int start;
dfd8645e 1832
54bfd872
JB
1833 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1834 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1835 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH
1836
1837 if (!skb_is_gso(skb)) {
1838 skb->ip_summed = CHECKSUM_NONE;
1839 skb->encapsulation = 0;
1840 }
1841 }
1842
af33c1ad
TH
1843 if (vxflags & VXLAN_F_GBP)
1844 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB
1845 if (vxflags & VXLAN_F_GPE) {
1846 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1847 if (err < 0)
c46b7897 1848 return err;
e1e5314d
JB
1849 inner_protocol = skb->protocol;
1850 }
3511494c 1851
e1e5314d 1852 skb_set_inner_protocol(skb, inner_protocol);
039f5062 1853 return 0;
49560532 1854}
49560532 1855
655c3de1 1856static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
1857 struct vxlan_sock *sock4,
1a8496ba 1858 struct sk_buff *skb, int oif, u8 tos,
4ecb1d83 1859 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
0c1d70af 1860 struct dst_cache *dst_cache,
db3c6139 1861 const struct ip_tunnel_info *info)
1a8496ba 1862{
db3c6139 1863 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB
1864 struct rtable *rt = NULL;
1865 struct flowi4 fl4;
1866
655c3de1 1867 if (!sock4)
1868 return ERR_PTR(-EIO);
1869
db3c6139
DB
1870 if (tos && !info)
1871 use_cache = false;
1872 if (use_cache) {
0c1d70af
PA
1873 rt = dst_cache_get_ip4(dst_cache, saddr);
1874 if (rt)
1875 return rt;
1876 }
1877
1a8496ba
JB
1878 memset(&fl4, 0, sizeof(fl4));
1879 fl4.flowi4_oif = oif;
1880 fl4.flowi4_tos = RT_TOS(tos);
1881 fl4.flowi4_mark = skb->mark;
1882 fl4.flowi4_proto = IPPROTO_UDP;
1883 fl4.daddr = daddr;
272d96a5 1884 fl4.saddr = *saddr;
4ecb1d83
MP
1885 fl4.fl4_dport = dport;
1886 fl4.fl4_sport = sport;
1a8496ba
JB
1887
1888 rt = ip_route_output_key(vxlan->net, &fl4);
655c3de1 1889 if (likely(!IS_ERR(rt))) {
1890 if (rt->dst.dev == dev) {
1891 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
1892 ip_rt_put(rt);
1893 return ERR_PTR(-ELOOP);
1894 }
1895
1a8496ba 1896 *saddr = fl4.saddr;
0c1d70af
PA
1897 if (use_cache)
1898 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
655c3de1 1899 } else {
1900 netdev_dbg(dev, "no route to %pI4\n", &daddr);
1901 return ERR_PTR(-ENETUNREACH);
0c1d70af 1902 }
1a8496ba
JB
1903 return rt;
1904}
1905
e5d4b29f
JB
1906#if IS_ENABLED(CONFIG_IPV6)
1907static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
655c3de1 1908 struct net_device *dev,
03dc52a8 1909 struct vxlan_sock *sock6,
1400615d 1910 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 1911 __be32 label,
e5d4b29f 1912 const struct in6_addr *daddr,
0c1d70af 1913 struct in6_addr *saddr,
4ecb1d83 1914 __be16 dport, __be16 sport,
db3c6139
DB
1915 struct dst_cache *dst_cache,
1916 const struct ip_tunnel_info *info)
e5d4b29f 1917{
db3c6139 1918 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB
1919 struct dst_entry *ndst;
1920 struct flowi6 fl6;
1921 int err;
1922
c6fcc4fc 1923 if (!sock6)
1924 return ERR_PTR(-EIO);
1925
1400615d
DB
1926 if (tos && !info)
1927 use_cache = false;
db3c6139 1928 if (use_cache) {
0c1d70af
PA
1929 ndst = dst_cache_get_ip6(dst_cache, saddr);
1930 if (ndst)
1931 return ndst;
1932 }
1933
e5d4b29f
JB
1934 memset(&fl6, 0, sizeof(fl6));
1935 fl6.flowi6_oif = oif;
1936 fl6.daddr = *daddr;
272d96a5 1937 fl6.saddr = *saddr;
eaa93bf4 1938 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB
1939 fl6.flowi6_mark = skb->mark;
1940 fl6.flowi6_proto = IPPROTO_UDP;
4ecb1d83
MP
1941 fl6.fl6_dport = dport;
1942 fl6.fl6_sport = sport;
e5d4b29f
JB
1943
1944 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
c6fcc4fc 1945 sock6->sock->sk,
e5d4b29f 1946 &ndst, &fl6);
655c3de1 1947 if (unlikely(err < 0)) {
1948 netdev_dbg(dev, "no route to %pI6\n", daddr);
1949 return ERR_PTR(-ENETUNREACH);
1950 }
1951
1952 if (unlikely(ndst->dev == dev)) {
1953 netdev_dbg(dev, "circular route to %pI6\n", daddr);
1954 dst_release(ndst);
1955 return ERR_PTR(-ELOOP);
1956 }
e5d4b29f
JB
1957
1958 *saddr = fl6.saddr;
db3c6139 1959 if (use_cache)
0c1d70af 1960 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB
1961 return ndst;
1962}
1963#endif
1964
9dcc71e1
SS
1965/* Bypass encapsulation if the destination is local */
1966static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
3ad7a4b1 1967 struct vxlan_dev *dst_vxlan, __be32 vni)
9dcc71e1 1968{
8f84985f 1969 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW
1970 union vxlan_addr loopback;
1971 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR
1972 struct net_device *dev = skb->dev;
1973 int len = skb->len;
9dcc71e1 1974
8f84985f
LR
1975 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1976 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS
1977 skb->pkt_type = PACKET_HOST;
1978 skb->encapsulation = 0;
1979 skb->dev = dst_vxlan->dev;
1980 __skb_pull(skb, skb_network_offset(skb));
1981
e4c7ed41
CW
1982 if (remote_ip->sa.sa_family == AF_INET) {
1983 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1984 loopback.sa.sa_family = AF_INET;
1985#if IS_ENABLED(CONFIG_IPV6)
1986 } else {
1987 loopback.sin6.sin6_addr = in6addr_loopback;
1988 loopback.sa.sa_family = AF_INET6;
1989#endif
1990 }
1991
dc5321d7 1992 if (dst_vxlan->cfg.flags & VXLAN_F_LEARN)
87613de9
MS
1993 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source, 0,
1994 vni);
9dcc71e1
SS
1995
1996 u64_stats_update_begin(&tx_stats->syncp);
1997 tx_stats->tx_packets++;
ce6502a8 1998 tx_stats->tx_bytes += len;
9dcc71e1
SS
1999 u64_stats_update_end(&tx_stats->syncp);
2000
2001 if (netif_rx(skb) == NET_RX_SUCCESS) {
2002 u64_stats_update_begin(&rx_stats->syncp);
2003 rx_stats->rx_packets++;
ce6502a8 2004 rx_stats->rx_bytes += len;
9dcc71e1
SS
2005 u64_stats_update_end(&rx_stats->syncp);
2006 } else {
ce6502a8 2007 dev->stats.rx_dropped++;
9dcc71e1
SS
2008 }
2009}
2010
fee1fad7 2011static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
49f810f0
MS
2012 struct vxlan_dev *vxlan,
2013 union vxlan_addr *daddr,
2014 __be16 dst_port, int dst_ifindex, __be32 vni,
2015 struct dst_entry *dst,
fee1fad7 2016 u32 rt_flags)
2017{
2018#if IS_ENABLED(CONFIG_IPV6)
2019 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2020 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2021 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2022 */
2023 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2024#endif
2025 /* Bypass encapsulation if the destination is local */
2026 if (rt_flags & RTCF_LOCAL &&
2027 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2028 struct vxlan_dev *dst_vxlan;
2029
2030 dst_release(dst);
49f810f0 2031 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
fee1fad7 2032 daddr->sa.sa_family, dst_port,
dc5321d7 2033 vxlan->cfg.flags);
fee1fad7 2034 if (!dst_vxlan) {
2035 dev->stats.tx_errors++;
2036 kfree_skb(skb);
2037
2038 return -ENOENT;
2039 }
3ad7a4b1 2040 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
fee1fad7 2041 return 1;
2042 }
2043
2044 return 0;
2045}
2046
4ad16930 2047static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
3ad7a4b1
RP
2048 __be32 default_vni, struct vxlan_rdst *rdst,
2049 bool did_rsc)
d342894c 2050{
d71785ff 2051 struct dst_cache *dst_cache;
3093fbe7 2052 struct ip_tunnel_info *info;
d342894c 2053 struct vxlan_dev *vxlan = netdev_priv(dev);
0770b53b 2054 const struct iphdr *old_iph = ip_hdr(skb);
e4c7ed41 2055 union vxlan_addr *dst;
272d96a5 2056 union vxlan_addr remote_ip, local_ip;
ee122c79
TG
2057 struct vxlan_metadata _md;
2058 struct vxlan_metadata *md = &_md;
e4c7ed41 2059 __be16 src_port = 0, dst_port;
655c3de1 2060 struct dst_entry *ndst = NULL;
e7f70af1 2061 __be32 vni, label;
d342894c 2062 __u8 tos, ttl;
49f810f0 2063 int ifindex;
0e6fbc5b 2064 int err;
dc5321d7 2065 u32 flags = vxlan->cfg.flags;
b4ed5cad 2066 bool udp_sum = false;
f491e56d 2067 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 2068
61adedf3 2069 info = skb_tunnel_info(skb);
3093fbe7 2070
ee122c79 2071 if (rdst) {
0770b53b 2072 dst = &rdst->remote_ip;
2073 if (vxlan_addr_any(dst)) {
2074 if (did_rsc) {
2075 /* short-circuited back to local bridge */
3ad7a4b1 2076 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
0770b53b 2077 return;
2078 }
2079 goto drop;
2080 }
2081
0dfbdf41 2082 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
3ad7a4b1 2083 vni = (rdst->remote_vni) ? : default_vni;
49f810f0 2084 ifindex = rdst->remote_ifindex;
1158632b 2085 local_ip = vxlan->cfg.saddr;
d71785ff 2086 dst_cache = &rdst->dst_cache;
0770b53b 2087 md->gbp = skb->mark;
2088 ttl = vxlan->cfg.ttl;
2089 if (!ttl && vxlan_addr_multicast(dst))
2090 ttl = 1;
2091
2092 tos = vxlan->cfg.tos;
2093 if (tos == 1)
2094 tos = ip_tunnel_get_dsfield(old_iph, skb);
2095
2096 if (dst->sa.sa_family == AF_INET)
2097 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2098 else
2099 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2100 label = vxlan->cfg.label;
ee122c79
TG
2101 } else {
2102 if (!info) {
2103 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2104 dev->name);
2105 goto drop;
2106 }
b1be00a6 2107 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
272d96a5 2108 if (remote_ip.sa.sa_family == AF_INET) {
a725e514 2109 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
272d96a5 2110 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2111 } else {
a725e514 2112 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
272d96a5 2113 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2114 }
ee122c79 2115 dst = &remote_ip;
0770b53b 2116 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2117 vni = tunnel_id_to_key32(info->key.tun_id);
49f810f0 2118 ifindex = 0;
d71785ff 2119 dst_cache = &info->dst_cache;
0770b53b 2120 if (info->options_len)
2121 md = ip_tunnel_info_opts(info);
a725e514
JB
2122 ttl = info->key.ttl;
2123 tos = info->key.tos;
e7f70af1 2124 label = info->key.label;
b4ed5cad 2125 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514 2126 }
0770b53b 2127 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2128 vxlan->cfg.port_max, true);
a725e514 2129
56de859e 2130 rcu_read_lock();
e4c7ed41 2131 if (dst->sa.sa_family == AF_INET) {
c6fcc4fc 2132 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
c46b7897 2133 struct rtable *rt;
0770b53b 2134 __be16 df = 0;
c6fcc4fc 2135
49f810f0 2136 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
272d96a5 2137 dst->sin.sin_addr.s_addr,
1158632b 2138 &local_ip.sin.sin_addr.s_addr,
4ecb1d83 2139 dst_port, src_port,
d71785ff 2140 dst_cache, info);
8ebd115b
DM
2141 if (IS_ERR(rt)) {
2142 err = PTR_ERR(rt);
c46b7897 2143 goto tx_error;
8ebd115b 2144 }
e4c7ed41
CW
2145
2146 /* Bypass encapsulation if the destination is local */
fee1fad7 2147 if (!info) {
2148 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS
2149 dst_port, ifindex, vni,
2150 &rt->dst, rt->rt_flags);
fee1fad7 2151 if (err)
56de859e 2152 goto out_unlock;
fee1fad7 2153 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
6ceb31ca 2154 df = htons(IP_DF);
fee1fad7 2155 }
6ceb31ca 2156
c46b7897 2157 ndst = &rt->dst;
e4c7ed41
CW
2158 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2159 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
c46b7897 2160 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
54bfd872 2161 vni, md, flags, udp_sum);
f491e56d 2162 if (err < 0)
c46b7897 2163 goto tx_error;
f491e56d 2164
1158632b 2165 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
f491e56d
JB
2166 dst->sin.sin_addr.s_addr, tos, ttl, df,
2167 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW
2168#if IS_ENABLED(CONFIG_IPV6)
2169 } else {
c6fcc4fc 2170 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
e4c7ed41 2171
49f810f0 2172 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
272d96a5 2173 label, &dst->sin6.sin6_addr,
1158632b 2174 &local_ip.sin6.sin6_addr,
4ecb1d83 2175 dst_port, src_port,
db3c6139 2176 dst_cache, info);
e5d4b29f 2177 if (IS_ERR(ndst)) {
8ebd115b 2178 err = PTR_ERR(ndst);
c46b7897 2179 ndst = NULL;
9dcc71e1 2180 goto tx_error;
e4c7ed41 2181 }
655c3de1 2182
fee1fad7 2183 if (!info) {
2184 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
49560532 2185
fee1fad7 2186 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS
2187 dst_port, ifindex, vni,
2188 ndst, rt6i_flags);
fee1fad7 2189 if (err)
56de859e 2190 goto out_unlock;
fee1fad7 2191 }
35e2d115 2192
1400615d 2193 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2194 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB
2195 skb_scrub_packet(skb, xnet);
2196 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2197 vni, md, flags, udp_sum);
c46b7897 2198 if (err < 0)
2199 goto tx_error;
2200
0770b53b 2201 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
1158632b 2202 &local_ip.sin6.sin6_addr,
272d96a5 2203 &dst->sin6.sin6_addr, tos, ttl,
e7f70af1 2204 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW
2205#endif
2206 }
56de859e
JK
2207out_unlock:
2208 rcu_read_unlock();
4ad16930 2209 return;
d342894c 2210
2211drop:
2212 dev->stats.tx_dropped++;
c46b7897 2213 dev_kfree_skb(skb);
2214 return;
d342894c 2215
2216tx_error:
56de859e 2217 rcu_read_unlock();
655c3de1 2218 if (err == -ELOOP)
2219 dev->stats.collisions++;
2220 else if (err == -ENETUNREACH)
2221 dev->stats.tx_carrier_errors++;
c46b7897 2222 dst_release(ndst);
d342894c 2223 dev->stats.tx_errors++;
c46b7897 2224 kfree_skb(skb);
d342894c 2225}
2226
6681712d
DS
2227/* Transmit local packets over Vxlan
2228 *
2229 * Outer IP header inherits ECN and DF from inner header.
2230 * Outer UDP destination is the VXLAN assigned port.
2231 * source port is based on hash of flow
2232 */
2233static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2234{
2235 struct vxlan_dev *vxlan = netdev_priv(dev);
8bff3685 2236 struct vxlan_rdst *rdst, *fdst = NULL;
3093fbe7 2237 const struct ip_tunnel_info *info;
6681712d 2238 bool did_rsc = false;
6681712d 2239 struct vxlan_fdb *f;
8bff3685 2240 struct ethhdr *eth;
3ad7a4b1 2241 __be32 vni = 0;
6681712d 2242
61adedf3 2243 info = skb_tunnel_info(skb);
3093fbe7 2244
6681712d 2245 skb_reset_mac_header(skb);
6681712d 2246
dc5321d7 2247 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP
2248 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2249 info->mode & IP_TUNNEL_INFO_TX) {
2250 vni = tunnel_id_to_key32(info->key.tun_id);
2251 } else {
2252 if (info && info->mode & IP_TUNNEL_INFO_TX)
2253 vxlan_xmit_one(skb, dev, vni, NULL, false);
2254 else
2255 kfree_skb(skb);
2256 return NETDEV_TX_OK;
2257 }
47e5d1b0
JB
2258 }
2259
dc5321d7 2260 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
47e5d1b0 2261 eth = eth_hdr(skb);
f564f45c 2262 if (ntohs(eth->h_proto) == ETH_P_ARP)
3ad7a4b1 2263 return arp_reduce(dev, skb, vni);
f564f45c 2264#if IS_ENABLED(CONFIG_IPV6)
8bff3685
XL
2265 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2266 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2267 sizeof(struct nd_msg)) &&
2268 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2269 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2270
2271 if (m->icmph.icmp6_code == 0 &&
2272 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
f1fb08f6 2273 return neigh_reduce(dev, skb, vni);
f564f45c
CW
2274 }
2275#endif
2276 }
6681712d 2277
47e5d1b0 2278 eth = eth_hdr(skb);
3ad7a4b1 2279 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS
2280 did_rsc = false;
2281
dc5321d7 2282 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
e15a00aa
CW
2283 (ntohs(eth->h_proto) == ETH_P_IP ||
2284 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS
2285 did_rsc = route_shortcircuit(dev, skb);
2286 if (did_rsc)
3ad7a4b1 2287 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS
2288 }
2289
6681712d 2290 if (f == NULL) {
3ad7a4b1 2291 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 2292 if (f == NULL) {
dc5321d7 2293 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
afbd8bae
MR
2294 !is_multicast_ether_addr(eth->h_dest))
2295 vxlan_fdb_miss(vxlan, eth->h_dest);
2296
2297 dev->stats.tx_dropped++;
8f646c92 2298 kfree_skb(skb);
afbd8bae
MR
2299 return NETDEV_TX_OK;
2300 }
2301 }
6681712d 2302
afbd8bae
MR
2303 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2304 struct sk_buff *skb1;
6681712d 2305
8f646c92
ED
2306 if (!fdst) {
2307 fdst = rdst;
2308 continue;
2309 }
afbd8bae
MR
2310 skb1 = skb_clone(skb, GFP_ATOMIC);
2311 if (skb1)
3ad7a4b1 2312 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
6681712d
DS
2313 }
2314
8f646c92 2315 if (fdst)
3ad7a4b1 2316 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
8f646c92
ED
2317 else
2318 kfree_skb(skb);
4ad16930 2319 return NETDEV_TX_OK;
6681712d
DS
2320}
2321
d342894c 2322/* Walk the forwarding table and purge stale entries */
df7e828c 2323static void vxlan_cleanup(struct timer_list *t)
d342894c 2324{
df7e828c 2325 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
d342894c 2326 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2327 unsigned int h;
2328
2329 if (!netif_running(vxlan->dev))
2330 return;
2331
d342894c 2332 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2333 struct hlist_node *p, *n;
14e1d0fa
SD
2334
2335 spin_lock_bh(&vxlan->hash_lock);
d342894c 2336 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2337 struct vxlan_fdb *f
2338 = container_of(p, struct vxlan_fdb, hlist);
2339 unsigned long timeout;
2340
efb5f68f 2341 if (f->state & (NUD_PERMANENT | NUD_NOARP))
d342894c 2342 continue;
2343
def499c9
RP
2344 if (f->flags & NTF_EXT_LEARNED)
2345 continue;
2346
0dfbdf41 2347 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2348 if (time_before_eq(timeout, jiffies)) {
2349 netdev_dbg(vxlan->dev,
2350 "garbage collect %pM\n",
2351 f->eth_addr);
2352 f->state = NUD_STALE;
2353 vxlan_fdb_destroy(vxlan, f);
2354 } else if (time_before(timeout, next_timer))
2355 next_timer = timeout;
2356 }
14e1d0fa 2357 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2358 }
d342894c 2359
2360 mod_timer(&vxlan->age_timer, next_timer);
2361}
2362
a53cb29b
MB
2363static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2364{
2365 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2366
2367 spin_lock(&vn->sock_lock);
69e76661
JB
2368 hlist_del_init_rcu(&vxlan->hlist4.hlist);
2369#if IS_ENABLED(CONFIG_IPV6)
2370 hlist_del_init_rcu(&vxlan->hlist6.hlist);
2371#endif
a53cb29b
MB
2372 spin_unlock(&vn->sock_lock);
2373}
2374
69e76661
JB
2375static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2376 struct vxlan_dev_node *node)
9c2e24e1 2377{
56ef9c90 2378 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2379 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2380
69e76661 2381 node->vxlan = vxlan;
56ef9c90 2382 spin_lock(&vn->sock_lock);
69e76661 2383 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
56ef9c90 2384 spin_unlock(&vn->sock_lock);
9c2e24e1
PS
2385}
2386
d342894c 2387/* Setup stats when device is created */
2388static int vxlan_init(struct net_device *dev)
2389{
1c213bd2 2390 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2391 if (!dev->tstats)
d342894c 2392 return -ENOMEM;
2393
2394 return 0;
2395}
2396
3ad7a4b1 2397static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
afbd8bae
MR
2398{
2399 struct vxlan_fdb *f;
2400
2401 spin_lock_bh(&vxlan->hash_lock);
3ad7a4b1 2402 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae
MR
2403 if (f)
2404 vxlan_fdb_destroy(vxlan, f);
2405 spin_unlock_bh(&vxlan->hash_lock);
2406}
2407
ebf4063e
SH
2408static void vxlan_uninit(struct net_device *dev)
2409{
2410 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2411
3ad7a4b1 2412 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
afbd8bae 2413
ebf4063e
SH
2414 free_percpu(dev->tstats);
2415}
2416
d342894c 2417/* Start ageing timer and join group when device is brought up */
2418static int vxlan_open(struct net_device *dev)
2419{
2420 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2421 int ret;
56ef9c90 2422
205f356d
JB
2423 ret = vxlan_sock_add(vxlan);
2424 if (ret < 0)
2425 return ret;
d342894c 2426
79d4a94f 2427 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2428 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL
2429 if (ret == -EADDRINUSE)
2430 ret = 0;
56ef9c90 2431 if (ret) {
205f356d 2432 vxlan_sock_release(vxlan);
56ef9c90
MRL
2433 return ret;
2434 }
d342894c 2435 }
2436
0dfbdf41 2437 if (vxlan->cfg.age_interval)
d342894c 2438 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2439
56ef9c90 2440 return ret;
d342894c 2441}
2442
2443/* Purge the forwarding table */
8b3f9337 2444static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
d342894c 2445{
31fec5aa 2446 unsigned int h;
d342894c 2447
2448 spin_lock_bh(&vxlan->hash_lock);
2449 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2450 struct hlist_node *p, *n;
2451 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2452 struct vxlan_fdb *f
2453 = container_of(p, struct vxlan_fdb, hlist);
8b3f9337
RP
2454 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2455 continue;
afbd8bae
MR
2456 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2457 if (!is_zero_ether_addr(f->eth_addr))
2458 vxlan_fdb_destroy(vxlan, f);
d342894c 2459 }
2460 }
2461 spin_unlock_bh(&vxlan->hash_lock);
2462}
2463
2464/* Cleanup timer and forwarding table on shutdown */
2465static int vxlan_stop(struct net_device *dev)
2466{
2467 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2468 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2469 int ret = 0;
d342894c 2470
24c0e683 2471 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2472 !vxlan_group_used(vn, vxlan))
56ef9c90 2473 ret = vxlan_igmp_leave(vxlan);
d342894c 2474
2475 del_timer_sync(&vxlan->age_timer);
2476
8b3f9337 2477 vxlan_flush(vxlan, false);
205f356d 2478 vxlan_sock_release(vxlan);
d342894c 2479
56ef9c90 2480 return ret;
d342894c 2481}
2482
d342894c 2483/* Stub, nothing needs to be done. */
2484static void vxlan_set_multicast_list(struct net_device *dev)
2485{
2486}
2487
91572088 2488static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
345010b5 2489{
91572088
JW
2490 struct vxlan_dev *vxlan = netdev_priv(dev);
2491 struct vxlan_rdst *dst = &vxlan->default_dst;
2492 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2493 dst->remote_ifindex);
ce44a4ae 2494 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
345010b5 2495
91572088
JW
2496 /* This check is different than dev->max_mtu, because it looks at
2497 * the lowerdev->mtu, rather than the static dev->max_mtu
2498 */
2499 if (lowerdev) {
2500 int max_mtu = lowerdev->mtu -
2501 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2502 if (new_mtu > max_mtu)
72564b59 2503 return -EINVAL;
72564b59
DW
2504 }
2505
345010b5
DB
2506 dev->mtu = new_mtu;
2507 return 0;
2508}
2509
fc4099f1
PS
2510static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2511{
2512 struct vxlan_dev *vxlan = netdev_priv(dev);
2513 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2514 __be16 sport, dport;
2515
2516 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2517 vxlan->cfg.port_max, true);
2518 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2519
239e944f 2520 if (ip_tunnel_info_af(info) == AF_INET) {
c6fcc4fc 2521 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
1a8496ba
JB
2522 struct rtable *rt;
2523
655c3de1 2524 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
1a8496ba 2525 info->key.u.ipv4.dst,
22f0708a
PA
2526 &info->key.u.ipv4.src, dport, sport,
2527 &info->dst_cache, info);
1a8496ba
JB
2528 if (IS_ERR(rt))
2529 return PTR_ERR(rt);
2530 ip_rt_put(rt);
239e944f
JB
2531 } else {
2532#if IS_ENABLED(CONFIG_IPV6)
03dc52a8 2533 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
239e944f
JB
2534 struct dst_entry *ndst;
2535
655c3de1 2536 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
e7f70af1 2537 info->key.label, &info->key.u.ipv6.dst,
22f0708a
PA
2538 &info->key.u.ipv6.src, dport, sport,
2539 &info->dst_cache, info);
239e944f
JB
2540 if (IS_ERR(ndst))
2541 return PTR_ERR(ndst);
2542 dst_release(ndst);
239e944f
JB
2543#else /* !CONFIG_IPV6 */
2544 return -EPFNOSUPPORT;
2545#endif
2546 }
1a8496ba
JB
2547 info->key.tp_src = sport;
2548 info->key.tp_dst = dport;
239e944f 2549 return 0;
fc4099f1
PS
2550}
2551
0c867c9b 2552static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 2553 .ndo_init = vxlan_init,
ebf4063e 2554 .ndo_uninit = vxlan_uninit,
d342894c 2555 .ndo_open = vxlan_open,
2556 .ndo_stop = vxlan_stop,
2557 .ndo_start_xmit = vxlan_xmit,
e8171045 2558 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2559 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2560 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2561 .ndo_validate_addr = eth_validate_addr,
2562 .ndo_set_mac_address = eth_mac_addr,
2563 .ndo_fdb_add = vxlan_fdb_add,
2564 .ndo_fdb_del = vxlan_fdb_delete,
2565 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2566 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2567};
2568
e1e5314d
JB
2569static const struct net_device_ops vxlan_netdev_raw_ops = {
2570 .ndo_init = vxlan_init,
2571 .ndo_uninit = vxlan_uninit,
2572 .ndo_open = vxlan_open,
2573 .ndo_stop = vxlan_stop,
2574 .ndo_start_xmit = vxlan_xmit,
2575 .ndo_get_stats64 = ip_tunnel_get_stats64,
2576 .ndo_change_mtu = vxlan_change_mtu,
2577 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2578};
2579
d342894c 2580/* Info for udev, that this is a virtual tunnel endpoint */
2581static struct device_type vxlan_type = {
2582 .name = "vxlan",
2583};
2584
e5de25dc 2585/* Calls the ndo_udp_tunnel_add of the caller in order to
35e42379 2586 * supply the listening VXLAN udp ports. Callers are expected
e5de25dc 2587 * to implement the ndo_udp_tunnel_add.
53cf5275 2588 */
2d2b13fc 2589static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
53cf5275
JG
2590{
2591 struct vxlan_sock *vs;
2592 struct net *net = dev_net(dev);
2593 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
35e42379 2594 unsigned int i;
53cf5275
JG
2595
2596 spin_lock(&vn->sock_lock);
2597 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2d2b13fc
SD
2598 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2599 unsigned short type;
2600
2601 if (vs->flags & VXLAN_F_GPE)
2602 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
2603 else
2604 type = UDP_TUNNEL_TYPE_VXLAN;
2605
2606 if (push)
2607 udp_tunnel_push_rx_port(dev, vs->sock, type);
2608 else
2609 udp_tunnel_drop_rx_port(dev, vs->sock, type);
2610 }
53cf5275
JG
2611 }
2612 spin_unlock(&vn->sock_lock);
2613}
53cf5275 2614
d342894c 2615/* Initialize the device structure. */
2616static void vxlan_setup(struct net_device *dev)
2617{
2618 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2619 unsigned int h;
d342894c 2620
65226ef8
JB
2621 eth_hw_addr_random(dev);
2622 ether_setup(dev);
2623
cf124db5 2624 dev->needs_free_netdev = true;
d342894c 2625 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2626
d342894c 2627 dev->features |= NETIF_F_LLTX;
d6727fe3 2628 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2629 dev->features |= NETIF_F_RXCSUM;
05c0db08 2630 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2631
1eaa8178 2632 dev->vlan_features = dev->features;
0afb1666 2633 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2634 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
02875878 2635 netif_keep_dst(dev);
0c867c9b 2636 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 2637
a985343b
MS
2638 /* MTU range: 68 - 65535 */
2639 dev->min_mtu = ETH_MIN_MTU;
2640 dev->max_mtu = ETH_MAX_MTU;
2641
553675fb 2642 INIT_LIST_HEAD(&vxlan->next);
d342894c 2643 spin_lock_init(&vxlan->hash_lock);
2644
df7e828c 2645 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
d342894c 2646
2647 vxlan->dev = dev;
2648
58ce31cc
TH
2649 gro_cells_init(&vxlan->gro_cells, dev);
2650
d342894c 2651 for (h = 0; h < FDB_HASH_SIZE; ++h)
2652 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2653}
2654
0c867c9b
JB
2655static void vxlan_ether_setup(struct net_device *dev)
2656{
0c867c9b
JB
2657 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2658 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2659 dev->netdev_ops = &vxlan_netdev_ether_ops;
2660}
2661
e1e5314d
JB
2662static void vxlan_raw_setup(struct net_device *dev)
2663{
65226ef8 2664 dev->header_ops = NULL;
e1e5314d
JB
2665 dev->type = ARPHRD_NONE;
2666 dev->hard_header_len = 0;
2667 dev->addr_len = 0;
e1e5314d
JB
2668 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2669 dev->netdev_ops = &vxlan_netdev_raw_ops;
2670}
2671
d342894c 2672static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2673 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2674 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2675 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2676 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2677 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2678 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2679 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2680 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 2681 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 2682 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2683 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2684 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2685 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS
2686 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2687 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2688 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2689 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2690 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2691 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH
2692 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2693 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2694 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH
2695 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2696 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2697 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 2698 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 2699 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2700};
2701
a8b8a889
MS
2702static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
2703 struct netlink_ext_ack *extack)
d342894c 2704{
2705 if (tb[IFLA_ADDRESS]) {
2706 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
653ef6a3
GM
2707 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2708 "Provided link layer address is not Ethernet");
d342894c 2709 return -EINVAL;
2710 }
2711
2712 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
653ef6a3
GM
2713 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2714 "Provided Ethernet address is not unicast");
d342894c 2715 return -EADDRNOTAVAIL;
2716 }
2717 }
2718
a985343b 2719 if (tb[IFLA_MTU]) {
019b13ae 2720 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
a985343b 2721
653ef6a3
GM
2722 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
2723 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
2724 "MTU must be between 68 and 65535");
a985343b 2725 return -EINVAL;
653ef6a3 2726 }
a985343b
MS
2727 }
2728
653ef6a3
GM
2729 if (!data) {
2730 NL_SET_ERR_MSG(extack,
2731 "Required attributes not provided to perform the operation");
d342894c 2732 return -EINVAL;
653ef6a3 2733 }
d342894c 2734
2735 if (data[IFLA_VXLAN_ID]) {
a985343b
MS
2736 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2737
653ef6a3
GM
2738 if (id >= VXLAN_N_VID) {
2739 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID],
2740 "VXLAN ID must be lower than 16777216");
d342894c 2741 return -ERANGE;
653ef6a3 2742 }
d342894c 2743 }
2744
05f47d69 2745 if (data[IFLA_VXLAN_PORT_RANGE]) {
2746 const struct ifla_vxlan_port_range *p
2747 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2748
2749 if (ntohs(p->high) < ntohs(p->low)) {
653ef6a3
GM
2750 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
2751 "Invalid source port range");
05f47d69 2752 return -EINVAL;
2753 }
2754 }
2755
d342894c 2756 return 0;
2757}
2758
1b13c97f
YB
2759static void vxlan_get_drvinfo(struct net_device *netdev,
2760 struct ethtool_drvinfo *drvinfo)
2761{
2762 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2763 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2764}
2765
2766static const struct ethtool_ops vxlan_ethtool_ops = {
2767 .get_drvinfo = vxlan_get_drvinfo,
2768 .get_link = ethtool_op_get_link,
2769};
2770
3ee64f39
TH
2771static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2772 __be16 port, u32 flags)
553675fb 2773{
e4c7ed41 2774 struct socket *sock;
3ee64f39
TH
2775 struct udp_port_cfg udp_conf;
2776 int err;
e4c7ed41 2777
3ee64f39 2778 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2779
3ee64f39
TH
2780 if (ipv6) {
2781 udp_conf.family = AF_INET6;
3ee64f39 2782 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2783 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2784 udp_conf.ipv6_v6only = 1;
3ee64f39
TH
2785 } else {
2786 udp_conf.family = AF_INET;
e4c7ed41 2787 }
e4c7ed41 2788
3ee64f39 2789 udp_conf.local_udp_port = port;
553675fb 2790
3ee64f39
TH
2791 /* Open UDP socket */
2792 err = udp_sock_create(net, &udp_conf, &sock);
2793 if (err < 0)
2794 return ERR_PTR(err);
e4c7ed41 2795
39deb2c7 2796 return sock;
e4c7ed41
CW
2797}
2798
2799/* Create new listen socket if needed */
b1be00a6
JB
2800static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2801 __be16 port, u32 flags)
e4c7ed41
CW
2802{
2803 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2804 struct vxlan_sock *vs;
2805 struct socket *sock;
e4c7ed41 2806 unsigned int h;
acbf74a7 2807 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2808
dc01e7d3 2809 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW
2810 if (!vs)
2811 return ERR_PTR(-ENOMEM);
2812
2813 for (h = 0; h < VNI_HASH_SIZE; ++h)
2814 INIT_HLIST_HEAD(&vs->vni_list[h]);
2815
3ee64f39 2816 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2817 if (IS_ERR(sock)) {
553675fb 2818 kfree(vs);
e50fddc8 2819 return ERR_CAST(sock);
553675fb 2820 }
e4c7ed41
CW
2821
2822 vs->sock = sock;
66af846f 2823 refcount_set(&vs->refcnt, 1);
af33c1ad 2824 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2825
9c2e24e1
PS
2826 spin_lock(&vn->sock_lock);
2827 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
e7b3db5e 2828 udp_tunnel_notify_add_rx_port(sock,
b9adcd69
AD
2829 (vs->flags & VXLAN_F_GPE) ?
2830 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 2831 UDP_TUNNEL_TYPE_VXLAN);
9c2e24e1 2832 spin_unlock(&vn->sock_lock);
553675fb 2833
2834 /* Mark socket as an encapsulation socket. */
5602c48c 2835 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ
2836 tunnel_cfg.sk_user_data = vs;
2837 tunnel_cfg.encap_type = 1;
f2d1968e 2838 tunnel_cfg.encap_rcv = vxlan_rcv;
acbf74a7 2839 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH
2840 tunnel_cfg.gro_receive = vxlan_gro_receive;
2841 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ
2842
2843 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2844
9c2e24e1
PS
2845 return vs;
2846}
2847
b1be00a6 2848static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2849{
205f356d
JB
2850 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2851 struct vxlan_sock *vs = NULL;
69e76661 2852 struct vxlan_dev_node *node;
9c2e24e1 2853
205f356d 2854 if (!vxlan->cfg.no_share) {
56ef9c90 2855 spin_lock(&vn->sock_lock);
205f356d 2856 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
dc5321d7 2857 vxlan->cfg.dst_port, vxlan->cfg.flags);
66af846f 2858 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
56ef9c90 2859 spin_unlock(&vn->sock_lock);
205f356d 2860 return -EBUSY;
56ef9c90
MRL
2861 }
2862 spin_unlock(&vn->sock_lock);
2863 }
205f356d 2864 if (!vs)
b1be00a6 2865 vs = vxlan_socket_create(vxlan->net, ipv6,
dc5321d7 2866 vxlan->cfg.dst_port, vxlan->cfg.flags);
205f356d
JB
2867 if (IS_ERR(vs))
2868 return PTR_ERR(vs);
b1be00a6 2869#if IS_ENABLED(CONFIG_IPV6)
69e76661 2870 if (ipv6) {
c6fcc4fc 2871 rcu_assign_pointer(vxlan->vn6_sock, vs);
69e76661
JB
2872 node = &vxlan->hlist6;
2873 } else
b1be00a6 2874#endif
69e76661 2875 {
c6fcc4fc 2876 rcu_assign_pointer(vxlan->vn4_sock, vs);
69e76661
JB
2877 node = &vxlan->hlist4;
2878 }
2879 vxlan_vs_add_dev(vs, vxlan, node);
205f356d 2880 return 0;
553675fb 2881}
2882
b1be00a6
JB
2883static int vxlan_sock_add(struct vxlan_dev *vxlan)
2884{
dc5321d7
MS
2885 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
2886 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
d074bf96 2887 bool ipv4 = !ipv6 || metadata;
b1be00a6
JB
2888 int ret = 0;
2889
c6fcc4fc 2890 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
b1be00a6 2891#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 2892 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
d074bf96 2893 if (ipv6) {
b1be00a6 2894 ret = __vxlan_sock_add(vxlan, true);
d074bf96
JB
2895 if (ret < 0 && ret != -EAFNOSUPPORT)
2896 ipv4 = false;
2897 }
b1be00a6 2898#endif
d074bf96 2899 if (ipv4)
b1be00a6
JB
2900 ret = __vxlan_sock_add(vxlan, false);
2901 if (ret < 0)
2902 vxlan_sock_release(vxlan);
2903 return ret;
2904}
2905
a985343b
MS
2906static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
2907 struct net_device **lower,
653ef6a3
GM
2908 struct vxlan_dev *old,
2909 struct netlink_ext_ack *extack)
d342894c 2910{
33564bbb 2911 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
a985343b 2912 struct vxlan_dev *tmp;
e4c7ed41 2913 bool use_ipv6 = false;
d342894c 2914
a985343b
MS
2915 if (conf->flags & VXLAN_F_GPE) {
2916 /* For now, allow GPE only together with
2917 * COLLECT_METADATA. This can be relaxed later; in such
2918 * case, the other side of the PtP link will have to be
2919 * provided.
2920 */
2921 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
2922 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
653ef6a3
GM
2923 NL_SET_ERR_MSG(extack,
2924 "VXLAN GPE does not support this combination of attributes");
a985343b 2925 return -EINVAL;
3555621d 2926 }
e1e5314d 2927 }
0c867c9b 2928
ce44a4ae
MS
2929 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
2930 /* Unless IPv6 is explicitly requested, assume IPv4 */
a985343b 2931 conf->remote_ip.sa.sa_family = AF_INET;
ce44a4ae
MS
2932 conf->saddr.sa.sa_family = AF_INET;
2933 } else if (!conf->remote_ip.sa.sa_family) {
2934 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
2935 } else if (!conf->saddr.sa.sa_family) {
2936 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
2937 }
2938
653ef6a3
GM
2939 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
2940 NL_SET_ERR_MSG(extack,
2941 "Local and remote address must be from the same family");
ce44a4ae 2942 return -EINVAL;
653ef6a3 2943 }
e4c7ed41 2944
653ef6a3
GM
2945 if (vxlan_addr_multicast(&conf->saddr)) {
2946 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
0f22a3c6 2947 return -EINVAL;
653ef6a3 2948 }
0f22a3c6 2949
ce44a4ae 2950 if (conf->saddr.sa.sa_family == AF_INET6) {
653ef6a3
GM
2951 if (!IS_ENABLED(CONFIG_IPV6)) {
2952 NL_SET_ERR_MSG(extack,
2953 "IPv6 support not enabled in the kernel");
057ba29b 2954 return -EPFNOSUPPORT;
653ef6a3 2955 }
e4c7ed41 2956 use_ipv6 = true;
a985343b 2957 conf->flags |= VXLAN_F_IPV6;
0f22a3c6
MS
2958
2959 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
2960 int local_type =
2961 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
2962 int remote_type =
2963 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
2964
2965 if (local_type & IPV6_ADDR_LINKLOCAL) {
2966 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
653ef6a3
GM
2967 (remote_type != IPV6_ADDR_ANY)) {
2968 NL_SET_ERR_MSG(extack,
2969 "Invalid combination of local and remote address scopes");
0f22a3c6 2970 return -EINVAL;
653ef6a3 2971 }
0f22a3c6
MS
2972
2973 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
2974 } else {
2975 if (remote_type ==
653ef6a3
GM
2976 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
2977 NL_SET_ERR_MSG(extack,
2978 "Invalid combination of local and remote address scopes");
0f22a3c6 2979 return -EINVAL;
653ef6a3 2980 }
0f22a3c6
MS
2981
2982 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
2983 }
2984 }
057ba29b 2985 }
d342894c 2986
653ef6a3
GM
2987 if (conf->label && !use_ipv6) {
2988 NL_SET_ERR_MSG(extack,
2989 "Label attribute only applies to IPv6 VXLAN devices");
e7f70af1 2990 return -EINVAL;
653ef6a3 2991 }
e7f70af1 2992
a985343b
MS
2993 if (conf->remote_ifindex) {
2994 struct net_device *lowerdev;
34e02aa1 2995
a985343b 2996 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
653ef6a3
GM
2997 if (!lowerdev) {
2998 NL_SET_ERR_MSG(extack,
2999 "Invalid local interface, device not found");
34e02aa1 3000 return -ENODEV;
653ef6a3 3001 }
d342894c 3002
e4c7ed41
CW
3003#if IS_ENABLED(CONFIG_IPV6)
3004 if (use_ipv6) {
3005 struct inet6_dev *idev = __in6_dev_get(lowerdev);
653ef6a3
GM
3006 if (idev && idev->cnf.disable_ipv6) {
3007 NL_SET_ERR_MSG(extack,
3008 "IPv6 support disabled by administrator");
e4c7ed41 3009 return -EPERM;
653ef6a3 3010 }
e4c7ed41
CW
3011 }
3012#endif
3013
a985343b
MS
3014 *lower = lowerdev;
3015 } else {
653ef6a3
GM
3016 if (vxlan_addr_multicast(&conf->remote_ip)) {
3017 NL_SET_ERR_MSG(extack,
3018 "Local interface required for multicast remote destination");
3019
a985343b 3020 return -EINVAL;
653ef6a3 3021 }
1ba56fb4 3022
0f22a3c6 3023#if IS_ENABLED(CONFIG_IPV6)
653ef6a3
GM
3024 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3025 NL_SET_ERR_MSG(extack,
3026 "Local interface required for link-local local/remote addresses");
0f22a3c6 3027 return -EINVAL;
653ef6a3 3028 }
0f22a3c6
MS
3029#endif
3030
a985343b 3031 *lower = NULL;
9dc2ad10 3032 }
d342894c 3033
a985343b
MS
3034 if (!conf->dst_port) {
3035 if (conf->flags & VXLAN_F_GPE)
3036 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3037 else
3038 conf->dst_port = htons(vxlan_port);
d6acfeb1
FM
3039 }
3040
a985343b
MS
3041 if (!conf->age_interval)
3042 conf->age_interval = FDB_AGE_DEFAULT;
91572088 3043
a985343b
MS
3044 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3045 if (tmp == old)
3046 continue;
91572088 3047
49f810f0
MS
3048 if (tmp->cfg.vni != conf->vni)
3049 continue;
3050 if (tmp->cfg.dst_port != conf->dst_port)
3051 continue;
3052 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
ce44a4ae 3053 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
49f810f0
MS
3054 continue;
3055
3056 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3057 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3058 continue;
3059
653ef6a3
GM
3060 NL_SET_ERR_MSG(extack,
3061 "A VXLAN device with the specified VNI already exists");
49f810f0 3062 return -EEXIST;
a985343b 3063 }
91572088 3064
a985343b
MS
3065 return 0;
3066}
3067
3068static void vxlan_config_apply(struct net_device *dev,
3069 struct vxlan_config *conf,
889ce937
SD
3070 struct net_device *lowerdev,
3071 struct net *src_net,
3072 bool changelink)
a985343b
MS
3073{
3074 struct vxlan_dev *vxlan = netdev_priv(dev);
3075 struct vxlan_rdst *dst = &vxlan->default_dst;
3076 unsigned short needed_headroom = ETH_HLEN;
3077 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3078 int max_mtu = ETH_MAX_MTU;
3079
3080 if (!changelink) {
3081 if (conf->flags & VXLAN_F_GPE)
3082 vxlan_raw_setup(dev);
3083 else
3084 vxlan_ether_setup(dev);
3085
3086 if (conf->mtu)
3087 dev->mtu = conf->mtu;
889ce937
SD
3088
3089 vxlan->net = src_net;
a985343b
MS
3090 }
3091
3092 dst->remote_vni = conf->vni;
3093
3094 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3095
3096 if (lowerdev) {
3097 dst->remote_ifindex = conf->remote_ifindex;
3098
3099 dev->gso_max_size = lowerdev->gso_max_size;
3100 dev->gso_max_segs = lowerdev->gso_max_segs;
91572088 3101
a985343b 3102 needed_headroom = lowerdev->hard_header_len;
91572088 3103
a985343b
MS
3104 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3105 VXLAN_HEADROOM);
7e059158
DW
3106 }
3107
a985343b
MS
3108 if (dev->mtu > max_mtu)
3109 dev->mtu = max_mtu;
3110
b1be00a6
JB
3111 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3112 needed_headroom += VXLAN6_HEADROOM;
3113 else
3114 needed_headroom += VXLAN_HEADROOM;
3115 dev->needed_headroom = needed_headroom;
3116
0dfbdf41 3117 memcpy(&vxlan->cfg, conf, sizeof(*conf));
a985343b 3118}
0dfbdf41 3119
a985343b 3120static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
653ef6a3
GM
3121 struct vxlan_config *conf, bool changelink,
3122 struct netlink_ext_ack *extack)
a985343b
MS
3123{
3124 struct vxlan_dev *vxlan = netdev_priv(dev);
3125 struct net_device *lowerdev;
3126 int ret;
0dfbdf41 3127
653ef6a3 3128 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
a985343b
MS
3129 if (ret)
3130 return ret;
8bcdc4f3 3131
889ce937 3132 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
0dfbdf41 3133
0dfbdf41
TG
3134 return 0;
3135}
3136
c80498e3 3137static int __vxlan_dev_create(struct net *net, struct net_device *dev,
653ef6a3
GM
3138 struct vxlan_config *conf,
3139 struct netlink_ext_ack *extack)
c80498e3
ND
3140{
3141 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3142 struct vxlan_dev *vxlan = netdev_priv(dev);
3143 int err;
3144
653ef6a3 3145 err = vxlan_dev_configure(net, dev, conf, false, extack);
c80498e3
ND
3146 if (err)
3147 return err;
3148
3149 dev->ethtool_ops = &vxlan_ethtool_ops;
3150
3151 /* create an fdb entry for a valid default destination */
3152 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
3153 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3154 &vxlan->default_dst.remote_ip,
3155 NUD_REACHABLE | NUD_PERMANENT,
3156 NLM_F_EXCL | NLM_F_CREATE,
3157 vxlan->cfg.dst_port,
3158 vxlan->default_dst.remote_vni,
3159 vxlan->default_dst.remote_vni,
3160 vxlan->default_dst.remote_ifindex,
3161 NTF_SELF);
3162 if (err)
3163 return err;
3164 }
3165
3166 err = register_netdevice(dev);
3167 if (err) {
3168 vxlan_fdb_delete_default(vxlan, vxlan->default_dst.remote_vni);
3169 return err;
3170 }
3171
3172 list_add(&vxlan->next, &vn->vxlan_list);
3173 return 0;
3174}
3175
8bcdc4f3
RP
3176static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3177 struct net_device *dev, struct vxlan_config *conf,
3178 bool changelink)
0dfbdf41 3179{
8bcdc4f3 3180 struct vxlan_dev *vxlan = netdev_priv(dev);
0dfbdf41 3181
8bcdc4f3 3182 memset(conf, 0, sizeof(*conf));
e277de5f 3183
8bcdc4f3
RP
3184 /* if changelink operation, start with old existing cfg */
3185 if (changelink)
3186 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3187
3188 if (data[IFLA_VXLAN_ID]) {
3189 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3190
3191 if (changelink && (vni != conf->vni))
3192 return -EOPNOTSUPP;
3193 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3194 }
0dfbdf41
TG
3195
3196 if (data[IFLA_VXLAN_GROUP]) {
ce44a4ae
MS
3197 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET))
3198 return -EOPNOTSUPP;
3199
8bcdc4f3 3200 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
ce44a4ae 3201 conf->remote_ip.sa.sa_family = AF_INET;
0dfbdf41
TG
3202 } else if (data[IFLA_VXLAN_GROUP6]) {
3203 if (!IS_ENABLED(CONFIG_IPV6))
3204 return -EPFNOSUPPORT;
3205
ce44a4ae
MS
3206 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6))
3207 return -EOPNOTSUPP;
3208
8bcdc4f3
RP
3209 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3210 conf->remote_ip.sa.sa_family = AF_INET6;
0dfbdf41
TG
3211 }
3212
3213 if (data[IFLA_VXLAN_LOCAL]) {
ce44a4ae
MS
3214 if (changelink && (conf->saddr.sa.sa_family != AF_INET))
3215 return -EOPNOTSUPP;
3216
8bcdc4f3
RP
3217 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3218 conf->saddr.sa.sa_family = AF_INET;
0dfbdf41
TG
3219 } else if (data[IFLA_VXLAN_LOCAL6]) {
3220 if (!IS_ENABLED(CONFIG_IPV6))
3221 return -EPFNOSUPPORT;
3222
ce44a4ae
MS
3223 if (changelink && (conf->saddr.sa.sa_family != AF_INET6))
3224 return -EOPNOTSUPP;
3225
0dfbdf41 3226 /* TODO: respect scope id */
8bcdc4f3
RP
3227 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3228 conf->saddr.sa.sa_family = AF_INET6;
0dfbdf41
TG
3229 }
3230
3231 if (data[IFLA_VXLAN_LINK])
8bcdc4f3 3232 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
0dfbdf41 3233
d342894c 3234 if (data[IFLA_VXLAN_TOS])
8bcdc4f3 3235 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 3236
afb97186 3237 if (data[IFLA_VXLAN_TTL])
8bcdc4f3 3238 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 3239
e7f70af1 3240 if (data[IFLA_VXLAN_LABEL])
8bcdc4f3 3241 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
e7f70af1
DB
3242 IPV6_FLOWLABEL_MASK;
3243
8bcdc4f3 3244 if (data[IFLA_VXLAN_LEARNING]) {
dc5321d7 3245 if (nla_get_u8(data[IFLA_VXLAN_LEARNING]))
8bcdc4f3 3246 conf->flags |= VXLAN_F_LEARN;
dc5321d7 3247 else
8bcdc4f3 3248 conf->flags &= ~VXLAN_F_LEARN;
8bcdc4f3
RP
3249 } else if (!changelink) {
3250 /* default to learn on a new device */
3251 conf->flags |= VXLAN_F_LEARN;
3252 }
d342894c 3253
8bcdc4f3
RP
3254 if (data[IFLA_VXLAN_AGEING]) {
3255 if (changelink)
3256 return -EOPNOTSUPP;
3257 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3258 }
d342894c 3259
8bcdc4f3
RP
3260 if (data[IFLA_VXLAN_PROXY]) {
3261 if (changelink)
3262 return -EOPNOTSUPP;
3263 if (nla_get_u8(data[IFLA_VXLAN_PROXY]))
3264 conf->flags |= VXLAN_F_PROXY;
3265 }
e4f67add 3266
8bcdc4f3
RP
3267 if (data[IFLA_VXLAN_RSC]) {
3268 if (changelink)
3269 return -EOPNOTSUPP;
3270 if (nla_get_u8(data[IFLA_VXLAN_RSC]))
3271 conf->flags |= VXLAN_F_RSC;
3272 }
e4f67add 3273
8bcdc4f3
RP
3274 if (data[IFLA_VXLAN_L2MISS]) {
3275 if (changelink)
3276 return -EOPNOTSUPP;
3277 if (nla_get_u8(data[IFLA_VXLAN_L2MISS]))
3278 conf->flags |= VXLAN_F_L2MISS;
3279 }
e4f67add 3280
8bcdc4f3
RP
3281 if (data[IFLA_VXLAN_L3MISS]) {
3282 if (changelink)
3283 return -EOPNOTSUPP;
3284 if (nla_get_u8(data[IFLA_VXLAN_L3MISS]))
3285 conf->flags |= VXLAN_F_L3MISS;
3286 }
e4f67add 3287
8bcdc4f3
RP
3288 if (data[IFLA_VXLAN_LIMIT]) {
3289 if (changelink)
3290 return -EOPNOTSUPP;
3291 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3292 }
d342894c 3293
8bcdc4f3
RP
3294 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3295 if (changelink)
3296 return -EOPNOTSUPP;
3297 if (nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3298 conf->flags |= VXLAN_F_COLLECT_METADATA;
3299 }
f8a9b1bc 3300
05f47d69 3301 if (data[IFLA_VXLAN_PORT_RANGE]) {
8bcdc4f3
RP
3302 if (!changelink) {
3303 const struct ifla_vxlan_port_range *p
3304 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3305 conf->port_min = ntohs(p->low);
3306 conf->port_max = ntohs(p->high);
3307 } else {
3308 return -EOPNOTSUPP;
3309 }
05f47d69 3310 }
3311
8bcdc4f3
RP
3312 if (data[IFLA_VXLAN_PORT]) {
3313 if (changelink)
3314 return -EOPNOTSUPP;
3315 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3316 }
823aa873 3317
8bcdc4f3
RP
3318 if (data[IFLA_VXLAN_UDP_CSUM]) {
3319 if (changelink)
3320 return -EOPNOTSUPP;
3321 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3322 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3323 }
359a0ea9 3324
8bcdc4f3
RP
3325 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3326 if (changelink)
3327 return -EOPNOTSUPP;
3328 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
3329 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
3330 }
359a0ea9 3331
8bcdc4f3
RP
3332 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3333 if (changelink)
3334 return -EOPNOTSUPP;
3335 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
3336 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
3337 }
359a0ea9 3338
8bcdc4f3
RP
3339 if (data[IFLA_VXLAN_REMCSUM_TX]) {
3340 if (changelink)
3341 return -EOPNOTSUPP;
3342 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
3343 conf->flags |= VXLAN_F_REMCSUM_TX;
3344 }
dfd8645e 3345
8bcdc4f3
RP
3346 if (data[IFLA_VXLAN_REMCSUM_RX]) {
3347 if (changelink)
3348 return -EOPNOTSUPP;
3349 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3350 conf->flags |= VXLAN_F_REMCSUM_RX;
3351 }
3352
3353 if (data[IFLA_VXLAN_GBP]) {
3354 if (changelink)
3355 return -EOPNOTSUPP;
3356 conf->flags |= VXLAN_F_GBP;
3357 }
3358
3359 if (data[IFLA_VXLAN_GPE]) {
3360 if (changelink)
3361 return -EOPNOTSUPP;
3362 conf->flags |= VXLAN_F_GPE;
3363 }
3364
3365 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3366 if (changelink)
3367 return -EOPNOTSUPP;
3368 conf->flags |= VXLAN_F_REMCSUM_NOPARTIAL;
3369 }
3370
3371 if (tb[IFLA_MTU]) {
3372 if (changelink)
3373 return -EOPNOTSUPP;
3374 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3375 }
3376
3377 return 0;
3378}
3379
3380static int vxlan_newlink(struct net *src_net, struct net_device *dev,
7a3f4a18
MS
3381 struct nlattr *tb[], struct nlattr *data[],
3382 struct netlink_ext_ack *extack)
8bcdc4f3 3383{
8bcdc4f3
RP
3384 struct vxlan_config conf;
3385 int err;
3386
3387 err = vxlan_nl2conf(tb, data, dev, &conf, false);
3388 if (err)
3389 return err;
3390
653ef6a3 3391 return __vxlan_dev_create(src_net, dev, &conf, extack);
8bcdc4f3 3392}
dfd8645e 3393
8bcdc4f3 3394static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
ad744b22
MS
3395 struct nlattr *data[],
3396 struct netlink_ext_ack *extack)
8bcdc4f3
RP
3397{
3398 struct vxlan_dev *vxlan = netdev_priv(dev);
3399 struct vxlan_rdst *dst = &vxlan->default_dst;
3400 struct vxlan_rdst old_dst;
3401 struct vxlan_config conf;
3402 int err;
3403
3404 err = vxlan_nl2conf(tb, data,
3405 dev, &conf, true);
3406 if (err)
3407 return err;
3511494c 3408
8bcdc4f3 3409 memcpy(&old_dst, dst, sizeof(struct vxlan_rdst));
e1e5314d 3410
653ef6a3 3411 err = vxlan_dev_configure(vxlan->net, dev, &conf, true, extack);
8bcdc4f3
RP
3412 if (err)
3413 return err;
0ace2ca8 3414
8bcdc4f3
RP
3415 /* handle default dst entry */
3416 if (!vxlan_addr_equal(&dst->remote_ip, &old_dst.remote_ip)) {
3417 spin_lock_bh(&vxlan->hash_lock);
3418 if (!vxlan_addr_any(&old_dst.remote_ip))
3419 __vxlan_fdb_delete(vxlan, all_zeros_mac,
3420 old_dst.remote_ip,
3421 vxlan->cfg.dst_port,
3422 old_dst.remote_vni,
3423 old_dst.remote_vni,
3424 old_dst.remote_ifindex, 0);
3425
3426 if (!vxlan_addr_any(&dst->remote_ip)) {
3427 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3428 &dst->remote_ip,
3429 NUD_REACHABLE | NUD_PERMANENT,
3430 NLM_F_CREATE | NLM_F_APPEND,
3431 vxlan->cfg.dst_port,
3432 dst->remote_vni,
3433 dst->remote_vni,
3434 dst->remote_ifindex,
3435 NTF_SELF);
3436 if (err) {
3437 spin_unlock_bh(&vxlan->hash_lock);
3438 return err;
3439 }
3440 }
3441 spin_unlock_bh(&vxlan->hash_lock);
3442 }
ce577668 3443
8bcdc4f3 3444 return 0;
d342894c 3445}
3446
3447static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3448{
3449 struct vxlan_dev *vxlan = netdev_priv(dev);
3450
8b3f9337
RP
3451 vxlan_flush(vxlan, true);
3452
58ce31cc 3453 gro_cells_destroy(&vxlan->gro_cells);
553675fb 3454 list_del(&vxlan->next);
d342894c 3455 unregister_netdevice_queue(dev, head);
3456}
3457
3458static size_t vxlan_get_size(const struct net_device *dev)
3459{
3460
3461 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3462 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3463 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3464 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3465 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3466 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
e7f70af1 3467 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 3468 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS
3469 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3470 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3471 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3472 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3473 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3474 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3475 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3476 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH
3477 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3478 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3479 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3480 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH
3481 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3482 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3483 0;
3484}
3485
3486static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3487{
3488 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3489 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3490 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG
3491 .low = htons(vxlan->cfg.port_min),
3492 .high = htons(vxlan->cfg.port_max),
05f47d69 3493 };
d342894c 3494
54bfd872 3495 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3496 goto nla_put_failure;
3497
e4c7ed41
CW
3498 if (!vxlan_addr_any(&dst->remote_ip)) {
3499 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB
3500 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3501 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW
3502 goto nla_put_failure;
3503#if IS_ENABLED(CONFIG_IPV6)
3504 } else {
930345ea
JB
3505 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3506 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW
3507 goto nla_put_failure;
3508#endif
3509 }
3510 }
d342894c 3511
c7995c43 3512 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3513 goto nla_put_failure;
3514
0dfbdf41
TG
3515 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3516 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3517 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3518 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW
3519 goto nla_put_failure;
3520#if IS_ENABLED(CONFIG_IPV6)
3521 } else {
930345ea 3522 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3523 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW
3524 goto nla_put_failure;
3525#endif
3526 }
3527 }
d342894c 3528
0dfbdf41
TG
3529 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3530 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e7f70af1 3531 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add 3532 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
dc5321d7 3533 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
e4f67add 3534 nla_put_u8(skb, IFLA_VXLAN_PROXY,
dc5321d7
MS
3535 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
3536 nla_put_u8(skb, IFLA_VXLAN_RSC,
3537 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
e4f67add 3538 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
dc5321d7 3539 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
e4f67add 3540 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
dc5321d7 3541 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
da8b43c0 3542 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
dc5321d7 3543 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG
3544 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3545 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3546 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3547 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
dc5321d7 3548 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9 3549 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
dc5321d7 3550 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
359a0ea9 3551 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dc5321d7 3552 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
dfd8645e 3553 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
dc5321d7 3554 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
dfd8645e 3555 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
dc5321d7 3556 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
d342894c 3557 goto nla_put_failure;
3558
05f47d69 3559 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3560 goto nla_put_failure;
3561
dc5321d7 3562 if (vxlan->cfg.flags & VXLAN_F_GBP &&
3511494c
TG
3563 nla_put_flag(skb, IFLA_VXLAN_GBP))
3564 goto nla_put_failure;
3565
dc5321d7 3566 if (vxlan->cfg.flags & VXLAN_F_GPE &&
e1e5314d
JB
3567 nla_put_flag(skb, IFLA_VXLAN_GPE))
3568 goto nla_put_failure;
3569
dc5321d7 3570 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
0ace2ca8
TH
3571 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3572 goto nla_put_failure;
3573
d342894c 3574 return 0;
3575
3576nla_put_failure:
3577 return -EMSGSIZE;
3578}
3579
1728d4fa
ND
3580static struct net *vxlan_get_link_net(const struct net_device *dev)
3581{
3582 struct vxlan_dev *vxlan = netdev_priv(dev);
3583
3584 return vxlan->net;
3585}
3586
d342894c 3587static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3588 .kind = "vxlan",
3589 .maxtype = IFLA_VXLAN_MAX,
3590 .policy = vxlan_policy,
3591 .priv_size = sizeof(struct vxlan_dev),
3592 .setup = vxlan_setup,
3593 .validate = vxlan_validate,
3594 .newlink = vxlan_newlink,
8bcdc4f3 3595 .changelink = vxlan_changelink,
d342894c 3596 .dellink = vxlan_dellink,
3597 .get_size = vxlan_get_size,
3598 .fill_info = vxlan_fill_info,
1728d4fa 3599 .get_link_net = vxlan_get_link_net,
d342894c 3600};
3601
cf5da330
ND
3602struct net_device *vxlan_dev_create(struct net *net, const char *name,
3603 u8 name_assign_type,
3604 struct vxlan_config *conf)
3605{
3606 struct nlattr *tb[IFLA_MAX + 1];
3607 struct net_device *dev;
3608 int err;
3609
3610 memset(&tb, 0, sizeof(tb));
3611
3612 dev = rtnl_create_link(net, name, name_assign_type,
3613 &vxlan_link_ops, tb);
3614 if (IS_ERR(dev))
3615 return dev;
3616
653ef6a3 3617 err = __vxlan_dev_create(net, dev, conf, NULL);
cf5da330
ND
3618 if (err < 0) {
3619 free_netdev(dev);
3620 return ERR_PTR(err);
3621 }
3622
3623 err = rtnl_configure_link(dev, NULL);
3624 if (err < 0) {
3625 LIST_HEAD(list_kill);
3626
3627 vxlan_dellink(dev, &list_kill);
3628 unregister_netdevice_many(&list_kill);
3629 return ERR_PTR(err);
3630 }
3631
3632 return dev;
3633}
3634EXPORT_SYMBOL_GPL(vxlan_dev_create);
3635
acaf4e70
DB
3636static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3637 struct net_device *dev)
3638{
3639 struct vxlan_dev *vxlan, *next;
3640 LIST_HEAD(list_kill);
3641
3642 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3643 struct vxlan_rdst *dst = &vxlan->default_dst;
3644
3645 /* In case we created vxlan device with carrier
3646 * and we loose the carrier due to module unload
3647 * we also need to remove vxlan device. In other
3648 * cases, it's not necessary and remote_ifindex
3649 * is 0 here, so no matches.
3650 */
3651 if (dst->remote_ifindex == dev->ifindex)
3652 vxlan_dellink(vxlan->dev, &list_kill);
3653 }
3654
3655 unregister_netdevice_many(&list_kill);
3656}
3657
b7aade15
HFS
3658static int vxlan_netdevice_event(struct notifier_block *unused,
3659 unsigned long event, void *ptr)
acaf4e70
DB
3660{
3661 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3662 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3663
04584957
SD
3664 if (event == NETDEV_UNREGISTER) {
3665 vxlan_offload_rx_ports(dev, false);
acaf4e70 3666 vxlan_handle_lowerdev_unregister(vn, dev);
04584957
SD
3667 } else if (event == NETDEV_REGISTER) {
3668 vxlan_offload_rx_ports(dev, true);
3669 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
3670 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
2d2b13fc 3671 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
04584957 3672 }
acaf4e70
DB
3673
3674 return NOTIFY_DONE;
3675}
3676
3677static struct notifier_block vxlan_notifier_block __read_mostly = {
b7aade15 3678 .notifier_call = vxlan_netdevice_event,
acaf4e70
DB
3679};
3680
d342894c 3681static __net_init int vxlan_init_net(struct net *net)
3682{
3683 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3684 unsigned int h;
d342894c 3685
553675fb 3686 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3687 spin_lock_init(&vn->sock_lock);
d342894c 3688
553675fb 3689 for (h = 0; h < PORT_HASH_SIZE; ++h)
3690 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3691
3692 return 0;
3693}
3694
f01ec1c0
ND
3695static void __net_exit vxlan_exit_net(struct net *net)
3696{
3697 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3698 struct vxlan_dev *vxlan, *next;
3699 struct net_device *dev, *aux;
0e4ec5ac 3700 unsigned int h;
f01ec1c0
ND
3701 LIST_HEAD(list);
3702
3703 rtnl_lock();
3704 for_each_netdev_safe(net, dev, aux)
3705 if (dev->rtnl_link_ops == &vxlan_link_ops)
3706 unregister_netdevice_queue(dev, &list);
3707
3708 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3709 /* If vxlan->dev is in the same netns, it has already been added
3710 * to the list by the previous loop.
3711 */
58ce31cc
TH
3712 if (!net_eq(dev_net(vxlan->dev), net)) {
3713 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3714 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3715 }
f01ec1c0
ND
3716 }
3717
3718 unregister_netdevice_many(&list);
3719 rtnl_unlock();
0e4ec5ac
VA
3720
3721 for (h = 0; h < PORT_HASH_SIZE; ++h)
3722 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
f01ec1c0
ND
3723}
3724
d342894c 3725static struct pernet_operations vxlan_net_ops = {
3726 .init = vxlan_init_net,
f01ec1c0 3727 .exit = vxlan_exit_net,
d342894c 3728 .id = &vxlan_net_id,
3729 .size = sizeof(struct vxlan_net),
3730};
3731
3732static int __init vxlan_init_module(void)
3733{
3734 int rc;
3735
3736 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3737
783c1463 3738 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3739 if (rc)
3740 goto out1;
3741
acaf4e70 3742 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3743 if (rc)
3744 goto out2;
3745
acaf4e70
DB
3746 rc = rtnl_link_register(&vxlan_link_ops);
3747 if (rc)
3748 goto out3;
d342894c 3749
acaf4e70
DB
3750 return 0;
3751out3:
3752 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3753out2:
783c1463 3754 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3755out1:
3756 return rc;
3757}
7332a13b 3758late_initcall(vxlan_init_module);
d342894c 3759
3760static void __exit vxlan_cleanup_module(void)
3761{
b7153984 3762 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3763 unregister_netdevice_notifier(&vxlan_notifier_block);
783c1463
DB
3764 unregister_pernet_subsys(&vxlan_net_ops);
3765 /* rcu_barrier() is called by netns */
d342894c 3766}
3767module_exit(vxlan_cleanup_module);
3768
3769MODULE_LICENSE("GPL");
3770MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3771MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3772MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3773MODULE_ALIAS_RTNL_LINK("vxlan");