]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-hirsute-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
vxlan: make netlink notify in vxlan_fdb_destroy optional
[mirror_ubuntu-hirsute-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:
603d4cf8 626 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
dc01e7d3
OG		 627
628 return pp;
629}
630
5602c48c 631static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
dc01e7d3 632{
229740c6
JR		 633 /* Sets 'skb->inner_mac_header' since we are always called with
634 * 'skb->encapsulation' set.
635 */
9b174d88 636 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 637}
638
7431016b
RP		 639static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan,
640 const u8 *mac, __u16 state,
641 __be32 src_vni, __u8 ndm_flags)
642{
643 struct vxlan_fdb *f;
644
645 f = kmalloc(sizeof(*f), GFP_ATOMIC);
646 if (!f)
647 return NULL;
648 f->state = state;
649 f->flags = ndm_flags;
650 f->updated = f->used = jiffies;
651 f->vni = src_vni;
652 INIT_LIST_HEAD(&f->remotes);
653 memcpy(f->eth_addr, mac, ETH_ALEN);
654
655 return f;
656}
657
d342894c 658static int vxlan_fdb_create(struct vxlan_dev *vxlan,
7431016b
RP		 659 const u8 *mac, union vxlan_addr *ip,
660 __u16 state, __be16 port, __be32 src_vni,
661 __be32 vni, __u32 ifindex, __u8 ndm_flags,
662 struct vxlan_fdb **fdb)
663{
664 struct vxlan_rdst *rd = NULL;
665 struct vxlan_fdb *f;
666 int rc;
667
668 if (vxlan->cfg.addrmax &&
669 vxlan->addrcnt >= vxlan->cfg.addrmax)
670 return -ENOSPC;
671
672 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
673 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
674 if (!f)
675 return -ENOMEM;
676
677 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
678 if (rc < 0) {
679 kfree(f);
680 return rc;
681 }
682
683 ++vxlan->addrcnt;
684 hlist_add_head_rcu(&f->hlist,
685 vxlan_fdb_head(vxlan, mac, src_vni));
686
687 *fdb = f;
688
689 return 0;
690}
691
692/* Add new entry to forwarding table -- assumes lock held */
693static int vxlan_fdb_update(struct vxlan_dev *vxlan,
e4c7ed41 694 const u8 *mac, union vxlan_addr *ip,
6681712d 695 __u16 state, __u16 flags,
3ad7a4b1
RP		 696 __be16 port, __be32 src_vni, __be32 vni,
697 __u32 ifindex, __u8 ndm_flags)
d342894c 698{
9e4b93f9 699 struct vxlan_rdst *rd = NULL;
d342894c 700 struct vxlan_fdb *f;
701 int notify = 0;
17b46365 702 int rc;
d342894c 703
3ad7a4b1 704 f = __vxlan_find_mac(vxlan, mac, src_vni);
d342894c 705 if (f) {
706 if (flags & NLM_F_EXCL) {
707 netdev_dbg(vxlan->dev,
708 "lost race to create %pM\n", mac);
709 return -EEXIST;
710 }
711 if (f->state != state) {
712 f->state = state;
713 f->updated = jiffies;
714 notify = 1;
715 }
ae884082
DS		 716 if (f->flags != ndm_flags) {
717 f->flags = ndm_flags;
718 f->updated = jiffies;
719 notify = 1;
720 }
906dc186
TR		 721 if ((flags & NLM_F_REPLACE)) {
722 /* Only change unicasts */
723 if (!(is_multicast_ether_addr(f->eth_addr) ||
724 is_zero_ether_addr(f->eth_addr))) {
60840429 725 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 726 ifindex);
906dc186
TR		 727 } else
728 return -EOPNOTSUPP;
729 }
6681712d 730 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 731 (is_multicast_ether_addr(f->eth_addr) ||
732 is_zero_ether_addr(f->eth_addr))) {
17b46365 733 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
6681712d
DS		 734
735 if (rc < 0)
736 return rc;
737 notify |= rc;
738 }
d342894c 739 } else {
740 if (!(flags & NLM_F_CREATE))
741 return -ENOENT;
742
906dc186
TR		 743 /* Disallow replace to add a multicast entry */
744 if ((flags & NLM_F_REPLACE) &&
745 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
746 return -EOPNOTSUPP;
747
e4c7ed41 748 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
7431016b
RP		 749 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
750 vni, ifindex, ndm_flags, &f);
751 if (rc < 0)
17b46365 752 return rc;
7431016b 753 notify = 1;
d342894c 754 }
755
9e4b93f9
ND		 756 if (notify) {
757 if (rd == NULL)
758 rd = first_remote_rtnl(f);
759 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
760 }
d342894c 761
762 return 0;
763}
764
6706c82e 765static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 766{
767 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 768 struct vxlan_rdst *rd, *nd;
6681712d 769
0c1d70af
PA		 770 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
771 dst_cache_destroy(&rd->dst_cache);
6681712d 772 kfree(rd);
0c1d70af 773 }
6681712d
DS		 774 kfree(f);
775}
776
f6e05385
RP		 777static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
778 bool do_notify)
d342894c 779{
780 netdev_dbg(vxlan->dev,
781 "delete %pM\n", f->eth_addr);
782
783 --vxlan->addrcnt;
f6e05385
RP		 784 if (do_notify)
785 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 786
787 hlist_del_rcu(&f->hlist);
6681712d 788 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 789}
790
35cf2845
LR		 791static void vxlan_dst_free(struct rcu_head *head)
792{
793 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
794
795 dst_cache_destroy(&rd->dst_cache);
796 kfree(rd);
797}
798
799static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
800 struct vxlan_rdst *rd)
801{
802 list_del_rcu(&rd->list);
803 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
804 call_rcu(&rd->rcu, vxlan_dst_free);
805}
806
f0b074be 807static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
3ad7a4b1
RP		 808 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
809 __be32 *vni, u32 *ifindex)
d342894c 810{
6681712d 811 struct net *net = dev_net(vxlan->dev);
e4c7ed41 812 int err;
d342894c 813
f0b074be 814 if (tb[NDA_DST]) {
e4c7ed41
CW		 815 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
816 if (err)
817 return err;
f0b074be 818 } else {
e4c7ed41
CW		 819 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
820 if (remote->sa.sa_family == AF_INET) {
821 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
822 ip->sa.sa_family = AF_INET;
823#if IS_ENABLED(CONFIG_IPV6)
824 } else {
825 ip->sin6.sin6_addr = in6addr_any;
826 ip->sa.sa_family = AF_INET6;
827#endif
828 }
f0b074be 829 }
d342894c 830
6681712d 831 if (tb[NDA_PORT]) {
73cf3317 832 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 833 return -EINVAL;
f0b074be
MR		 834 *port = nla_get_be16(tb[NDA_PORT]);
835 } else {
0dfbdf41 836 *port = vxlan->cfg.dst_port;
f0b074be 837 }
6681712d
DS		 838
839 if (tb[NDA_VNI]) {
840 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
841 return -EINVAL;
54bfd872 842 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 843 } else {
844 *vni = vxlan->default_dst.remote_vni;
845 }
6681712d 846
3ad7a4b1
RP		 847 if (tb[NDA_SRC_VNI]) {
848 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
849 return -EINVAL;
850 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
851 } else {
852 *src_vni = vxlan->default_dst.remote_vni;
853 }
854
6681712d 855 if (tb[NDA_IFINDEX]) {
5abb0029 856 struct net_device *tdev;
6681712d
DS		 857
858 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
859 return -EINVAL;
f0b074be 860 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 861 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 862 if (!tdev)
6681712d 863 return -EADDRNOTAVAIL;
f0b074be
MR		 864 } else {
865 *ifindex = 0;
866 }
867
868 return 0;
869}
870
871/* Add static entry (via netlink) */
872static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
873 struct net_device *dev,
f6f6424b 874 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 875{
876 struct vxlan_dev *vxlan = netdev_priv(dev);
877 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 878 union vxlan_addr ip;
f0b074be 879 __be16 port;
3ad7a4b1 880 __be32 src_vni, vni;
54bfd872 881 u32 ifindex;
f0b074be
MR		 882 int err;
883
884 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
885 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
886 ndm->ndm_state);
887 return -EINVAL;
888 }
889
890 if (tb[NDA_DST] == NULL)
891 return -EINVAL;
892
3ad7a4b1 893 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
f0b074be
MR		 894 if (err)
895 return err;
6681712d 896
5933a7bb
MR		 897 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
898 return -EAFNOSUPPORT;
899
d342894c 900 spin_lock_bh(&vxlan->hash_lock);
7431016b 901 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
3ad7a4b1 902 port, src_vni, vni, ifindex, ndm->ndm_flags);
d342894c 903 spin_unlock_bh(&vxlan->hash_lock);
904
905 return err;
906}
907
3ad7a4b1
RP		 908static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
909 const unsigned char *addr, union vxlan_addr ip,
fc39c38b
XL		 910 __be16 port, __be32 src_vni, __be32 vni,
911 u32 ifindex, u16 vid)
d342894c 912{
d342894c 913 struct vxlan_fdb *f;
bc7892ba 914 struct vxlan_rdst *rd = NULL;
3ad7a4b1 915 int err = -ENOENT;
bc7892ba 916
3ad7a4b1 917 f = vxlan_find_mac(vxlan, addr, src_vni);
bc7892ba 918 if (!f)
3ad7a4b1 919 return err;
bc7892ba 920
e4c7ed41
CW		 921 if (!vxlan_addr_any(&ip)) {
922 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 923 if (!rd)
924 goto out;
925 }
926
bc7892ba
MR		 927 /* remove a destination if it's not the only one on the list,
928 * otherwise destroy the fdb entry
929 */
930 if (rd && !list_is_singular(&f->remotes)) {
35cf2845 931 vxlan_fdb_dst_destroy(vxlan, f, rd);
bc7892ba 932 goto out;
d342894c 933 }
bc7892ba 934
f6e05385 935 vxlan_fdb_destroy(vxlan, f, true);
bc7892ba
MR		 936
937out:
3ad7a4b1
RP		 938 return 0;
939}
940
941/* Delete entry (via netlink) */
942static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
943 struct net_device *dev,
944 const unsigned char *addr, u16 vid)
945{
946 struct vxlan_dev *vxlan = netdev_priv(dev);
947 union vxlan_addr ip;
948 __be32 src_vni, vni;
949 __be16 port;
950 u32 ifindex;
951 int err;
952
953 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
954 if (err)
955 return err;
956
957 spin_lock_bh(&vxlan->hash_lock);
958 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
959 vid);
d342894c 960 spin_unlock_bh(&vxlan->hash_lock);
961
962 return err;
963}
964
965/* Dump forwarding table */
966static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 967 struct net_device *dev,
d297653d 968 struct net_device *filter_dev, int *idx)
d342894c 969{
970 struct vxlan_dev *vxlan = netdev_priv(dev);
971 unsigned int h;
d297653d 972 int err = 0;
d342894c 973
974 for (h = 0; h < FDB_HASH_SIZE; ++h) {
975 struct vxlan_fdb *f;
d342894c 976
b67bfe0d 977 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 978 struct vxlan_rdst *rd;
6681712d 979
3e61aa8f 980 list_for_each_entry_rcu(rd, &f->remotes, list) {
d297653d 981 if (*idx < cb->args[2])
07a51cd3
AW		 982 goto skip;
983
6681712d
DS		 984 err = vxlan_fdb_info(skb, vxlan, f,
985 NETLINK_CB(cb->skb).portid,
986 cb->nlh->nlmsg_seq,
987 RTM_NEWNEIGH,
988 NLM_F_MULTI, rd);
d297653d 989 if (err < 0)
3e61aa8f 990 goto out;
3e61aa8f 991skip:
d297653d 992 *idx += 1;
07a51cd3 993 }
d342894c 994 }
995 }
3e61aa8f 996out:
d297653d 997 return err;
d342894c 998}
999
1000/* Watch incoming packets to learn mapping between Ethernet address
1001 * and Tunnel endpoint.
c4b49512 1002 * Return true if packet is bogus and should be dropped.
d342894c 1003 */
26a41ae6 1004static bool vxlan_snoop(struct net_device *dev,
3ad7a4b1 1005 union vxlan_addr *src_ip, const u8 *src_mac,
87613de9 1006 u32 src_ifindex, __be32 vni)
d342894c 1007{
1008 struct vxlan_dev *vxlan = netdev_priv(dev);
1009 struct vxlan_fdb *f;
87613de9
MS		 1010 u32 ifindex = 0;
1011
1012#if IS_ENABLED(CONFIG_IPV6)
1013 if (src_ip->sa.sa_family == AF_INET6 &&
1014 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1015 ifindex = src_ifindex;
1016#endif
d342894c 1017
3ad7a4b1 1018 f = vxlan_find_mac(vxlan, src_mac, vni);
d342894c 1019 if (likely(f)) {
5ca5461c 1020 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 1021
87613de9
MS		 1022 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1023 rdst->remote_ifindex == ifindex))
26a41ae6 1024 return false;
1025
1026 /* Don't migrate static entries, drop packets */
e0090a9e 1027 if (f->state & (NUD_PERMANENT | NUD_NOARP))
26a41ae6 1028 return true;
d342894c 1029
1030 if (net_ratelimit())
1031 netdev_info(dev,
e4c7ed41 1032 "%pM migrated from %pIS to %pIS\n",
a4870f79 1033 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 1034
e4c7ed41 1035 rdst->remote_ip = *src_ip;
d342894c 1036 f->updated = jiffies;
9e4b93f9 1037 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 1038 } else {
1039 /* learned new entry */
1040 spin_lock(&vxlan->hash_lock);
3bf74b1a 1041
1042 /* close off race between vxlan_flush and incoming packets */
1043 if (netif_running(dev))
7431016b 1044 vxlan_fdb_update(vxlan, src_mac, src_ip,
3bf74b1a 1045 NUD_REACHABLE,
1046 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 1047 vxlan->cfg.dst_port,
3ad7a4b1 1048 vni,
3bf74b1a 1049 vxlan->default_dst.remote_vni,
87613de9 1050 ifindex, NTF_SELF);
d342894c 1051 spin_unlock(&vxlan->hash_lock);
1052 }
26a41ae6 1053
1054 return false;
d342894c 1055}
1056
d342894c 1057/* See if multicast group is already in use by other ID */
95ab0991 1058static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1059{
553675fb 1060 struct vxlan_dev *vxlan;
c6fcc4fc 1061 struct vxlan_sock *sock4;
4053ab1b
AB		 1062#if IS_ENABLED(CONFIG_IPV6)
1063 struct vxlan_sock *sock6;
1064#endif
b1be00a6 1065 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1066
c6fcc4fc 1067 sock4 = rtnl_dereference(dev->vn4_sock);
1068
95ab0991
G		 1069 /* The vxlan_sock is only used by dev, leaving group has
1070 * no effect on other vxlan devices.
1071 */
66af846f 1072 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
95ab0991 1073 return false;
b1be00a6 1074#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1075 sock6 = rtnl_dereference(dev->vn6_sock);
66af846f 1076 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
b1be00a6
JB		 1077 return false;
1078#endif
95ab0991 1079
553675fb 1080 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1081 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1082 continue;
d342894c 1083
c6fcc4fc 1084 if (family == AF_INET &&
1085 rtnl_dereference(vxlan->vn4_sock) != sock4)
95ab0991 1086 continue;
b1be00a6 1087#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1088 if (family == AF_INET6 &&
1089 rtnl_dereference(vxlan->vn6_sock) != sock6)
b1be00a6
JB		 1090 continue;
1091#endif
95ab0991
G		 1092
1093 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1094 &dev->default_dst.remote_ip))
1095 continue;
1096
1097 if (vxlan->default_dst.remote_ifindex !=
1098 dev->default_dst.remote_ifindex)
1099 continue;
1100
1101 return true;
553675fb 1102 }
d342894c 1103
1104 return false;
1105}
1106
544a773a 1107static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
7c47cedf 1108{
b1be00a6 1109 struct vxlan_net *vn;
012a5729 1110
b1be00a6 1111 if (!vs)
544a773a 1112 return false;
66af846f 1113 if (!refcount_dec_and_test(&vs->refcnt))
544a773a 1114 return false;
d342894c 1115
b1be00a6 1116 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1117 spin_lock(&vn->sock_lock);
7c47cedf 1118 hlist_del_rcu(&vs->hlist);
e7b3db5e 1119 udp_tunnel_notify_del_rx_port(vs->sock,
b9adcd69
AD		 1120 (vs->flags & VXLAN_F_GPE) ?
1121 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 1122 UDP_TUNNEL_TYPE_VXLAN);
1c51a915
SH		 1123 spin_unlock(&vn->sock_lock);
1124
544a773a 1125 return true;
d342894c 1126}
1127
b1be00a6
JB		 1128static void vxlan_sock_release(struct vxlan_dev *vxlan)
1129{
c6fcc4fc 1130 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
b1be00a6 1131#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1132 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1133
57d88182 1134 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
544a773a
HFS		 1135#endif
1136
57d88182 1137 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
544a773a
HFS		 1138 synchronize_net();
1139
a53cb29b
MB		 1140 vxlan_vs_del_dev(vxlan);
1141
c6fcc4fc 1142 if (__vxlan_sock_release_prep(sock4)) {
1143 udp_tunnel_sock_release(sock4->sock);
1144 kfree(sock4);
544a773a
HFS		 1145 }
1146
1147#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1148 if (__vxlan_sock_release_prep(sock6)) {
1149 udp_tunnel_sock_release(sock6->sock);
1150 kfree(sock6);
544a773a 1151 }
b1be00a6
JB		 1152#endif
1153}
1154
56ef9c90 1155/* Update multicast group membership when first VNI on
c4b49512 1156 * multicast address is brought up
7c47cedf 1157 */
56ef9c90 1158static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1159{
b1be00a6 1160 struct sock *sk;
e4c7ed41
CW		 1161 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1162 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1163 int ret = -EINVAL;
d342894c 1164
e4c7ed41 1165 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1166 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1167 struct ip_mreqn mreq = {
1168 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1169 .imr_ifindex = ifindex,
1170 };
1171
c6fcc4fc 1172 sk = sock4->sock->sk;
b1be00a6 1173 lock_sock(sk);
56ef9c90 1174 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1175 release_sock(sk);
e4c7ed41
CW		 1176#if IS_ENABLED(CONFIG_IPV6)
1177 } else {
c6fcc4fc 1178 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1179
1180 sk = sock6->sock->sk;
b1be00a6 1181 lock_sock(sk);
56ef9c90
MRL		 1182 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1183 &ip->sin6.sin6_addr);
b1be00a6 1184 release_sock(sk);
e4c7ed41
CW		 1185#endif
1186 }
3fc2de2f 1187
56ef9c90 1188 return ret;
3fc2de2f 1189}
1190
1191/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1192static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1193{
b1be00a6 1194 struct sock *sk;
e4c7ed41
CW		 1195 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1196 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1197 int ret = -EINVAL;
3fc2de2f 1198
e4c7ed41 1199 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1200 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1201 struct ip_mreqn mreq = {
1202 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1203 .imr_ifindex = ifindex,
1204 };
1205
c6fcc4fc 1206 sk = sock4->sock->sk;
b1be00a6 1207 lock_sock(sk);
56ef9c90 1208 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1209 release_sock(sk);
e4c7ed41
CW		 1210#if IS_ENABLED(CONFIG_IPV6)
1211 } else {
c6fcc4fc 1212 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1213
1214 sk = sock6->sock->sk;
b1be00a6 1215 lock_sock(sk);
56ef9c90
MRL		 1216 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1217 &ip->sin6.sin6_addr);
b1be00a6 1218 release_sock(sk);
e4c7ed41
CW		 1219#endif
1220 }
d342894c 1221
56ef9c90 1222 return ret;
d342894c 1223}
1224
f14ecebb
JB		 1225static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1226 struct sk_buff *skb, u32 vxflags)
dfd8645e 1227{
7d34fa75 1228 size_t start, offset;
dfd8645e 1229
f14ecebb
JB		 1230 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1231 goto out;
b7fe10e5 1232
f14ecebb
JB		 1233 start = vxlan_rco_start(unparsed->vx_vni);
1234 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1235
7d34fa75 1236 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1237 return false;
dfd8645e 1238
be5cfeab
JB		 1239 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1240 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1241out:
1242 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1243 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1244 return true;
dfd8645e
TH		 1245}
1246
f14ecebb 1247static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1248 struct sk_buff *skb, u32 vxflags,
10a5af23 1249 struct vxlan_metadata *md)
3288af08 1250{
f14ecebb 1251 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1252 struct metadata_dst *tun_dst;
f14ecebb
JB		 1253
1254 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1255 goto out;
3288af08 1256
3288af08
JB		 1257 md->gbp = ntohs(gbp->policy_id);
1258
10a5af23 1259 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1260 if (tun_dst) {
3288af08 1261 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1262 tun_dst->u.tun_info.options_len = sizeof(*md);
1263 }
3288af08
JB		 1264 if (gbp->dont_learn)
1265 md->gbp |= VXLAN_GBP_DONT_LEARN;
1266
1267 if (gbp->policy_applied)
1268 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1269
64f87d36
JB		 1270 /* In flow-based mode, GBP is carried in dst_metadata */
1271 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1272 skb->mark = md->gbp;
f14ecebb
JB		 1273out:
1274 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1275}
1276
e1e5314d 1277static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
61618eea 1278 __be16 *protocol,
e1e5314d
JB		 1279 struct sk_buff *skb, u32 vxflags)
1280{
1281 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1282
1283 /* Need to have Next Protocol set for interfaces in GPE mode. */
1284 if (!gpe->np_applied)
1285 return false;
1286 /* "The initial version is 0. If a receiver does not support the
1287 * version indicated it MUST drop the packet.
1288 */
1289 if (gpe->version != 0)
1290 return false;
1291 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1292 * processing MUST occur." However, we don't implement OAM
1293 * processing, thus drop the packet.
1294 */
1295 if (gpe->oam_flag)
1296 return false;
1297
fa20e0e3
JB		 1298 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1299 if (!*protocol)
e1e5314d 1300 return false;
e1e5314d
JB		 1301
1302 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1303 return true;
1304}
1305
1ab016e2
JB		 1306static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1307 struct vxlan_sock *vs,
3ad7a4b1 1308 struct sk_buff *skb, __be32 vni)
614732ea 1309{
614732ea 1310 union vxlan_addr saddr;
87613de9 1311 u32 ifindex = skb->dev->ifindex;
614732ea 1312
614732ea 1313 skb_reset_mac_header(skb);
614732ea
TG		 1314 skb->protocol = eth_type_trans(skb, vxlan->dev);
1315 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1316
1317 /* Ignore packet loops (and multicast echo) */
1318 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1319 return false;
614732ea 1320
760c6805 1321 /* Get address from the outer IP header */
ce212d0f 1322 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1323 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1324 saddr.sa.sa_family = AF_INET;
1325#if IS_ENABLED(CONFIG_IPV6)
1326 } else {
1ab016e2 1327 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1328 saddr.sa.sa_family = AF_INET6;
1329#endif
1330 }
1331
dc5321d7 1332 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
87613de9 1333 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1ab016e2
JB		 1334 return false;
1335
1336 return true;
1337}
1338
760c6805
JB		 1339static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1340 struct sk_buff *skb)
1341{
1342 int err = 0;
1343
1344 if (vxlan_get_sk_family(vs) == AF_INET)
1345 err = IP_ECN_decapsulate(oiph, skb);
1346#if IS_ENABLED(CONFIG_IPV6)
1347 else
1348 err = IP6_ECN_decapsulate(oiph, skb);
1349#endif
1350
1351 if (unlikely(err) && log_ecn_error) {
1352 if (vxlan_get_sk_family(vs) == AF_INET)
1353 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1354 &((struct iphdr *)oiph)->saddr,
1355 ((struct iphdr *)oiph)->tos);
1356 else
1357 net_info_ratelimited("non-ECT from %pI6\n",
1358 &((struct ipv6hdr *)oiph)->saddr);
1359 }
1360 return err <= 1;
1361}
1362
d342894c 1363/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1364static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1365{
f2d1968e 1366 struct pcpu_sw_netstats *stats;
c9e78efb 1367 struct vxlan_dev *vxlan;
5cfccc5a 1368 struct vxlan_sock *vs;
f14ecebb 1369 struct vxlanhdr unparsed;
ee122c79
TG		 1370 struct vxlan_metadata _md;
1371 struct vxlan_metadata *md = &_md;
61618eea 1372 __be16 protocol = htons(ETH_P_TEB);
e1e5314d 1373 bool raw_proto = false;
f2d1968e 1374 void *oiph;
3ad7a4b1 1375 __be32 vni = 0;
d342894c 1376
e1e5314d 1377 /* Need UDP and VXLAN header to be present */
7ce04758 1378 if (!pskb_may_pull(skb, VXLAN_HLEN))
e5aed006 1379 goto drop;
d342894c 1380
f14ecebb 1381 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1382 /* VNI flag always required to be set */
1383 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1384 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1385 ntohl(vxlan_hdr(skb)->vx_flags),
1386 ntohl(vxlan_hdr(skb)->vx_vni));
1387 /* Return non vxlan pkt */
e5aed006 1388 goto drop;
d342894c 1389 }
288b01c8
JB		 1390 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1391 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1392
559835ea 1393 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1394 if (!vs)
d342894c 1395 goto drop;
d342894c 1396
3ad7a4b1
RP		 1397 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1398
49f810f0 1399 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
c9e78efb
JB		 1400 if (!vxlan)
1401 goto drop;
1402
e1e5314d
JB		 1403 /* For backwards compatibility, only allow reserved fields to be
1404 * used by VXLAN extensions if explicitly requested.
1405 */
1406 if (vs->flags & VXLAN_F_GPE) {
1407 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1408 goto drop;
1409 raw_proto = true;
1410 }
1411
1412 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1413 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1414 goto drop;
c9e78efb 1415
ee122c79 1416 if (vxlan_collect_metadata(vs)) {
10a5af23 1417 struct metadata_dst *tun_dst;
07dabf20 1418
c29a70d2 1419 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
d817f432 1420 key32_to_tunnel_id(vni), sizeof(*md));
c29a70d2 1421
ee122c79
TG		 1422 if (!tun_dst)
1423 goto drop;
1424
0f1b7354 1425 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 1426
1427 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 1428 } else {
1429 memset(md, 0, sizeof(*md));
1430 }
1431
f14ecebb
JB		 1432 if (vs->flags & VXLAN_F_REMCSUM_RX)
1433 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1434 goto drop;
1435 if (vs->flags & VXLAN_F_GBP)
10a5af23 1436 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB		 1437 /* Note that GBP and GPE can never be active together. This is
1438 * ensured in vxlan_dev_configure.
1439 */
3511494c 1440
f14ecebb 1441 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1442 /* If there are any unprocessed flags remaining treat
1443 * this as a malformed packet. This behavior diverges from
1444 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1445 * in reserved fields are to be ignored. The approach here
c4b49512 1446 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1447 * is more robust and provides a little more security in
1448 * adding extensions to VXLAN.
1449 */
288b01c8 1450 goto drop;
3bf39475
TH		 1451 }
1452
e1e5314d 1453 if (!raw_proto) {
3ad7a4b1 1454 if (!vxlan_set_mac(vxlan, vs, skb, vni))
e1e5314d
JB		 1455 goto drop;
1456 } else {
8be0cfa4 1457 skb_reset_mac_header(skb);
e1e5314d
JB		 1458 skb->dev = vxlan->dev;
1459 skb->pkt_type = PACKET_HOST;
1460 }
f2d1968e 1461
f2d1968e
JB		 1462 oiph = skb_network_header(skb);
1463 skb_reset_network_header(skb);
1464
1465 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1466 ++vxlan->dev->stats.rx_frame_errors;
1467 ++vxlan->dev->stats.rx_errors;
1468 goto drop;
1469 }
1470
1471 stats = this_cpu_ptr(vxlan->dev->tstats);
1472 u64_stats_update_begin(&stats->syncp);
1473 stats->rx_packets++;
1474 stats->rx_bytes += skb->len;
1475 u64_stats_update_end(&stats->syncp);
1476
1477 gro_cells_receive(&vxlan->gro_cells, skb);
5cfccc5a
PS		 1478 return 0;
1479
1480drop:
288b01c8
JB		 1481 /* Consume bad packet */
1482 kfree_skb(skb);
1483 return 0;
5cfccc5a
PS		 1484}
1485
3ad7a4b1 1486static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
e4f67add
DS		 1487{
1488 struct vxlan_dev *vxlan = netdev_priv(dev);
1489 struct arphdr *parp;
1490 u8 *arpptr, *sha;
1491 __be32 sip, tip;
1492 struct neighbour *n;
1493
1494 if (dev->flags & IFF_NOARP)
1495 goto out;
1496
1497 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1498 dev->stats.tx_dropped++;
1499 goto out;
1500 }
1501 parp = arp_hdr(skb);
1502
1503 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1504 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1505 parp->ar_pro != htons(ETH_P_IP) ||
1506 parp->ar_op != htons(ARPOP_REQUEST) ||
1507 parp->ar_hln != dev->addr_len ||
1508 parp->ar_pln != 4)
1509 goto out;
1510 arpptr = (u8 *)parp + sizeof(struct arphdr);
1511 sha = arpptr;
1512 arpptr += dev->addr_len; /* sha */
1513 memcpy(&sip, arpptr, sizeof(sip));
1514 arpptr += sizeof(sip);
1515 arpptr += dev->addr_len; /* tha */
1516 memcpy(&tip, arpptr, sizeof(tip));
1517
1518 if (ipv4_is_loopback(tip) ||
1519 ipv4_is_multicast(tip))
1520 goto out;
1521
1522 n = neigh_lookup(&arp_tbl, &tip, dev);
1523
1524 if (n) {
e4f67add
DS		 1525 struct vxlan_fdb *f;
1526 struct sk_buff *reply;
1527
1528 if (!(n->nud_state & NUD_CONNECTED)) {
1529 neigh_release(n);
1530 goto out;
1531 }
1532
3ad7a4b1 1533 f = vxlan_find_mac(vxlan, n->ha, vni);
e4c7ed41 1534 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1535 /* bridge-local neighbor */
1536 neigh_release(n);
1537 goto out;
1538 }
1539
1540 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1541 n->ha, sha);
1542
1543 neigh_release(n);
1544
7346135d
DS		 1545 if (reply == NULL)
1546 goto out;
1547
e4f67add
DS		 1548 skb_reset_mac_header(reply);
1549 __skb_pull(reply, skb_network_offset(reply));
1550 reply->ip_summed = CHECKSUM_UNNECESSARY;
1551 reply->pkt_type = PACKET_HOST;
1552
1553 if (netif_rx_ni(reply) == NET_RX_DROP)
1554 dev->stats.rx_dropped++;
dc5321d7 1555 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
e4c7ed41
CW		 1556 union vxlan_addr ipa = {
1557 .sin.sin_addr.s_addr = tip,
a45e92a5 1558 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1559 };
1560
1561 vxlan_ip_miss(dev, &ipa);
1562 }
e4f67add
DS		 1563out:
1564 consume_skb(skb);
1565 return NETDEV_TX_OK;
1566}
1567
f564f45c 1568#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1569static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1570 struct neighbour *n, bool isrouter)
1571{
1572 struct net_device *dev = request->dev;
1573 struct sk_buff *reply;
1574 struct nd_msg *ns, *na;
1575 struct ipv6hdr *pip6;
1576 u8 *daddr;
1577 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1578 int ns_olen;
1579 int i, len;
1580
f1fb08f6 1581 if (dev == NULL || !pskb_may_pull(request, request->len))
4b29dba9
DS		 1582 return NULL;
1583
1584 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1585 sizeof(*na) + na_olen + dev->needed_tailroom;
1586 reply = alloc_skb(len, GFP_ATOMIC);
1587 if (reply == NULL)
1588 return NULL;
1589
1590 reply->protocol = htons(ETH_P_IPV6);
1591 reply->dev = dev;
1592 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1593 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1594 skb_reset_mac_header(reply);
4b29dba9 1595
f1fb08f6 1596 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
4b29dba9
DS		 1597
1598 daddr = eth_hdr(request)->h_source;
f1fb08f6
VB		 1599 ns_olen = request->len - skb_network_offset(request) -
1600 sizeof(struct ipv6hdr) - sizeof(*ns);
4b29dba9
DS		 1601 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1602 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1603 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1604 break;
1605 }
1606 }
1607
1608 /* Ethernet header */
1609 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1610 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1611 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1612 reply->protocol = htons(ETH_P_IPV6);
1613
1614 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1615 skb_reset_network_header(reply);
4b29dba9
DS		 1616 skb_put(reply, sizeof(struct ipv6hdr));
1617
1618 /* IPv6 header */
1619
1620 pip6 = ipv6_hdr(reply);
1621 memset(pip6, 0, sizeof(struct ipv6hdr));
1622 pip6->version = 6;
1623 pip6->priority = ipv6_hdr(request)->priority;
1624 pip6->nexthdr = IPPROTO_ICMPV6;
1625 pip6->hop_limit = 255;
1626 pip6->daddr = ipv6_hdr(request)->saddr;
1627 pip6->saddr = *(struct in6_addr *)n->primary_key;
1628
1629 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1630 skb_reset_transport_header(reply);
4b29dba9 1631
4b29dba9 1632 /* Neighbor Advertisement */
b080db58 1633 na = skb_put_zero(reply, sizeof(*na) + na_olen);
4b29dba9
DS		 1634 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1635 na->icmph.icmp6_router = isrouter;
1636 na->icmph.icmp6_override = 1;
1637 na->icmph.icmp6_solicited = 1;
1638 na->target = ns->target;
1639 ether_addr_copy(&na->opt[2], n->ha);
1640 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1641 na->opt[1] = na_olen >> 3;
1642
1643 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1644 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1645 csum_partial(na, sizeof(*na)+na_olen, 0));
1646
1647 pip6->payload_len = htons(sizeof(*na)+na_olen);
1648
1649 skb_push(reply, sizeof(struct ipv6hdr));
1650
1651 reply->ip_summed = CHECKSUM_UNNECESSARY;
1652
1653 return reply;
1654}
1655
3ad7a4b1 1656static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
f564f45c
CW		 1657{
1658 struct vxlan_dev *vxlan = netdev_priv(dev);
8dcd81a9 1659 const struct in6_addr *daddr;
8bff3685 1660 const struct ipv6hdr *iphdr;
4b29dba9 1661 struct inet6_dev *in6_dev;
8bff3685
XL		 1662 struct neighbour *n;
1663 struct nd_msg *msg;
f564f45c
CW		 1664
1665 in6_dev = __in6_dev_get(dev);
1666 if (!in6_dev)
1667 goto out;
1668
f564f45c 1669 iphdr = ipv6_hdr(skb);
f564f45c 1670 daddr = &iphdr->daddr;
f1fb08f6 1671 msg = (struct nd_msg *)(iphdr + 1);
f564f45c 1672
4b29dba9
DS		 1673 if (ipv6_addr_loopback(daddr) ||
1674 ipv6_addr_is_multicast(&msg->target))
1675 goto out;
1676
1677 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1678
1679 if (n) {
1680 struct vxlan_fdb *f;
4b29dba9 1681 struct sk_buff *reply;
f564f45c
CW		 1682
1683 if (!(n->nud_state & NUD_CONNECTED)) {
1684 neigh_release(n);
1685 goto out;
1686 }
1687
3ad7a4b1 1688 f = vxlan_find_mac(vxlan, n->ha, vni);
f564f45c
CW		 1689 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1690 /* bridge-local neighbor */
1691 neigh_release(n);
1692 goto out;
1693 }
1694
4b29dba9
DS		 1695 reply = vxlan_na_create(skb, n,
1696 !!(f ? f->flags & NTF_ROUTER : 0));
1697
f564f45c 1698 neigh_release(n);
4b29dba9
DS		 1699
1700 if (reply == NULL)
1701 goto out;
1702
1703 if (netif_rx_ni(reply) == NET_RX_DROP)
1704 dev->stats.rx_dropped++;
1705
dc5321d7 1706 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1707 union vxlan_addr ipa = {
1708 .sin6.sin6_addr = msg->target,
a45e92a5 1709 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1710 };
1711
f564f45c
CW		 1712 vxlan_ip_miss(dev, &ipa);
1713 }
1714
1715out:
1716 consume_skb(skb);
1717 return NETDEV_TX_OK;
1718}
1719#endif
1720
e4f67add
DS		 1721static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1722{
1723 struct vxlan_dev *vxlan = netdev_priv(dev);
1724 struct neighbour *n;
e4f67add
DS		 1725
1726 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1727 return false;
1728
1729 n = NULL;
1730 switch (ntohs(eth_hdr(skb)->h_proto)) {
1731 case ETH_P_IP:
e15a00aa
CW		 1732 {
1733 struct iphdr *pip;
1734
e4f67add
DS		 1735 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1736 return false;
1737 pip = ip_hdr(skb);
1738 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
dc5321d7 1739 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e4c7ed41
CW		 1740 union vxlan_addr ipa = {
1741 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1742 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1743 };
1744
1745 vxlan_ip_miss(dev, &ipa);
1746 return false;
1747 }
1748
e4f67add 1749 break;
e15a00aa
CW		 1750 }
1751#if IS_ENABLED(CONFIG_IPV6)
1752 case ETH_P_IPV6:
1753 {
1754 struct ipv6hdr *pip6;
1755
1756 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1757 return false;
1758 pip6 = ipv6_hdr(skb);
1759 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
dc5321d7 1760 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e15a00aa
CW		 1761 union vxlan_addr ipa = {
1762 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1763 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1764 };
1765
1766 vxlan_ip_miss(dev, &ipa);
1767 return false;
1768 }
1769
1770 break;
1771 }
1772#endif
e4f67add
DS		 1773 default:
1774 return false;
1775 }
1776
1777 if (n) {
1778 bool diff;
1779
7367d0b5 1780 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1781 if (diff) {
1782 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1783 dev->addr_len);
1784 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1785 }
1786 neigh_release(n);
1787 return diff;
e4c7ed41
CW		 1788 }
1789
e4f67add
DS		 1790 return false;
1791}
1792
af33c1ad 1793static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1794 struct vxlan_metadata *md)
1795{
1796 struct vxlanhdr_gbp *gbp;
1797
db79a621
TG		 1798 if (!md->gbp)
1799 return;
1800
3511494c 1801 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1802 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 1803
1804 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1805 gbp->dont_learn = 1;
1806
1807 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1808 gbp->policy_applied = 1;
1809
1810 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1811}
1812
e1e5314d
JB		 1813static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1814 __be16 protocol)
1815{
1816 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1817
1818 gpe->np_applied = 1;
fa20e0e3
JB		 1819 gpe->next_protocol = tun_p_from_eth_p(protocol);
1820 if (!gpe->next_protocol)
1821 return -EPFNOSUPPORT;
1822 return 0;
e1e5314d
JB		 1823}
1824
f491e56d
JB		 1825static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1826 int iphdr_len, __be32 vni,
1827 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1828 bool udp_sum)
e4c7ed41 1829{
e4c7ed41 1830 struct vxlanhdr *vxh;
e4c7ed41
CW		 1831 int min_headroom;
1832 int err;
dfd8645e 1833 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 1834 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 1835
af33c1ad 1836 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1837 skb->ip_summed == CHECKSUM_PARTIAL) {
1838 int csum_start = skb_checksum_start_offset(skb);
1839
1840 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1841 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1842 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1843 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1844 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1845 }
e4c7ed41 1846
e4c7ed41 1847 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
4a4f86cc 1848 + VXLAN_HLEN + iphdr_len;
649c5b8b
PS		 1849
1850 /* Need space for new headers (invalidates iph ptr) */
1851 err = skb_cow_head(skb, min_headroom);
e1e5314d 1852 if (unlikely(err))
c46b7897 1853 return err;
649c5b8b 1854
aed069df
AD		 1855 err = iptunnel_handle_offloads(skb, type);
1856 if (err)
c46b7897 1857 return err;
b736a623 1858
d58ff351 1859 vxh = __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 1860 vxh->vx_flags = VXLAN_HF_VNI;
1861 vxh->vx_vni = vxlan_vni_field(vni);
49560532 1862
dfd8645e 1863 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 1864 unsigned int start;
dfd8645e 1865
54bfd872
JB		 1866 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1867 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1868 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 1869
1870 if (!skb_is_gso(skb)) {
1871 skb->ip_summed = CHECKSUM_NONE;
1872 skb->encapsulation = 0;
1873 }
1874 }
1875
af33c1ad
TH		 1876 if (vxflags & VXLAN_F_GBP)
1877 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB		 1878 if (vxflags & VXLAN_F_GPE) {
1879 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1880 if (err < 0)
c46b7897 1881 return err;
e1e5314d
JB		 1882 inner_protocol = skb->protocol;
1883 }
3511494c 1884
e1e5314d 1885 skb_set_inner_protocol(skb, inner_protocol);
039f5062 1886 return 0;
49560532 1887}
49560532 1888
655c3de1 1889static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
1890 struct vxlan_sock *sock4,
1a8496ba 1891 struct sk_buff *skb, int oif, u8 tos,
4ecb1d83 1892 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
0c1d70af 1893 struct dst_cache *dst_cache,
db3c6139 1894 const struct ip_tunnel_info *info)
1a8496ba 1895{
db3c6139 1896 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 1897 struct rtable *rt = NULL;
1898 struct flowi4 fl4;
1899
655c3de1 1900 if (!sock4)
1901 return ERR_PTR(-EIO);
1902
db3c6139
DB		 1903 if (tos && !info)
1904 use_cache = false;
1905 if (use_cache) {
0c1d70af
PA		 1906 rt = dst_cache_get_ip4(dst_cache, saddr);
1907 if (rt)
1908 return rt;
1909 }
1910
1a8496ba
JB		 1911 memset(&fl4, 0, sizeof(fl4));
1912 fl4.flowi4_oif = oif;
1913 fl4.flowi4_tos = RT_TOS(tos);
1914 fl4.flowi4_mark = skb->mark;
1915 fl4.flowi4_proto = IPPROTO_UDP;
1916 fl4.daddr = daddr;
272d96a5 1917 fl4.saddr = *saddr;
4ecb1d83
MP		 1918 fl4.fl4_dport = dport;
1919 fl4.fl4_sport = sport;
1a8496ba
JB		 1920
1921 rt = ip_route_output_key(vxlan->net, &fl4);
655c3de1 1922 if (likely(!IS_ERR(rt))) {
1923 if (rt->dst.dev == dev) {
1924 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
1925 ip_rt_put(rt);
1926 return ERR_PTR(-ELOOP);
1927 }
1928
1a8496ba 1929 *saddr = fl4.saddr;
0c1d70af
PA		 1930 if (use_cache)
1931 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
655c3de1 1932 } else {
1933 netdev_dbg(dev, "no route to %pI4\n", &daddr);
1934 return ERR_PTR(-ENETUNREACH);
0c1d70af 1935 }
1a8496ba
JB		 1936 return rt;
1937}
1938
e5d4b29f
JB		 1939#if IS_ENABLED(CONFIG_IPV6)
1940static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
655c3de1 1941 struct net_device *dev,
03dc52a8 1942 struct vxlan_sock *sock6,
1400615d 1943 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 1944 __be32 label,
e5d4b29f 1945 const struct in6_addr *daddr,
0c1d70af 1946 struct in6_addr *saddr,
4ecb1d83 1947 __be16 dport, __be16 sport,
db3c6139
DB		 1948 struct dst_cache *dst_cache,
1949 const struct ip_tunnel_info *info)
e5d4b29f 1950{
db3c6139 1951 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 1952 struct dst_entry *ndst;
1953 struct flowi6 fl6;
1954 int err;
1955
c6fcc4fc 1956 if (!sock6)
1957 return ERR_PTR(-EIO);
1958
1400615d
DB		 1959 if (tos && !info)
1960 use_cache = false;
db3c6139 1961 if (use_cache) {
0c1d70af
PA		 1962 ndst = dst_cache_get_ip6(dst_cache, saddr);
1963 if (ndst)
1964 return ndst;
1965 }
1966
e5d4b29f
JB		 1967 memset(&fl6, 0, sizeof(fl6));
1968 fl6.flowi6_oif = oif;
1969 fl6.daddr = *daddr;
272d96a5 1970 fl6.saddr = *saddr;
eaa93bf4 1971 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB		 1972 fl6.flowi6_mark = skb->mark;
1973 fl6.flowi6_proto = IPPROTO_UDP;
4ecb1d83
MP		 1974 fl6.fl6_dport = dport;
1975 fl6.fl6_sport = sport;
e5d4b29f
JB		 1976
1977 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
c6fcc4fc 1978 sock6->sock->sk,
e5d4b29f 1979 &ndst, &fl6);
655c3de1 1980 if (unlikely(err < 0)) {
1981 netdev_dbg(dev, "no route to %pI6\n", daddr);
1982 return ERR_PTR(-ENETUNREACH);
1983 }
1984
1985 if (unlikely(ndst->dev == dev)) {
1986 netdev_dbg(dev, "circular route to %pI6\n", daddr);
1987 dst_release(ndst);
1988 return ERR_PTR(-ELOOP);
1989 }
e5d4b29f
JB		 1990
1991 *saddr = fl6.saddr;
db3c6139 1992 if (use_cache)
0c1d70af 1993 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 1994 return ndst;
1995}
1996#endif
1997
9dcc71e1
SS		 1998/* Bypass encapsulation if the destination is local */
1999static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
3ad7a4b1 2000 struct vxlan_dev *dst_vxlan, __be32 vni)
9dcc71e1 2001{
8f84985f 2002 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 2003 union vxlan_addr loopback;
2004 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 2005 struct net_device *dev = skb->dev;
2006 int len = skb->len;
9dcc71e1 2007
8f84985f
LR		 2008 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2009 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 2010 skb->pkt_type = PACKET_HOST;
2011 skb->encapsulation = 0;
2012 skb->dev = dst_vxlan->dev;
2013 __skb_pull(skb, skb_network_offset(skb));
2014
e4c7ed41
CW		 2015 if (remote_ip->sa.sa_family == AF_INET) {
2016 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2017 loopback.sa.sa_family = AF_INET;
2018#if IS_ENABLED(CONFIG_IPV6)
2019 } else {
2020 loopback.sin6.sin6_addr = in6addr_loopback;
2021 loopback.sa.sa_family = AF_INET6;
2022#endif
2023 }
2024
dc5321d7 2025 if (dst_vxlan->cfg.flags & VXLAN_F_LEARN)
87613de9
MS		 2026 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source, 0,
2027 vni);
9dcc71e1
SS		 2028
2029 u64_stats_update_begin(&tx_stats->syncp);
2030 tx_stats->tx_packets++;
ce6502a8 2031 tx_stats->tx_bytes += len;
9dcc71e1
SS		 2032 u64_stats_update_end(&tx_stats->syncp);
2033
2034 if (netif_rx(skb) == NET_RX_SUCCESS) {
2035 u64_stats_update_begin(&rx_stats->syncp);
2036 rx_stats->rx_packets++;
ce6502a8 2037 rx_stats->rx_bytes += len;
9dcc71e1
SS		 2038 u64_stats_update_end(&rx_stats->syncp);
2039 } else {
ce6502a8 2040 dev->stats.rx_dropped++;
9dcc71e1
SS		 2041 }
2042}
2043
fee1fad7 2044static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
49f810f0
MS		 2045 struct vxlan_dev *vxlan,
2046 union vxlan_addr *daddr,
2047 __be16 dst_port, int dst_ifindex, __be32 vni,
2048 struct dst_entry *dst,
fee1fad7 2049 u32 rt_flags)
2050{
2051#if IS_ENABLED(CONFIG_IPV6)
2052 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2053 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2054 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2055 */
2056 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2057#endif
2058 /* Bypass encapsulation if the destination is local */
2059 if (rt_flags & RTCF_LOCAL &&
2060 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2061 struct vxlan_dev *dst_vxlan;
2062
2063 dst_release(dst);
49f810f0 2064 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
fee1fad7 2065 daddr->sa.sa_family, dst_port,
dc5321d7 2066 vxlan->cfg.flags);
fee1fad7 2067 if (!dst_vxlan) {
2068 dev->stats.tx_errors++;
2069 kfree_skb(skb);
2070
2071 return -ENOENT;
2072 }
3ad7a4b1 2073 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
fee1fad7 2074 return 1;
2075 }
2076
2077 return 0;
2078}
2079
4ad16930 2080static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
3ad7a4b1
RP		 2081 __be32 default_vni, struct vxlan_rdst *rdst,
2082 bool did_rsc)
d342894c 2083{
d71785ff 2084 struct dst_cache *dst_cache;
3093fbe7 2085 struct ip_tunnel_info *info;
d342894c 2086 struct vxlan_dev *vxlan = netdev_priv(dev);
0770b53b 2087 const struct iphdr *old_iph = ip_hdr(skb);
e4c7ed41 2088 union vxlan_addr *dst;
272d96a5 2089 union vxlan_addr remote_ip, local_ip;
ee122c79
TG		 2090 struct vxlan_metadata _md;
2091 struct vxlan_metadata *md = &_md;
e4c7ed41 2092 __be16 src_port = 0, dst_port;
655c3de1 2093 struct dst_entry *ndst = NULL;
e7f70af1 2094 __be32 vni, label;
d342894c 2095 __u8 tos, ttl;
49f810f0 2096 int ifindex;
0e6fbc5b 2097 int err;
dc5321d7 2098 u32 flags = vxlan->cfg.flags;
b4ed5cad 2099 bool udp_sum = false;
f491e56d 2100 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 2101
61adedf3 2102 info = skb_tunnel_info(skb);
3093fbe7 2103
ee122c79 2104 if (rdst) {
0770b53b 2105 dst = &rdst->remote_ip;
2106 if (vxlan_addr_any(dst)) {
2107 if (did_rsc) {
2108 /* short-circuited back to local bridge */
3ad7a4b1 2109 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
0770b53b 2110 return;
2111 }
2112 goto drop;
2113 }
2114
0dfbdf41 2115 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
3ad7a4b1 2116 vni = (rdst->remote_vni) ? : default_vni;
49f810f0 2117 ifindex = rdst->remote_ifindex;
1158632b 2118 local_ip = vxlan->cfg.saddr;
d71785ff 2119 dst_cache = &rdst->dst_cache;
0770b53b 2120 md->gbp = skb->mark;
72f6d71e
HL		 2121 if (flags & VXLAN_F_TTL_INHERIT) {
2122 ttl = ip_tunnel_get_ttl(old_iph, skb);
2123 } else {
2124 ttl = vxlan->cfg.ttl;
2125 if (!ttl && vxlan_addr_multicast(dst))
2126 ttl = 1;
2127 }
0770b53b 2128
2129 tos = vxlan->cfg.tos;
2130 if (tos == 1)
2131 tos = ip_tunnel_get_dsfield(old_iph, skb);
2132
2133 if (dst->sa.sa_family == AF_INET)
2134 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2135 else
2136 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2137 label = vxlan->cfg.label;
ee122c79
TG		 2138 } else {
2139 if (!info) {
2140 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2141 dev->name);
2142 goto drop;
2143 }
b1be00a6 2144 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
272d96a5 2145 if (remote_ip.sa.sa_family == AF_INET) {
a725e514 2146 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
272d96a5 2147 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2148 } else {
a725e514 2149 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
272d96a5 2150 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2151 }
ee122c79 2152 dst = &remote_ip;
0770b53b 2153 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2154 vni = tunnel_id_to_key32(info->key.tun_id);
49f810f0 2155 ifindex = 0;
d71785ff 2156 dst_cache = &info->dst_cache;
0770b53b 2157 if (info->options_len)
2158 md = ip_tunnel_info_opts(info);
a725e514
JB		 2159 ttl = info->key.ttl;
2160 tos = info->key.tos;
e7f70af1 2161 label = info->key.label;
b4ed5cad 2162 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514 2163 }
0770b53b 2164 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2165 vxlan->cfg.port_max, true);
a725e514 2166
56de859e 2167 rcu_read_lock();
e4c7ed41 2168 if (dst->sa.sa_family == AF_INET) {
c6fcc4fc 2169 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
c46b7897 2170 struct rtable *rt;
0770b53b 2171 __be16 df = 0;
c6fcc4fc 2172
49f810f0 2173 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
272d96a5 2174 dst->sin.sin_addr.s_addr,
1158632b 2175 &local_ip.sin.sin_addr.s_addr,
4ecb1d83 2176 dst_port, src_port,
d71785ff 2177 dst_cache, info);
8ebd115b
DM		 2178 if (IS_ERR(rt)) {
2179 err = PTR_ERR(rt);
c46b7897 2180 goto tx_error;
8ebd115b 2181 }
e4c7ed41
CW		 2182
2183 /* Bypass encapsulation if the destination is local */
fee1fad7 2184 if (!info) {
2185 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2186 dst_port, ifindex, vni,
2187 &rt->dst, rt->rt_flags);
fee1fad7 2188 if (err)
56de859e 2189 goto out_unlock;
fee1fad7 2190 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
6ceb31ca 2191 df = htons(IP_DF);
fee1fad7 2192 }
6ceb31ca 2193
c46b7897 2194 ndst = &rt->dst;
a93bf0ff
XL		 2195 if (skb_dst(skb)) {
2196 int mtu = dst_mtu(ndst) - VXLAN_HEADROOM;
2197
f15ca723 2198 skb_dst_update_pmtu(skb, mtu);
a93bf0ff
XL		 2199 }
2200
e4c7ed41
CW		 2201 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2202 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
c46b7897 2203 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
54bfd872 2204 vni, md, flags, udp_sum);
f491e56d 2205 if (err < 0)
c46b7897 2206 goto tx_error;
f491e56d 2207
1158632b 2208 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
f491e56d
JB		 2209 dst->sin.sin_addr.s_addr, tos, ttl, df,
2210 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2211#if IS_ENABLED(CONFIG_IPV6)
2212 } else {
c6fcc4fc 2213 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
e4c7ed41 2214
49f810f0 2215 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
272d96a5 2216 label, &dst->sin6.sin6_addr,
1158632b 2217 &local_ip.sin6.sin6_addr,
4ecb1d83 2218 dst_port, src_port,
db3c6139 2219 dst_cache, info);
e5d4b29f 2220 if (IS_ERR(ndst)) {
8ebd115b 2221 err = PTR_ERR(ndst);
c46b7897 2222 ndst = NULL;
9dcc71e1 2223 goto tx_error;
e4c7ed41 2224 }
655c3de1 2225
fee1fad7 2226 if (!info) {
2227 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
49560532 2228
fee1fad7 2229 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2230 dst_port, ifindex, vni,
2231 ndst, rt6i_flags);
fee1fad7 2232 if (err)
56de859e 2233 goto out_unlock;
fee1fad7 2234 }
35e2d115 2235
a93bf0ff
XL		 2236 if (skb_dst(skb)) {
2237 int mtu = dst_mtu(ndst) - VXLAN6_HEADROOM;
2238
f15ca723 2239 skb_dst_update_pmtu(skb, mtu);
a93bf0ff
XL		 2240 }
2241
1400615d 2242 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2243 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2244 skb_scrub_packet(skb, xnet);
2245 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2246 vni, md, flags, udp_sum);
c46b7897 2247 if (err < 0)
2248 goto tx_error;
2249
0770b53b 2250 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
1158632b 2251 &local_ip.sin6.sin6_addr,
272d96a5 2252 &dst->sin6.sin6_addr, tos, ttl,
e7f70af1 2253 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2254#endif
2255 }
56de859e
JK		 2256out_unlock:
2257 rcu_read_unlock();
4ad16930 2258 return;
d342894c 2259
2260drop:
2261 dev->stats.tx_dropped++;
c46b7897 2262 dev_kfree_skb(skb);
2263 return;
d342894c 2264
2265tx_error:
56de859e 2266 rcu_read_unlock();
655c3de1 2267 if (err == -ELOOP)
2268 dev->stats.collisions++;
2269 else if (err == -ENETUNREACH)
2270 dev->stats.tx_carrier_errors++;
c46b7897 2271 dst_release(ndst);
d342894c 2272 dev->stats.tx_errors++;
c46b7897 2273 kfree_skb(skb);
d342894c 2274}
2275
6681712d
DS		 2276/* Transmit local packets over Vxlan
2277 *
2278 * Outer IP header inherits ECN and DF from inner header.
2279 * Outer UDP destination is the VXLAN assigned port.
2280 * source port is based on hash of flow
2281 */
2282static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2283{
2284 struct vxlan_dev *vxlan = netdev_priv(dev);
8bff3685 2285 struct vxlan_rdst *rdst, *fdst = NULL;
3093fbe7 2286 const struct ip_tunnel_info *info;
6681712d 2287 bool did_rsc = false;
6681712d 2288 struct vxlan_fdb *f;
8bff3685 2289 struct ethhdr *eth;
3ad7a4b1 2290 __be32 vni = 0;
6681712d 2291
61adedf3 2292 info = skb_tunnel_info(skb);
3093fbe7 2293
6681712d 2294 skb_reset_mac_header(skb);
6681712d 2295
dc5321d7 2296 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 2297 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2298 info->mode & IP_TUNNEL_INFO_TX) {
2299 vni = tunnel_id_to_key32(info->key.tun_id);
2300 } else {
2301 if (info && info->mode & IP_TUNNEL_INFO_TX)
2302 vxlan_xmit_one(skb, dev, vni, NULL, false);
2303 else
2304 kfree_skb(skb);
2305 return NETDEV_TX_OK;
2306 }
47e5d1b0
JB		 2307 }
2308
dc5321d7 2309 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
47e5d1b0 2310 eth = eth_hdr(skb);
f564f45c 2311 if (ntohs(eth->h_proto) == ETH_P_ARP)
3ad7a4b1 2312 return arp_reduce(dev, skb, vni);
f564f45c 2313#if IS_ENABLED(CONFIG_IPV6)
8bff3685
XL		 2314 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2315 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2316 sizeof(struct nd_msg)) &&
2317 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2318 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2319
2320 if (m->icmph.icmp6_code == 0 &&
2321 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
f1fb08f6 2322 return neigh_reduce(dev, skb, vni);
f564f45c
CW		 2323 }
2324#endif
2325 }
6681712d 2326
47e5d1b0 2327 eth = eth_hdr(skb);
3ad7a4b1 2328 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2329 did_rsc = false;
2330
dc5321d7 2331 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2332 (ntohs(eth->h_proto) == ETH_P_IP ||
2333 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2334 did_rsc = route_shortcircuit(dev, skb);
2335 if (did_rsc)
3ad7a4b1 2336 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2337 }
2338
6681712d 2339 if (f == NULL) {
3ad7a4b1 2340 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 2341 if (f == NULL) {
dc5321d7 2342 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
afbd8bae
MR		 2343 !is_multicast_ether_addr(eth->h_dest))
2344 vxlan_fdb_miss(vxlan, eth->h_dest);
2345
2346 dev->stats.tx_dropped++;
8f646c92 2347 kfree_skb(skb);
afbd8bae
MR		 2348 return NETDEV_TX_OK;
2349 }
2350 }
6681712d 2351
afbd8bae
MR		 2352 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2353 struct sk_buff *skb1;
6681712d 2354
8f646c92
ED		 2355 if (!fdst) {
2356 fdst = rdst;
2357 continue;
2358 }
afbd8bae
MR		 2359 skb1 = skb_clone(skb, GFP_ATOMIC);
2360 if (skb1)
3ad7a4b1 2361 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
6681712d
DS		 2362 }
2363
8f646c92 2364 if (fdst)
3ad7a4b1 2365 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
8f646c92
ED		 2366 else
2367 kfree_skb(skb);
4ad16930 2368 return NETDEV_TX_OK;
6681712d
DS		 2369}
2370
d342894c 2371/* Walk the forwarding table and purge stale entries */
df7e828c 2372static void vxlan_cleanup(struct timer_list *t)
d342894c 2373{
df7e828c 2374 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
d342894c 2375 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2376 unsigned int h;
2377
2378 if (!netif_running(vxlan->dev))
2379 return;
2380
d342894c 2381 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2382 struct hlist_node *p, *n;
14e1d0fa
SD		 2383
2384 spin_lock_bh(&vxlan->hash_lock);
d342894c 2385 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2386 struct vxlan_fdb *f
2387 = container_of(p, struct vxlan_fdb, hlist);
2388 unsigned long timeout;
2389
efb5f68f 2390 if (f->state & (NUD_PERMANENT | NUD_NOARP))
d342894c 2391 continue;
2392
def499c9
RP		 2393 if (f->flags & NTF_EXT_LEARNED)
2394 continue;
2395
0dfbdf41 2396 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2397 if (time_before_eq(timeout, jiffies)) {
2398 netdev_dbg(vxlan->dev,
2399 "garbage collect %pM\n",
2400 f->eth_addr);
2401 f->state = NUD_STALE;
f6e05385 2402 vxlan_fdb_destroy(vxlan, f, true);
d342894c 2403 } else if (time_before(timeout, next_timer))
2404 next_timer = timeout;
2405 }
14e1d0fa 2406 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2407 }
d342894c 2408
2409 mod_timer(&vxlan->age_timer, next_timer);
2410}
2411
a53cb29b
MB		 2412static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2413{
2414 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2415
2416 spin_lock(&vn->sock_lock);
69e76661
JB		 2417 hlist_del_init_rcu(&vxlan->hlist4.hlist);
2418#if IS_ENABLED(CONFIG_IPV6)
2419 hlist_del_init_rcu(&vxlan->hlist6.hlist);
2420#endif
a53cb29b
MB		 2421 spin_unlock(&vn->sock_lock);
2422}
2423
69e76661
JB		 2424static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2425 struct vxlan_dev_node *node)
9c2e24e1 2426{
56ef9c90 2427 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2428 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2429
69e76661 2430 node->vxlan = vxlan;
56ef9c90 2431 spin_lock(&vn->sock_lock);
69e76661 2432 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
56ef9c90 2433 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2434}
2435
d342894c 2436/* Setup stats when device is created */
2437static int vxlan_init(struct net_device *dev)
2438{
1c213bd2 2439 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2440 if (!dev->tstats)
d342894c 2441 return -ENOMEM;
2442
2443 return 0;
2444}
2445
3ad7a4b1 2446static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
afbd8bae
MR		 2447{
2448 struct vxlan_fdb *f;
2449
2450 spin_lock_bh(&vxlan->hash_lock);
3ad7a4b1 2451 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 2452 if (f)
f6e05385 2453 vxlan_fdb_destroy(vxlan, f, true);
afbd8bae
MR		 2454 spin_unlock_bh(&vxlan->hash_lock);
2455}
2456
ebf4063e
SH		 2457static void vxlan_uninit(struct net_device *dev)
2458{
2459 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2460
3ad7a4b1 2461 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
afbd8bae 2462
ebf4063e
SH		 2463 free_percpu(dev->tstats);
2464}
2465
d342894c 2466/* Start ageing timer and join group when device is brought up */
2467static int vxlan_open(struct net_device *dev)
2468{
2469 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2470 int ret;
56ef9c90 2471
205f356d
JB		 2472 ret = vxlan_sock_add(vxlan);
2473 if (ret < 0)
2474 return ret;
d342894c 2475
79d4a94f 2476 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2477 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2478 if (ret == -EADDRINUSE)
2479 ret = 0;
56ef9c90 2480 if (ret) {
205f356d 2481 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2482 return ret;
2483 }
d342894c 2484 }
2485
0dfbdf41 2486 if (vxlan->cfg.age_interval)
d342894c 2487 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2488
56ef9c90 2489 return ret;
d342894c 2490}
2491
2492/* Purge the forwarding table */
8b3f9337 2493static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
d342894c 2494{
31fec5aa 2495 unsigned int h;
d342894c 2496
2497 spin_lock_bh(&vxlan->hash_lock);
2498 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2499 struct hlist_node *p, *n;
2500 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2501 struct vxlan_fdb *f
2502 = container_of(p, struct vxlan_fdb, hlist);
8b3f9337
RP		 2503 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2504 continue;
afbd8bae
MR		 2505 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2506 if (!is_zero_ether_addr(f->eth_addr))
f6e05385 2507 vxlan_fdb_destroy(vxlan, f, true);
d342894c 2508 }
2509 }
2510 spin_unlock_bh(&vxlan->hash_lock);
2511}
2512
2513/* Cleanup timer and forwarding table on shutdown */
2514static int vxlan_stop(struct net_device *dev)
2515{
2516 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2517 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2518 int ret = 0;
d342894c 2519
24c0e683 2520 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2521 !vxlan_group_used(vn, vxlan))
56ef9c90 2522 ret = vxlan_igmp_leave(vxlan);
d342894c 2523
2524 del_timer_sync(&vxlan->age_timer);
2525
8b3f9337 2526 vxlan_flush(vxlan, false);
205f356d 2527 vxlan_sock_release(vxlan);
d342894c 2528
56ef9c90 2529 return ret;
d342894c 2530}
2531
d342894c 2532/* Stub, nothing needs to be done. */
2533static void vxlan_set_multicast_list(struct net_device *dev)
2534{
2535}
2536
91572088 2537static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
345010b5 2538{
91572088
JW		 2539 struct vxlan_dev *vxlan = netdev_priv(dev);
2540 struct vxlan_rdst *dst = &vxlan->default_dst;
2541 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2542 dst->remote_ifindex);
ce44a4ae 2543 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
345010b5 2544
91572088
JW		 2545 /* This check is different than dev->max_mtu, because it looks at
2546 * the lowerdev->mtu, rather than the static dev->max_mtu
2547 */
2548 if (lowerdev) {
2549 int max_mtu = lowerdev->mtu -
2550 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2551 if (new_mtu > max_mtu)
72564b59 2552 return -EINVAL;
72564b59
DW		 2553 }
2554
345010b5
DB		 2555 dev->mtu = new_mtu;
2556 return 0;
2557}
2558
fc4099f1
PS		 2559static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2560{
2561 struct vxlan_dev *vxlan = netdev_priv(dev);
2562 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2563 __be16 sport, dport;
2564
2565 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2566 vxlan->cfg.port_max, true);
2567 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2568
239e944f 2569 if (ip_tunnel_info_af(info) == AF_INET) {
c6fcc4fc 2570 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
1a8496ba
JB		 2571 struct rtable *rt;
2572
655c3de1 2573 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
1a8496ba 2574 info->key.u.ipv4.dst,
22f0708a
PA		 2575 &info->key.u.ipv4.src, dport, sport,
2576 &info->dst_cache, info);
1a8496ba
JB		 2577 if (IS_ERR(rt))
2578 return PTR_ERR(rt);
2579 ip_rt_put(rt);
239e944f
JB		 2580 } else {
2581#if IS_ENABLED(CONFIG_IPV6)
03dc52a8 2582 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
239e944f
JB		 2583 struct dst_entry *ndst;
2584
655c3de1 2585 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
e7f70af1 2586 info->key.label, &info->key.u.ipv6.dst,
22f0708a
PA		 2587 &info->key.u.ipv6.src, dport, sport,
2588 &info->dst_cache, info);
239e944f
JB		 2589 if (IS_ERR(ndst))
2590 return PTR_ERR(ndst);
2591 dst_release(ndst);
239e944f
JB		 2592#else /* !CONFIG_IPV6 */
2593 return -EPFNOSUPPORT;
2594#endif
2595 }
1a8496ba
JB		 2596 info->key.tp_src = sport;
2597 info->key.tp_dst = dport;
239e944f 2598 return 0;
fc4099f1
PS		 2599}
2600
0c867c9b 2601static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 2602 .ndo_init = vxlan_init,
ebf4063e 2603 .ndo_uninit = vxlan_uninit,
d342894c 2604 .ndo_open = vxlan_open,
2605 .ndo_stop = vxlan_stop,
2606 .ndo_start_xmit = vxlan_xmit,
e8171045 2607 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2608 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2609 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2610 .ndo_validate_addr = eth_validate_addr,
2611 .ndo_set_mac_address = eth_mac_addr,
2612 .ndo_fdb_add = vxlan_fdb_add,
2613 .ndo_fdb_del = vxlan_fdb_delete,
2614 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2615 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2616};
2617
e1e5314d
JB		 2618static const struct net_device_ops vxlan_netdev_raw_ops = {
2619 .ndo_init = vxlan_init,
2620 .ndo_uninit = vxlan_uninit,
2621 .ndo_open = vxlan_open,
2622 .ndo_stop = vxlan_stop,
2623 .ndo_start_xmit = vxlan_xmit,
2624 .ndo_get_stats64 = ip_tunnel_get_stats64,
2625 .ndo_change_mtu = vxlan_change_mtu,
2626 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2627};
2628
d342894c 2629/* Info for udev, that this is a virtual tunnel endpoint */
2630static struct device_type vxlan_type = {
2631 .name = "vxlan",
2632};
2633
e5de25dc 2634/* Calls the ndo_udp_tunnel_add of the caller in order to
35e42379 2635 * supply the listening VXLAN udp ports. Callers are expected
e5de25dc 2636 * to implement the ndo_udp_tunnel_add.
53cf5275 2637 */
2d2b13fc 2638static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
53cf5275
JG		 2639{
2640 struct vxlan_sock *vs;
2641 struct net *net = dev_net(dev);
2642 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
35e42379 2643 unsigned int i;
53cf5275
JG		 2644
2645 spin_lock(&vn->sock_lock);
2646 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2d2b13fc
SD		 2647 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2648 unsigned short type;
2649
2650 if (vs->flags & VXLAN_F_GPE)
2651 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
2652 else
2653 type = UDP_TUNNEL_TYPE_VXLAN;
2654
2655 if (push)
2656 udp_tunnel_push_rx_port(dev, vs->sock, type);
2657 else
2658 udp_tunnel_drop_rx_port(dev, vs->sock, type);
2659 }
53cf5275
JG		 2660 }
2661 spin_unlock(&vn->sock_lock);
2662}
53cf5275 2663
d342894c 2664/* Initialize the device structure. */
2665static void vxlan_setup(struct net_device *dev)
2666{
2667 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2668 unsigned int h;
d342894c 2669
65226ef8
JB		 2670 eth_hw_addr_random(dev);
2671 ether_setup(dev);
2672
cf124db5 2673 dev->needs_free_netdev = true;
d342894c 2674 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2675
d342894c 2676 dev->features |= NETIF_F_LLTX;
d6727fe3 2677 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2678 dev->features |= NETIF_F_RXCSUM;
05c0db08 2679 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2680
1eaa8178 2681 dev->vlan_features = dev->features;
0afb1666 2682 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2683 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
02875878 2684 netif_keep_dst(dev);
0c867c9b 2685 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 2686
a985343b
MS		 2687 /* MTU range: 68 - 65535 */
2688 dev->min_mtu = ETH_MIN_MTU;
2689 dev->max_mtu = ETH_MAX_MTU;
2690
553675fb 2691 INIT_LIST_HEAD(&vxlan->next);
d342894c 2692 spin_lock_init(&vxlan->hash_lock);
2693
df7e828c 2694 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
d342894c 2695
2696 vxlan->dev = dev;
2697
58ce31cc
TH		 2698 gro_cells_init(&vxlan->gro_cells, dev);
2699
d342894c 2700 for (h = 0; h < FDB_HASH_SIZE; ++h)
2701 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2702}
2703
0c867c9b
JB		 2704static void vxlan_ether_setup(struct net_device *dev)
2705{
0c867c9b
JB		 2706 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2707 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2708 dev->netdev_ops = &vxlan_netdev_ether_ops;
2709}
2710
e1e5314d
JB		 2711static void vxlan_raw_setup(struct net_device *dev)
2712{
65226ef8 2713 dev->header_ops = NULL;
e1e5314d
JB		 2714 dev->type = ARPHRD_NONE;
2715 dev->hard_header_len = 0;
2716 dev->addr_len = 0;
e1e5314d
JB		 2717 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2718 dev->netdev_ops = &vxlan_netdev_raw_ops;
2719}
2720
d342894c 2721static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2722 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2723 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2724 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2725 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2726 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2727 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2728 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2729 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 2730 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 2731 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2732 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2733 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2734 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2735 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2736 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2737 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2738 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2739 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2740 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2741 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2742 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2743 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2744 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2745 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2746 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 2747 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 2748 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
72f6d71e 2749 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
d342894c 2750};
2751
a8b8a889
MS		 2752static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
2753 struct netlink_ext_ack *extack)
d342894c 2754{
2755 if (tb[IFLA_ADDRESS]) {
2756 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
653ef6a3
GM		 2757 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2758 "Provided link layer address is not Ethernet");
d342894c 2759 return -EINVAL;
2760 }
2761
2762 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
653ef6a3
GM		 2763 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2764 "Provided Ethernet address is not unicast");
d342894c 2765 return -EADDRNOTAVAIL;
2766 }
2767 }
2768
a985343b 2769 if (tb[IFLA_MTU]) {
019b13ae 2770 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
a985343b 2771
653ef6a3
GM		 2772 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
2773 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
2774 "MTU must be between 68 and 65535");
a985343b 2775 return -EINVAL;
653ef6a3 2776 }
a985343b
MS		 2777 }
2778
653ef6a3
GM		 2779 if (!data) {
2780 NL_SET_ERR_MSG(extack,
2781 "Required attributes not provided to perform the operation");
d342894c 2782 return -EINVAL;
653ef6a3 2783 }
d342894c 2784
2785 if (data[IFLA_VXLAN_ID]) {
a985343b
MS		 2786 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2787
653ef6a3
GM		 2788 if (id >= VXLAN_N_VID) {
2789 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID],
2790 "VXLAN ID must be lower than 16777216");
d342894c 2791 return -ERANGE;
653ef6a3 2792 }
d342894c 2793 }
2794
05f47d69 2795 if (data[IFLA_VXLAN_PORT_RANGE]) {
2796 const struct ifla_vxlan_port_range *p
2797 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2798
2799 if (ntohs(p->high) < ntohs(p->low)) {
653ef6a3
GM		 2800 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
2801 "Invalid source port range");
05f47d69 2802 return -EINVAL;
2803 }
2804 }
2805
d342894c 2806 return 0;
2807}
2808
1b13c97f
YB		 2809static void vxlan_get_drvinfo(struct net_device *netdev,
2810 struct ethtool_drvinfo *drvinfo)
2811{
2812 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2813 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2814}
2815
2816static const struct ethtool_ops vxlan_ethtool_ops = {
2817 .get_drvinfo = vxlan_get_drvinfo,
2818 .get_link = ethtool_op_get_link,
2819};
2820
3ee64f39
TH		 2821static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2822 __be16 port, u32 flags)
553675fb 2823{
e4c7ed41 2824 struct socket *sock;
3ee64f39
TH		 2825 struct udp_port_cfg udp_conf;
2826 int err;
e4c7ed41 2827
3ee64f39 2828 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2829
3ee64f39
TH		 2830 if (ipv6) {
2831 udp_conf.family = AF_INET6;
3ee64f39 2832 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2833 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2834 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2835 } else {
2836 udp_conf.family = AF_INET;
e4c7ed41 2837 }
e4c7ed41 2838
3ee64f39 2839 udp_conf.local_udp_port = port;
553675fb 2840
3ee64f39
TH		 2841 /* Open UDP socket */
2842 err = udp_sock_create(net, &udp_conf, &sock);
2843 if (err < 0)
2844 return ERR_PTR(err);
e4c7ed41 2845
39deb2c7 2846 return sock;
e4c7ed41
CW		 2847}
2848
2849/* Create new listen socket if needed */
b1be00a6
JB		 2850static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2851 __be16 port, u32 flags)
e4c7ed41
CW		 2852{
2853 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2854 struct vxlan_sock *vs;
2855 struct socket *sock;
e4c7ed41 2856 unsigned int h;
acbf74a7 2857 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2858
dc01e7d3 2859 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2860 if (!vs)
2861 return ERR_PTR(-ENOMEM);
2862
2863 for (h = 0; h < VNI_HASH_SIZE; ++h)
2864 INIT_HLIST_HEAD(&vs->vni_list[h]);
2865
3ee64f39 2866 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2867 if (IS_ERR(sock)) {
553675fb 2868 kfree(vs);
e50fddc8 2869 return ERR_CAST(sock);
553675fb 2870 }
e4c7ed41
CW		 2871
2872 vs->sock = sock;
66af846f 2873 refcount_set(&vs->refcnt, 1);
af33c1ad 2874 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2875
9c2e24e1
PS		 2876 spin_lock(&vn->sock_lock);
2877 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
e7b3db5e 2878 udp_tunnel_notify_add_rx_port(sock,
b9adcd69
AD		 2879 (vs->flags & VXLAN_F_GPE) ?
2880 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 2881 UDP_TUNNEL_TYPE_VXLAN);
9c2e24e1 2882 spin_unlock(&vn->sock_lock);
553675fb 2883
2884 /* Mark socket as an encapsulation socket. */
5602c48c 2885 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ		 2886 tunnel_cfg.sk_user_data = vs;
2887 tunnel_cfg.encap_type = 1;
f2d1968e 2888 tunnel_cfg.encap_rcv = vxlan_rcv;
acbf74a7 2889 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH		 2890 tunnel_cfg.gro_receive = vxlan_gro_receive;
2891 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ		 2892
2893 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2894
9c2e24e1
PS		 2895 return vs;
2896}
2897
b1be00a6 2898static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2899{
205f356d
JB		 2900 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2901 struct vxlan_sock *vs = NULL;
69e76661 2902 struct vxlan_dev_node *node;
9c2e24e1 2903
205f356d 2904 if (!vxlan->cfg.no_share) {
56ef9c90 2905 spin_lock(&vn->sock_lock);
205f356d 2906 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
dc5321d7 2907 vxlan->cfg.dst_port, vxlan->cfg.flags);
66af846f 2908 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
56ef9c90 2909 spin_unlock(&vn->sock_lock);
205f356d 2910 return -EBUSY;
56ef9c90
MRL		 2911 }
2912 spin_unlock(&vn->sock_lock);
2913 }
205f356d 2914 if (!vs)
b1be00a6 2915 vs = vxlan_socket_create(vxlan->net, ipv6,
dc5321d7 2916 vxlan->cfg.dst_port, vxlan->cfg.flags);
205f356d
JB		 2917 if (IS_ERR(vs))
2918 return PTR_ERR(vs);
b1be00a6 2919#if IS_ENABLED(CONFIG_IPV6)
69e76661 2920 if (ipv6) {
c6fcc4fc 2921 rcu_assign_pointer(vxlan->vn6_sock, vs);
69e76661
JB		 2922 node = &vxlan->hlist6;
2923 } else
b1be00a6 2924#endif
69e76661 2925 {
c6fcc4fc 2926 rcu_assign_pointer(vxlan->vn4_sock, vs);
69e76661
JB		 2927 node = &vxlan->hlist4;
2928 }
2929 vxlan_vs_add_dev(vs, vxlan, node);
205f356d 2930 return 0;
553675fb 2931}
2932
b1be00a6
JB		 2933static int vxlan_sock_add(struct vxlan_dev *vxlan)
2934{
dc5321d7
MS		 2935 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
2936 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
d074bf96 2937 bool ipv4 = !ipv6 || metadata;
b1be00a6
JB		 2938 int ret = 0;
2939
c6fcc4fc 2940 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
b1be00a6 2941#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 2942 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
d074bf96 2943 if (ipv6) {
b1be00a6 2944 ret = __vxlan_sock_add(vxlan, true);
d074bf96
JB		 2945 if (ret < 0 && ret != -EAFNOSUPPORT)
2946 ipv4 = false;
2947 }
b1be00a6 2948#endif
d074bf96 2949 if (ipv4)
b1be00a6
JB		 2950 ret = __vxlan_sock_add(vxlan, false);
2951 if (ret < 0)
2952 vxlan_sock_release(vxlan);
2953 return ret;
2954}
2955
a985343b
MS		 2956static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
2957 struct net_device **lower,
653ef6a3
GM		 2958 struct vxlan_dev *old,
2959 struct netlink_ext_ack *extack)
d342894c 2960{
33564bbb 2961 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
a985343b 2962 struct vxlan_dev *tmp;
e4c7ed41 2963 bool use_ipv6 = false;
d342894c 2964
a985343b
MS		 2965 if (conf->flags & VXLAN_F_GPE) {
2966 /* For now, allow GPE only together with
2967 * COLLECT_METADATA. This can be relaxed later; in such
2968 * case, the other side of the PtP link will have to be
2969 * provided.
2970 */
2971 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
2972 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
653ef6a3
GM		 2973 NL_SET_ERR_MSG(extack,
2974 "VXLAN GPE does not support this combination of attributes");
a985343b 2975 return -EINVAL;
3555621d 2976 }
e1e5314d 2977 }
0c867c9b 2978
ce44a4ae
MS		 2979 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
2980 /* Unless IPv6 is explicitly requested, assume IPv4 */
a985343b 2981 conf->remote_ip.sa.sa_family = AF_INET;
ce44a4ae
MS		 2982 conf->saddr.sa.sa_family = AF_INET;
2983 } else if (!conf->remote_ip.sa.sa_family) {
2984 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
2985 } else if (!conf->saddr.sa.sa_family) {
2986 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
2987 }
2988
653ef6a3
GM		 2989 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
2990 NL_SET_ERR_MSG(extack,
2991 "Local and remote address must be from the same family");
ce44a4ae 2992 return -EINVAL;
653ef6a3 2993 }
e4c7ed41 2994
653ef6a3
GM		 2995 if (vxlan_addr_multicast(&conf->saddr)) {
2996 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
0f22a3c6 2997 return -EINVAL;
653ef6a3 2998 }
0f22a3c6 2999
ce44a4ae 3000 if (conf->saddr.sa.sa_family == AF_INET6) {
653ef6a3
GM		 3001 if (!IS_ENABLED(CONFIG_IPV6)) {
3002 NL_SET_ERR_MSG(extack,
3003 "IPv6 support not enabled in the kernel");
057ba29b 3004 return -EPFNOSUPPORT;
653ef6a3 3005 }
e4c7ed41 3006 use_ipv6 = true;
a985343b 3007 conf->flags |= VXLAN_F_IPV6;
0f22a3c6
MS		 3008
3009 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3010 int local_type =
3011 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3012 int remote_type =
3013 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3014
3015 if (local_type & IPV6_ADDR_LINKLOCAL) {
3016 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
653ef6a3
GM		 3017 (remote_type != IPV6_ADDR_ANY)) {
3018 NL_SET_ERR_MSG(extack,
3019 "Invalid combination of local and remote address scopes");
0f22a3c6 3020 return -EINVAL;
653ef6a3 3021 }
0f22a3c6
MS		 3022
3023 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3024 } else {
3025 if (remote_type ==
653ef6a3
GM		 3026 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3027 NL_SET_ERR_MSG(extack,
3028 "Invalid combination of local and remote address scopes");
0f22a3c6 3029 return -EINVAL;
653ef6a3 3030 }
0f22a3c6
MS		 3031
3032 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3033 }
3034 }
057ba29b 3035 }
d342894c 3036
653ef6a3
GM		 3037 if (conf->label && !use_ipv6) {
3038 NL_SET_ERR_MSG(extack,
3039 "Label attribute only applies to IPv6 VXLAN devices");
e7f70af1 3040 return -EINVAL;
653ef6a3 3041 }
e7f70af1 3042
a985343b
MS		 3043 if (conf->remote_ifindex) {
3044 struct net_device *lowerdev;
34e02aa1 3045
a985343b 3046 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
653ef6a3
GM		 3047 if (!lowerdev) {
3048 NL_SET_ERR_MSG(extack,
3049 "Invalid local interface, device not found");
34e02aa1 3050 return -ENODEV;
653ef6a3 3051 }
d342894c 3052
e4c7ed41
CW		 3053#if IS_ENABLED(CONFIG_IPV6)
3054 if (use_ipv6) {
3055 struct inet6_dev *idev = __in6_dev_get(lowerdev);
653ef6a3
GM		 3056 if (idev && idev->cnf.disable_ipv6) {
3057 NL_SET_ERR_MSG(extack,
3058 "IPv6 support disabled by administrator");
e4c7ed41 3059 return -EPERM;
653ef6a3 3060 }
e4c7ed41
CW		 3061 }
3062#endif
3063
a985343b
MS		 3064 *lower = lowerdev;
3065 } else {
653ef6a3
GM		 3066 if (vxlan_addr_multicast(&conf->remote_ip)) {
3067 NL_SET_ERR_MSG(extack,
3068 "Local interface required for multicast remote destination");
3069
a985343b 3070 return -EINVAL;
653ef6a3 3071 }
1ba56fb4 3072
0f22a3c6 3073#if IS_ENABLED(CONFIG_IPV6)
653ef6a3
GM		 3074 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3075 NL_SET_ERR_MSG(extack,
3076 "Local interface required for link-local local/remote addresses");
0f22a3c6 3077 return -EINVAL;
653ef6a3 3078 }
0f22a3c6
MS		 3079#endif
3080
a985343b 3081 *lower = NULL;
9dc2ad10 3082 }
d342894c 3083
a985343b
MS		 3084 if (!conf->dst_port) {
3085 if (conf->flags & VXLAN_F_GPE)
3086 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3087 else
3088 conf->dst_port = htons(vxlan_port);
d6acfeb1
FM		 3089 }
3090
a985343b
MS		 3091 if (!conf->age_interval)
3092 conf->age_interval = FDB_AGE_DEFAULT;
91572088 3093
a985343b
MS		 3094 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3095 if (tmp == old)
3096 continue;
91572088 3097
49f810f0
MS		 3098 if (tmp->cfg.vni != conf->vni)
3099 continue;
3100 if (tmp->cfg.dst_port != conf->dst_port)
3101 continue;
3102 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
ce44a4ae 3103 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
49f810f0
MS		 3104 continue;
3105
3106 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3107 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3108 continue;
3109
653ef6a3
GM		 3110 NL_SET_ERR_MSG(extack,
3111 "A VXLAN device with the specified VNI already exists");
49f810f0 3112 return -EEXIST;
a985343b 3113 }
91572088 3114
a985343b
MS		 3115 return 0;
3116}
3117
3118static void vxlan_config_apply(struct net_device *dev,
3119 struct vxlan_config *conf,
889ce937
SD		 3120 struct net_device *lowerdev,
3121 struct net *src_net,
3122 bool changelink)
a985343b
MS		 3123{
3124 struct vxlan_dev *vxlan = netdev_priv(dev);
3125 struct vxlan_rdst *dst = &vxlan->default_dst;
3126 unsigned short needed_headroom = ETH_HLEN;
3127 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3128 int max_mtu = ETH_MAX_MTU;
3129
3130 if (!changelink) {
3131 if (conf->flags & VXLAN_F_GPE)
3132 vxlan_raw_setup(dev);
3133 else
3134 vxlan_ether_setup(dev);
3135
3136 if (conf->mtu)
3137 dev->mtu = conf->mtu;
889ce937
SD		 3138
3139 vxlan->net = src_net;
a985343b
MS		 3140 }
3141
3142 dst->remote_vni = conf->vni;
3143
3144 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3145
3146 if (lowerdev) {
3147 dst->remote_ifindex = conf->remote_ifindex;
3148
3149 dev->gso_max_size = lowerdev->gso_max_size;
3150 dev->gso_max_segs = lowerdev->gso_max_segs;
91572088 3151
a985343b 3152 needed_headroom = lowerdev->hard_header_len;
91572088 3153
a985343b
MS		 3154 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3155 VXLAN_HEADROOM);
f870c1ff
AK		 3156 if (max_mtu < ETH_MIN_MTU)
3157 max_mtu = ETH_MIN_MTU;
3158
3159 if (!changelink && !conf->mtu)
3160 dev->mtu = max_mtu;
7e059158
DW		 3161 }
3162
a985343b
MS		 3163 if (dev->mtu > max_mtu)
3164 dev->mtu = max_mtu;
3165
b1be00a6
JB		 3166 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3167 needed_headroom += VXLAN6_HEADROOM;
3168 else
3169 needed_headroom += VXLAN_HEADROOM;
3170 dev->needed_headroom = needed_headroom;
3171
0dfbdf41 3172 memcpy(&vxlan->cfg, conf, sizeof(*conf));
a985343b 3173}
0dfbdf41 3174
a985343b 3175static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
653ef6a3
GM		 3176 struct vxlan_config *conf, bool changelink,
3177 struct netlink_ext_ack *extack)
a985343b
MS		 3178{
3179 struct vxlan_dev *vxlan = netdev_priv(dev);
3180 struct net_device *lowerdev;
3181 int ret;
0dfbdf41 3182
653ef6a3 3183 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
a985343b
MS		 3184 if (ret)
3185 return ret;
8bcdc4f3 3186
889ce937 3187 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
0dfbdf41 3188
0dfbdf41
TG		 3189 return 0;
3190}
3191
c80498e3 3192static int __vxlan_dev_create(struct net *net, struct net_device *dev,
653ef6a3
GM		 3193 struct vxlan_config *conf,
3194 struct netlink_ext_ack *extack)
c80498e3
ND		 3195{
3196 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3197 struct vxlan_dev *vxlan = netdev_priv(dev);
3198 int err;
3199
653ef6a3 3200 err = vxlan_dev_configure(net, dev, conf, false, extack);
c80498e3
ND		 3201 if (err)
3202 return err;
3203
3204 dev->ethtool_ops = &vxlan_ethtool_ops;
3205
3206 /* create an fdb entry for a valid default destination */
3207 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
7431016b 3208 err = vxlan_fdb_update(vxlan, all_zeros_mac,
c80498e3
ND		 3209 &vxlan->default_dst.remote_ip,
3210 NUD_REACHABLE | NUD_PERMANENT,
3211 NLM_F_EXCL | NLM_F_CREATE,
3212 vxlan->cfg.dst_port,
3213 vxlan->default_dst.remote_vni,
3214 vxlan->default_dst.remote_vni,
3215 vxlan->default_dst.remote_ifindex,
3216 NTF_SELF);
3217 if (err)
3218 return err;
3219 }
3220
3221 err = register_netdevice(dev);
3222 if (err) {
3223 vxlan_fdb_delete_default(vxlan, vxlan->default_dst.remote_vni);
3224 return err;
3225 }
3226
3227 list_add(&vxlan->next, &vn->vxlan_list);
3228 return 0;
3229}
3230
8bcdc4f3
RP		 3231static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3232 struct net_device *dev, struct vxlan_config *conf,
3233 bool changelink)
0dfbdf41 3234{
8bcdc4f3 3235 struct vxlan_dev *vxlan = netdev_priv(dev);
0dfbdf41 3236
8bcdc4f3 3237 memset(conf, 0, sizeof(*conf));
e277de5f 3238
8bcdc4f3
RP		 3239 /* if changelink operation, start with old existing cfg */
3240 if (changelink)
3241 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3242
3243 if (data[IFLA_VXLAN_ID]) {
3244 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3245
3246 if (changelink && (vni != conf->vni))
3247 return -EOPNOTSUPP;
3248 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3249 }
0dfbdf41
TG		 3250
3251 if (data[IFLA_VXLAN_GROUP]) {
ce44a4ae
MS		 3252 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET))
3253 return -EOPNOTSUPP;
3254
8bcdc4f3 3255 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
ce44a4ae 3256 conf->remote_ip.sa.sa_family = AF_INET;
0dfbdf41
TG		 3257 } else if (data[IFLA_VXLAN_GROUP6]) {
3258 if (!IS_ENABLED(CONFIG_IPV6))
3259 return -EPFNOSUPPORT;
3260
ce44a4ae
MS		 3261 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6))
3262 return -EOPNOTSUPP;
3263
8bcdc4f3
RP		 3264 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3265 conf->remote_ip.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3266 }
3267
3268 if (data[IFLA_VXLAN_LOCAL]) {
ce44a4ae
MS		 3269 if (changelink && (conf->saddr.sa.sa_family != AF_INET))
3270 return -EOPNOTSUPP;
3271
8bcdc4f3
RP		 3272 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3273 conf->saddr.sa.sa_family = AF_INET;
0dfbdf41
TG		 3274 } else if (data[IFLA_VXLAN_LOCAL6]) {
3275 if (!IS_ENABLED(CONFIG_IPV6))
3276 return -EPFNOSUPPORT;
3277
ce44a4ae
MS		 3278 if (changelink && (conf->saddr.sa.sa_family != AF_INET6))
3279 return -EOPNOTSUPP;
3280
0dfbdf41 3281 /* TODO: respect scope id */
8bcdc4f3
RP		 3282 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3283 conf->saddr.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3284 }
3285
3286 if (data[IFLA_VXLAN_LINK])
8bcdc4f3 3287 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
0dfbdf41 3288
d342894c 3289 if (data[IFLA_VXLAN_TOS])
8bcdc4f3 3290 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 3291
afb97186 3292 if (data[IFLA_VXLAN_TTL])
8bcdc4f3 3293 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 3294
72f6d71e
HL		 3295 if (data[IFLA_VXLAN_TTL_INHERIT]) {
3296 if (changelink)
3297 return -EOPNOTSUPP;
3298 conf->flags |= VXLAN_F_TTL_INHERIT;
3299 }
3300
e7f70af1 3301 if (data[IFLA_VXLAN_LABEL])
8bcdc4f3 3302 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
e7f70af1
DB		 3303 IPV6_FLOWLABEL_MASK;
3304
8bcdc4f3 3305 if (data[IFLA_VXLAN_LEARNING]) {
dc5321d7 3306 if (nla_get_u8(data[IFLA_VXLAN_LEARNING]))
8bcdc4f3 3307 conf->flags |= VXLAN_F_LEARN;
dc5321d7 3308 else
8bcdc4f3 3309 conf->flags &= ~VXLAN_F_LEARN;
8bcdc4f3
RP		 3310 } else if (!changelink) {
3311 /* default to learn on a new device */
3312 conf->flags |= VXLAN_F_LEARN;
3313 }
d342894c 3314
8bcdc4f3
RP		 3315 if (data[IFLA_VXLAN_AGEING]) {
3316 if (changelink)
3317 return -EOPNOTSUPP;
3318 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3319 }
d342894c 3320
8bcdc4f3
RP		 3321 if (data[IFLA_VXLAN_PROXY]) {
3322 if (changelink)
3323 return -EOPNOTSUPP;
3324 if (nla_get_u8(data[IFLA_VXLAN_PROXY]))
3325 conf->flags |= VXLAN_F_PROXY;
3326 }
e4f67add 3327
8bcdc4f3
RP		 3328 if (data[IFLA_VXLAN_RSC]) {
3329 if (changelink)
3330 return -EOPNOTSUPP;
3331 if (nla_get_u8(data[IFLA_VXLAN_RSC]))
3332 conf->flags |= VXLAN_F_RSC;
3333 }
e4f67add 3334
8bcdc4f3
RP		 3335 if (data[IFLA_VXLAN_L2MISS]) {
3336 if (changelink)
3337 return -EOPNOTSUPP;
3338 if (nla_get_u8(data[IFLA_VXLAN_L2MISS]))
3339 conf->flags |= VXLAN_F_L2MISS;
3340 }
e4f67add 3341
8bcdc4f3
RP		 3342 if (data[IFLA_VXLAN_L3MISS]) {
3343 if (changelink)
3344 return -EOPNOTSUPP;
3345 if (nla_get_u8(data[IFLA_VXLAN_L3MISS]))
3346 conf->flags |= VXLAN_F_L3MISS;
3347 }
e4f67add 3348
8bcdc4f3
RP		 3349 if (data[IFLA_VXLAN_LIMIT]) {
3350 if (changelink)
3351 return -EOPNOTSUPP;
3352 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3353 }
d342894c 3354
8bcdc4f3
RP		 3355 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3356 if (changelink)
3357 return -EOPNOTSUPP;
3358 if (nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3359 conf->flags |= VXLAN_F_COLLECT_METADATA;
3360 }
f8a9b1bc 3361
05f47d69 3362 if (data[IFLA_VXLAN_PORT_RANGE]) {
8bcdc4f3
RP		 3363 if (!changelink) {
3364 const struct ifla_vxlan_port_range *p
3365 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3366 conf->port_min = ntohs(p->low);
3367 conf->port_max = ntohs(p->high);
3368 } else {
3369 return -EOPNOTSUPP;
3370 }
05f47d69 3371 }
3372
8bcdc4f3
RP		 3373 if (data[IFLA_VXLAN_PORT]) {
3374 if (changelink)
3375 return -EOPNOTSUPP;
3376 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3377 }
823aa873 3378
8bcdc4f3
RP		 3379 if (data[IFLA_VXLAN_UDP_CSUM]) {
3380 if (changelink)
3381 return -EOPNOTSUPP;
3382 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3383 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3384 }
359a0ea9 3385
8bcdc4f3
RP		 3386 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3387 if (changelink)
3388 return -EOPNOTSUPP;
3389 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
3390 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
3391 }
359a0ea9 3392
8bcdc4f3
RP		 3393 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3394 if (changelink)
3395 return -EOPNOTSUPP;
3396 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
3397 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
3398 }
359a0ea9 3399
8bcdc4f3
RP		 3400 if (data[IFLA_VXLAN_REMCSUM_TX]) {
3401 if (changelink)
3402 return -EOPNOTSUPP;
3403 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
3404 conf->flags |= VXLAN_F_REMCSUM_TX;
3405 }
dfd8645e 3406
8bcdc4f3
RP		 3407 if (data[IFLA_VXLAN_REMCSUM_RX]) {
3408 if (changelink)
3409 return -EOPNOTSUPP;
3410 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3411 conf->flags |= VXLAN_F_REMCSUM_RX;
3412 }
3413
3414 if (data[IFLA_VXLAN_GBP]) {
3415 if (changelink)
3416 return -EOPNOTSUPP;
3417 conf->flags |= VXLAN_F_GBP;
3418 }
3419
3420 if (data[IFLA_VXLAN_GPE]) {
3421 if (changelink)
3422 return -EOPNOTSUPP;
3423 conf->flags |= VXLAN_F_GPE;
3424 }
3425
3426 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3427 if (changelink)
3428 return -EOPNOTSUPP;
3429 conf->flags |= VXLAN_F_REMCSUM_NOPARTIAL;
3430 }
3431
3432 if (tb[IFLA_MTU]) {
3433 if (changelink)
3434 return -EOPNOTSUPP;
3435 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3436 }
3437
3438 return 0;
3439}
3440
3441static int vxlan_newlink(struct net *src_net, struct net_device *dev,
7a3f4a18
MS		 3442 struct nlattr *tb[], struct nlattr *data[],
3443 struct netlink_ext_ack *extack)
8bcdc4f3 3444{
8bcdc4f3
RP		 3445 struct vxlan_config conf;
3446 int err;
3447
3448 err = vxlan_nl2conf(tb, data, dev, &conf, false);
3449 if (err)
3450 return err;
3451
653ef6a3 3452 return __vxlan_dev_create(src_net, dev, &conf, extack);
8bcdc4f3 3453}
dfd8645e 3454
8bcdc4f3 3455static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
ad744b22
MS		 3456 struct nlattr *data[],
3457 struct netlink_ext_ack *extack)
8bcdc4f3
RP		 3458{
3459 struct vxlan_dev *vxlan = netdev_priv(dev);
3460 struct vxlan_rdst *dst = &vxlan->default_dst;
3461 struct vxlan_rdst old_dst;
3462 struct vxlan_config conf;
3463 int err;
3464
3465 err = vxlan_nl2conf(tb, data,
3466 dev, &conf, true);
3467 if (err)
3468 return err;
3511494c 3469
8bcdc4f3 3470 memcpy(&old_dst, dst, sizeof(struct vxlan_rdst));
e1e5314d 3471
653ef6a3 3472 err = vxlan_dev_configure(vxlan->net, dev, &conf, true, extack);
8bcdc4f3
RP		 3473 if (err)
3474 return err;
0ace2ca8 3475
8bcdc4f3
RP		 3476 /* handle default dst entry */
3477 if (!vxlan_addr_equal(&dst->remote_ip, &old_dst.remote_ip)) {
3478 spin_lock_bh(&vxlan->hash_lock);
3479 if (!vxlan_addr_any(&old_dst.remote_ip))
3480 __vxlan_fdb_delete(vxlan, all_zeros_mac,
3481 old_dst.remote_ip,
3482 vxlan->cfg.dst_port,
3483 old_dst.remote_vni,
3484 old_dst.remote_vni,
3485 old_dst.remote_ifindex, 0);
3486
3487 if (!vxlan_addr_any(&dst->remote_ip)) {
7431016b 3488 err = vxlan_fdb_update(vxlan, all_zeros_mac,
8bcdc4f3
RP		 3489 &dst->remote_ip,
3490 NUD_REACHABLE | NUD_PERMANENT,
3491 NLM_F_CREATE | NLM_F_APPEND,
3492 vxlan->cfg.dst_port,
3493 dst->remote_vni,
3494 dst->remote_vni,
3495 dst->remote_ifindex,
3496 NTF_SELF);
3497 if (err) {
3498 spin_unlock_bh(&vxlan->hash_lock);
3499 return err;
3500 }
3501 }
3502 spin_unlock_bh(&vxlan->hash_lock);
3503 }
ce577668 3504
8bcdc4f3 3505 return 0;
d342894c 3506}
3507
3508static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3509{
3510 struct vxlan_dev *vxlan = netdev_priv(dev);
3511
8b3f9337
RP		 3512 vxlan_flush(vxlan, true);
3513
58ce31cc 3514 gro_cells_destroy(&vxlan->gro_cells);
553675fb 3515 list_del(&vxlan->next);
d342894c 3516 unregister_netdevice_queue(dev, head);
3517}
3518
3519static size_t vxlan_get_size(const struct net_device *dev)
3520{
3521
3522 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3523 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3524 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3525 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3526 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3527 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
e7f70af1 3528 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 3529 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 3530 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3531 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3532 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3533 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3534 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3535 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3536 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3537 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3538 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3539 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3540 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3541 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3542 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3543 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3544 0;
3545}
3546
3547static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3548{
3549 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3550 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3551 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3552 .low = htons(vxlan->cfg.port_min),
3553 .high = htons(vxlan->cfg.port_max),
05f47d69 3554 };
d342894c 3555
54bfd872 3556 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3557 goto nla_put_failure;
3558
e4c7ed41
CW		 3559 if (!vxlan_addr_any(&dst->remote_ip)) {
3560 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3561 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3562 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3563 goto nla_put_failure;
3564#if IS_ENABLED(CONFIG_IPV6)
3565 } else {
930345ea
JB		 3566 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3567 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3568 goto nla_put_failure;
3569#endif
3570 }
3571 }
d342894c 3572
c7995c43 3573 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3574 goto nla_put_failure;
3575
0dfbdf41
TG		 3576 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3577 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3578 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3579 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3580 goto nla_put_failure;
3581#if IS_ENABLED(CONFIG_IPV6)
3582 } else {
930345ea 3583 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3584 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3585 goto nla_put_failure;
3586#endif
3587 }
3588 }
d342894c 3589
0dfbdf41
TG		 3590 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3591 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e7f70af1 3592 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add 3593 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
dc5321d7 3594 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
e4f67add 3595 nla_put_u8(skb, IFLA_VXLAN_PROXY,
dc5321d7
MS		 3596 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
3597 nla_put_u8(skb, IFLA_VXLAN_RSC,
3598 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
e4f67add 3599 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
dc5321d7 3600 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
e4f67add 3601 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
dc5321d7 3602 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
da8b43c0 3603 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
dc5321d7 3604 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3605 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3606 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3607 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3608 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
dc5321d7 3609 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9 3610 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
dc5321d7 3611 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
359a0ea9 3612 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dc5321d7 3613 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
dfd8645e 3614 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
dc5321d7 3615 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
dfd8645e 3616 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
dc5321d7 3617 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
d342894c 3618 goto nla_put_failure;
3619
05f47d69 3620 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3621 goto nla_put_failure;
3622
dc5321d7 3623 if (vxlan->cfg.flags & VXLAN_F_GBP &&
3511494c
TG		 3624 nla_put_flag(skb, IFLA_VXLAN_GBP))
3625 goto nla_put_failure;
3626
dc5321d7 3627 if (vxlan->cfg.flags & VXLAN_F_GPE &&
e1e5314d
JB		 3628 nla_put_flag(skb, IFLA_VXLAN_GPE))
3629 goto nla_put_failure;
3630
dc5321d7 3631 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
0ace2ca8
TH		 3632 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3633 goto nla_put_failure;
3634
d342894c 3635 return 0;
3636
3637nla_put_failure:
3638 return -EMSGSIZE;
3639}
3640
1728d4fa
ND		 3641static struct net *vxlan_get_link_net(const struct net_device *dev)
3642{
3643 struct vxlan_dev *vxlan = netdev_priv(dev);
3644
3645 return vxlan->net;
3646}
3647
d342894c 3648static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3649 .kind = "vxlan",
3650 .maxtype = IFLA_VXLAN_MAX,
3651 .policy = vxlan_policy,
3652 .priv_size = sizeof(struct vxlan_dev),
3653 .setup = vxlan_setup,
3654 .validate = vxlan_validate,
3655 .newlink = vxlan_newlink,
8bcdc4f3 3656 .changelink = vxlan_changelink,
d342894c 3657 .dellink = vxlan_dellink,
3658 .get_size = vxlan_get_size,
3659 .fill_info = vxlan_fill_info,
1728d4fa 3660 .get_link_net = vxlan_get_link_net,
d342894c 3661};
3662
cf5da330
ND		 3663struct net_device *vxlan_dev_create(struct net *net, const char *name,
3664 u8 name_assign_type,
3665 struct vxlan_config *conf)
3666{
3667 struct nlattr *tb[IFLA_MAX + 1];
3668 struct net_device *dev;
3669 int err;
3670
3671 memset(&tb, 0, sizeof(tb));
3672
3673 dev = rtnl_create_link(net, name, name_assign_type,
3674 &vxlan_link_ops, tb);
3675 if (IS_ERR(dev))
3676 return dev;
3677
653ef6a3 3678 err = __vxlan_dev_create(net, dev, conf, NULL);
cf5da330
ND		 3679 if (err < 0) {
3680 free_netdev(dev);
3681 return ERR_PTR(err);
3682 }
3683
3684 err = rtnl_configure_link(dev, NULL);
3685 if (err < 0) {
3686 LIST_HEAD(list_kill);
3687
3688 vxlan_dellink(dev, &list_kill);
3689 unregister_netdevice_many(&list_kill);
3690 return ERR_PTR(err);
3691 }
3692
3693 return dev;
3694}
3695EXPORT_SYMBOL_GPL(vxlan_dev_create);
3696
acaf4e70
DB		 3697static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3698 struct net_device *dev)
3699{
3700 struct vxlan_dev *vxlan, *next;
3701 LIST_HEAD(list_kill);
3702
3703 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3704 struct vxlan_rdst *dst = &vxlan->default_dst;
3705
3706 /* In case we created vxlan device with carrier
3707 * and we loose the carrier due to module unload
3708 * we also need to remove vxlan device. In other
3709 * cases, it's not necessary and remote_ifindex
3710 * is 0 here, so no matches.
3711 */
3712 if (dst->remote_ifindex == dev->ifindex)
3713 vxlan_dellink(vxlan->dev, &list_kill);
3714 }
3715
3716 unregister_netdevice_many(&list_kill);
3717}
3718
b7aade15
HFS		 3719static int vxlan_netdevice_event(struct notifier_block *unused,
3720 unsigned long event, void *ptr)
acaf4e70
DB		 3721{
3722 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3723 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3724
04584957
SD		 3725 if (event == NETDEV_UNREGISTER) {
3726 vxlan_offload_rx_ports(dev, false);
acaf4e70 3727 vxlan_handle_lowerdev_unregister(vn, dev);
04584957
SD		 3728 } else if (event == NETDEV_REGISTER) {
3729 vxlan_offload_rx_ports(dev, true);
3730 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
3731 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
2d2b13fc 3732 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
04584957 3733 }
acaf4e70
DB		 3734
3735 return NOTIFY_DONE;
3736}
3737
3738static struct notifier_block vxlan_notifier_block __read_mostly = {
b7aade15 3739 .notifier_call = vxlan_netdevice_event,
acaf4e70
DB		 3740};
3741
d342894c 3742static __net_init int vxlan_init_net(struct net *net)
3743{
3744 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3745 unsigned int h;
d342894c 3746
553675fb 3747 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3748 spin_lock_init(&vn->sock_lock);
d342894c 3749
553675fb 3750 for (h = 0; h < PORT_HASH_SIZE; ++h)
3751 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3752
3753 return 0;
3754}
3755
57b61127 3756static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
f01ec1c0
ND		 3757{
3758 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3759 struct vxlan_dev *vxlan, *next;
3760 struct net_device *dev, *aux;
0e4ec5ac 3761 unsigned int h;
f01ec1c0 3762
f01ec1c0
ND		 3763 for_each_netdev_safe(net, dev, aux)
3764 if (dev->rtnl_link_ops == &vxlan_link_ops)
57b61127 3765 unregister_netdevice_queue(dev, head);
f01ec1c0
ND		 3766
3767 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3768 /* If vxlan->dev is in the same netns, it has already been added
3769 * to the list by the previous loop.
3770 */
58ce31cc
TH		 3771 if (!net_eq(dev_net(vxlan->dev), net)) {
3772 gro_cells_destroy(&vxlan->gro_cells);
57b61127 3773 unregister_netdevice_queue(vxlan->dev, head);
58ce31cc 3774 }
f01ec1c0
ND		 3775 }
3776
0e4ec5ac
VA		 3777 for (h = 0; h < PORT_HASH_SIZE; ++h)
3778 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
f01ec1c0
ND		 3779}
3780
57b61127
HY		 3781static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
3782{
3783 struct net *net;
3784 LIST_HEAD(list);
3785
3786 rtnl_lock();
3787 list_for_each_entry(net, net_list, exit_list)
3788 vxlan_destroy_tunnels(net, &list);
3789
3790 unregister_netdevice_many(&list);
3791 rtnl_unlock();
3792}
3793
d342894c 3794static struct pernet_operations vxlan_net_ops = {
3795 .init = vxlan_init_net,
57b61127 3796 .exit_batch = vxlan_exit_batch_net,
d342894c 3797 .id = &vxlan_net_id,
3798 .size = sizeof(struct vxlan_net),
3799};
3800
3801static int __init vxlan_init_module(void)
3802{
3803 int rc;
3804
3805 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3806
783c1463 3807 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3808 if (rc)
3809 goto out1;
3810
acaf4e70 3811 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3812 if (rc)
3813 goto out2;
3814
acaf4e70
DB		 3815 rc = rtnl_link_register(&vxlan_link_ops);
3816 if (rc)
3817 goto out3;
d342894c 3818
acaf4e70
DB		 3819 return 0;
3820out3:
3821 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3822out2:
783c1463 3823 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3824out1:
3825 return rc;
3826}
7332a13b 3827late_initcall(vxlan_init_module);
d342894c 3828
3829static void __exit vxlan_cleanup_module(void)
3830{
b7153984 3831 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3832 unregister_netdevice_notifier(&vxlan_notifier_block);
783c1463
DB		 3833 unregister_pernet_subsys(&vxlan_net_ops);
3834 /* rcu_barrier() is called by netns */
d342894c 3835}
3836module_exit(vxlan_cleanup_module);
3837
3838MODULE_LICENSE("GPL");
3839MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3840MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3841MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3842MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Hirsute Linux kernel mirror