]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'mlx5-updates-2017-06-27' of git://git.kernel.org/pub/scm/linux/kernel...
[mirror_ubuntu-artful-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>
012a5729 29#include <net/vxlan.h>
40d29af0 30
e4c7ed41 31#if IS_ENABLED(CONFIG_IPV6)
e4c7ed41 32#include <net/ip6_tunnel.h>
660d98ca 33#include <net/ip6_checksum.h>
e4c7ed41 34#endif
d342894c 35
36#define VXLAN_VERSION "0.1"
37
553675fb 38#define PORT_HASH_BITS 8
39#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 40#define FDB_AGE_DEFAULT 300 /* 5 min */
41#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
42
23c578bf 43/* UDP port for VXLAN traffic.
44 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 45 * for compatibility with early adopters.
23c578bf 46 */
9daaa397
SH		 47static unsigned short vxlan_port __read_mostly = 8472;
48module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 49MODULE_PARM_DESC(udp_port, "Destination UDP port");
50
51static bool log_ecn_error = true;
52module_param(log_ecn_error, bool, 0644);
53MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
54
c7d03a00 55static unsigned int vxlan_net_id;
0dfbdf41 56static struct rtnl_link_ops vxlan_link_ops;
553675fb 57
7256eac1 58static const u8 all_zeros_mac[ETH_ALEN + 2];
afbd8bae 59
205f356d 60static int vxlan_sock_add(struct vxlan_dev *vxlan);
614732ea 61
a53cb29b
MB		 62static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
63
553675fb 64/* per-network namespace private data for this module */
65struct vxlan_net {
66 struct list_head vxlan_list;
67 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 68 spinlock_t sock_lock;
553675fb 69};
70
d342894c 71/* Forwarding table entry */
72struct vxlan_fdb {
73 struct hlist_node hlist; /* linked list of entries */
74 struct rcu_head rcu;
75 unsigned long updated; /* jiffies */
76 unsigned long used;
3e61aa8f 77 struct list_head remotes;
7177a3b0 78 u8 eth_addr[ETH_ALEN];
d342894c 79 u16 state; /* see ndm_state */
3ad7a4b1 80 __be32 vni;
ae884082 81 u8 flags; /* see ndm_flags */
d342894c 82};
83
d342894c 84/* salt for hash table */
85static u32 vxlan_salt __read_mostly;
86
ee122c79
TG		 87static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
88{
e7030878
TG		 89 return vs->flags & VXLAN_F_COLLECT_METADATA ||
90 ip_tunnel_collect_metadata();
ee122c79
TG		 91}
92
e4c7ed41
CW		 93#if IS_ENABLED(CONFIG_IPV6)
94static inline
95bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
96{
f0ef3126
JB		 97 if (a->sa.sa_family != b->sa.sa_family)
98 return false;
99 if (a->sa.sa_family == AF_INET6)
100 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
101 else
102 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 103}
104
105static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
106{
f0ef3126
JB		 107 if (ipa->sa.sa_family == AF_INET6)
108 return ipv6_addr_any(&ipa->sin6.sin6_addr);
109 else
110 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 111}
112
113static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
114{
f0ef3126
JB		 115 if (ipa->sa.sa_family == AF_INET6)
116 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
117 else
118 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 119}
120
121static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
122{
f0ef3126 123 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 124 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 125 ip->sa.sa_family = AF_INET6;
126 return 0;
127 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 128 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 129 ip->sa.sa_family = AF_INET;
130 return 0;
131 } else {
132 return -EAFNOSUPPORT;
133 }
e4c7ed41
CW		 134}
135
136static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 137 const union vxlan_addr *ip)
e4c7ed41 138{
f0ef3126 139 if (ip->sa.sa_family == AF_INET6)
930345ea 140 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 141 else
930345ea 142 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 143}
144
145#else /* !CONFIG_IPV6 */
146
147static inline
148bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
149{
f0ef3126 150 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 151}
152
153static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
154{
f0ef3126 155 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 156}
157
158static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
159{
f0ef3126 160 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 161}
162
163static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
164{
f0ef3126
JB		 165 if (nla_len(nla) >= sizeof(struct in6_addr)) {
166 return -EAFNOSUPPORT;
167 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 168 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 169 ip->sa.sa_family = AF_INET;
170 return 0;
171 } else {
172 return -EAFNOSUPPORT;
173 }
e4c7ed41
CW		 174}
175
176static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 177 const union vxlan_addr *ip)
e4c7ed41 178{
930345ea 179 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 180}
181#endif
182
553675fb 183/* Virtual Network hash table head */
54bfd872 184static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 185{
54bfd872 186 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 187}
188
189/* Socket hash table head */
190static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 191{
192 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
193
553675fb 194 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
195}
196
3e61aa8f
SH		 197/* First remote destination for a forwarding entry.
198 * Guaranteed to be non-NULL because remotes are never deleted.
199 */
5ca5461c 200static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 201{
5ca5461c 202 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
203}
204
205static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
206{
207 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 208}
209
ac5132d1
TG		 210/* Find VXLAN socket based on network namespace, address family and UDP port
211 * and enabled unshareable flags.
212 */
213static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
214 __be16 port, u32 flags)
553675fb 215{
216 struct vxlan_sock *vs;
af33c1ad
TH		 217
218 flags &= VXLAN_F_RCV_FLAGS;
553675fb 219
220 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 221 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 222 vxlan_get_sk_family(vs) == family &&
af33c1ad 223 vs->flags == flags)
553675fb 224 return vs;
225 }
226 return NULL;
d342894c 227}
228
49f810f0
MS		 229static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
230 __be32 vni)
d342894c 231{
232 struct vxlan_dev *vxlan;
d342894c 233
b9167b2e
JB		 234 /* For flow based devices, map all packets to VNI 0 */
235 if (vs->flags & VXLAN_F_COLLECT_METADATA)
236 vni = 0;
237
54bfd872 238 hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) {
49f810f0
MS		 239 if (vxlan->default_dst.remote_vni != vni)
240 continue;
241
242 if (IS_ENABLED(CONFIG_IPV6)) {
243 const struct vxlan_config *cfg = &vxlan->cfg;
244
245 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
246 cfg->remote_ifindex != ifindex)
247 continue;
248 }
249
250 return vxlan;
d342894c 251 }
252
253 return NULL;
254}
255
5cfccc5a 256/* Look up VNI in a per net namespace table */
49f810f0
MS		 257static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
258 __be32 vni, sa_family_t family,
259 __be16 port, u32 flags)
5cfccc5a
PS		 260{
261 struct vxlan_sock *vs;
262
ac5132d1 263 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 264 if (!vs)
265 return NULL;
266
49f810f0 267 return vxlan_vs_find_vni(vs, ifindex, vni);
5cfccc5a
PS		 268}
269
d342894c 270/* Fill in neighbour message in skbuff. */
271static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 272 const struct vxlan_fdb *fdb,
273 u32 portid, u32 seq, int type, unsigned int flags,
274 const struct vxlan_rdst *rdst)
d342894c 275{
276 unsigned long now = jiffies;
277 struct nda_cacheinfo ci;
278 struct nlmsghdr *nlh;
279 struct ndmsg *ndm;
e4f67add 280 bool send_ip, send_eth;
d342894c 281
282 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
283 if (nlh == NULL)
284 return -EMSGSIZE;
285
286 ndm = nlmsg_data(nlh);
287 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 288
289 send_eth = send_ip = true;
290
291 if (type == RTM_GETNEIGH) {
e4c7ed41 292 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add 293 send_eth = !is_zero_ether_addr(fdb->eth_addr);
8f48ba71 294 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
e4f67add
DS		 295 } else
296 ndm->ndm_family = AF_BRIDGE;
d342894c 297 ndm->ndm_state = fdb->state;
298 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 299 ndm->ndm_flags = fdb->flags;
545469f7 300 ndm->ndm_type = RTN_UNICAST;
d342894c 301
193523bf 302 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 303 nla_put_s32(skb, NDA_LINK_NETNSID,
38f507f1 304 peernet2id(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 305 goto nla_put_failure;
306
e4f67add 307 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 308 goto nla_put_failure;
309
e4c7ed41 310 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 311 goto nla_put_failure;
312
0dfbdf41 313 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 314 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
315 goto nla_put_failure;
c7995c43 316 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
54bfd872 317 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
6681712d 318 goto nla_put_failure;
dc5321d7 319 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
3ad7a4b1
RP		 320 nla_put_u32(skb, NDA_SRC_VNI,
321 be32_to_cpu(fdb->vni)))
322 goto nla_put_failure;
6681712d
DS		 323 if (rdst->remote_ifindex &&
324 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 325 goto nla_put_failure;
326
327 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
328 ci.ndm_confirmed = 0;
329 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
330 ci.ndm_refcnt = 0;
331
332 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
333 goto nla_put_failure;
334
053c095a
JB		 335 nlmsg_end(skb, nlh);
336 return 0;
d342894c 337
338nla_put_failure:
339 nlmsg_cancel(skb, nlh);
340 return -EMSGSIZE;
341}
342
343static inline size_t vxlan_nlmsg_size(void)
344{
345 return NLMSG_ALIGN(sizeof(struct ndmsg))
346 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 347 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 348 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 349 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
350 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 351 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 352 + nla_total_size(sizeof(struct nda_cacheinfo));
353}
354
9e4b93f9
ND		 355static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
356 struct vxlan_rdst *rd, int type)
d342894c 357{
358 struct net *net = dev_net(vxlan->dev);
359 struct sk_buff *skb;
360 int err = -ENOBUFS;
361
362 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
363 if (skb == NULL)
364 goto errout;
365
9e4b93f9 366 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 367 if (err < 0) {
368 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
369 WARN_ON(err == -EMSGSIZE);
370 kfree_skb(skb);
371 goto errout;
372 }
373
374 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
375 return;
376errout:
377 if (err < 0)
378 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
379}
380
e4c7ed41 381static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 382{
383 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 384 struct vxlan_fdb f = {
385 .state = NUD_STALE,
386 };
387 struct vxlan_rdst remote = {
e4c7ed41 388 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 389 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 390 };
3e61aa8f 391
9e4b93f9 392 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 393}
394
395static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
396{
bb3fd687
SH		 397 struct vxlan_fdb f = {
398 .state = NUD_STALE,
399 };
9e4b93f9 400 struct vxlan_rdst remote = { };
e4f67add 401
e4f67add
DS		 402 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
403
9e4b93f9 404 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 405}
406
d342894c 407/* Hash Ethernet address */
408static u32 eth_hash(const unsigned char *addr)
409{
410 u64 value = get_unaligned((u64 *)addr);
411
412 /* only want 6 bytes */
413#ifdef __BIG_ENDIAN
d342894c 414 value >>= 16;
321fb991 415#else
416 value <<= 16;
d342894c 417#endif
418 return hash_64(value, FDB_HASH_BITS);
419}
420
3ad7a4b1
RP		 421static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
422{
423 /* use 1 byte of OUI and 3 bytes of NIC */
424 u32 key = get_unaligned((u32 *)(addr + 2));
425
426 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
427}
428
d342894c 429/* Hash chain to use given mac address */
430static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
3ad7a4b1 431 const u8 *mac, __be32 vni)
d342894c 432{
dc5321d7 433 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
3ad7a4b1
RP		 434 return &vxlan->fdb_head[eth_vni_hash(mac, vni)];
435 else
436 return &vxlan->fdb_head[eth_hash(mac)];
d342894c 437}
438
439/* Look up Ethernet address in forwarding table */
014be2c8 440static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 441 const u8 *mac, __be32 vni)
d342894c 442{
3ad7a4b1 443 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
d342894c 444 struct vxlan_fdb *f;
d342894c 445
b67bfe0d 446 hlist_for_each_entry_rcu(f, head, hlist) {
3ad7a4b1 447 if (ether_addr_equal(mac, f->eth_addr)) {
dc5321d7 448 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 449 if (vni == f->vni)
450 return f;
451 } else {
452 return f;
453 }
454 }
d342894c 455 }
456
457 return NULL;
458}
459
014be2c8 460static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 461 const u8 *mac, __be32 vni)
014be2c8
SS		 462{
463 struct vxlan_fdb *f;
464
3ad7a4b1 465 f = __vxlan_find_mac(vxlan, mac, vni);
014be2c8
SS		 466 if (f)
467 f->used = jiffies;
468
469 return f;
470}
471
a5e7c10a
MR		 472/* caller should hold vxlan->hash_lock */
473static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 474 union vxlan_addr *ip, __be16 port,
54bfd872 475 __be32 vni, __u32 ifindex)
6681712d 476{
3e61aa8f 477 struct vxlan_rdst *rd;
6681712d 478
3e61aa8f 479 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 480 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 481 rd->remote_port == port &&
482 rd->remote_vni == vni &&
483 rd->remote_ifindex == ifindex)
a5e7c10a 484 return rd;
6681712d 485 }
3e61aa8f 486
a5e7c10a
MR		 487 return NULL;
488}
489
906dc186
TR		 490/* Replace destination of unicast mac */
491static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872
JB		 492 union vxlan_addr *ip, __be16 port, __be32 vni,
493 __u32 ifindex)
906dc186
TR		 494{
495 struct vxlan_rdst *rd;
496
497 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
498 if (rd)
499 return 0;
500
501 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
502 if (!rd)
503 return 0;
0c1d70af
PA		 504
505 dst_cache_reset(&rd->dst_cache);
e4c7ed41 506 rd->remote_ip = *ip;
906dc186
TR		 507 rd->remote_port = port;
508 rd->remote_vni = vni;
509 rd->remote_ifindex = ifindex;
510 return 1;
511}
512
a5e7c10a
MR		 513/* Add/update destinations for multicast */
514static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 515 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 516 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 517{
518 struct vxlan_rdst *rd;
519
520 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
521 if (rd)
522 return 0;
523
6681712d
DS		 524 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
525 if (rd == NULL)
526 return -ENOBUFS;
0c1d70af
PA		 527
528 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
529 kfree(rd);
530 return -ENOBUFS;
531 }
532
e4c7ed41 533 rd->remote_ip = *ip;
6681712d
DS		 534 rd->remote_port = port;
535 rd->remote_vni = vni;
536 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 537
538 list_add_tail_rcu(&rd->list, &f->remotes);
539
9e4b93f9 540 *rdp = rd;
6681712d
DS		 541 return 1;
542}
543
dfd8645e
TH		 544static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
545 unsigned int off,
546 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 547 __be32 vni_field,
548 struct gro_remcsum *grc,
0ace2ca8 549 bool nopartial)
dfd8645e 550{
b7fe10e5 551 size_t start, offset;
dfd8645e
TH		 552
553 if (skb->remcsum_offload)
b7fe10e5 554 return vh;
dfd8645e
TH		 555
556 if (!NAPI_GRO_CB(skb)->csum_valid)
557 return NULL;
558
54bfd872
JB		 559 start = vxlan_rco_start(vni_field);
560 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 561
b7fe10e5
TH		 562 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
563 start, offset, grc, nopartial);
dfd8645e
TH		 564
565 skb->remcsum_offload = 1;
566
567 return vh;
568}
569
5602c48c
TH		 570static struct sk_buff **vxlan_gro_receive(struct sock *sk,
571 struct sk_buff **head,
572 struct sk_buff *skb)
dc01e7d3
OG		 573{
574 struct sk_buff *p, **pp = NULL;
575 struct vxlanhdr *vh, *vh2;
9b174d88 576 unsigned int hlen, off_vx;
dc01e7d3 577 int flush = 1;
5602c48c 578 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
54bfd872 579 __be32 flags;
26c4f7da
TH		 580 struct gro_remcsum grc;
581
582 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 583
584 off_vx = skb_gro_offset(skb);
585 hlen = off_vx + sizeof(*vh);
586 vh = skb_gro_header_fast(skb, off_vx);
587 if (skb_gro_header_hard(skb, hlen)) {
588 vh = skb_gro_header_slow(skb, hlen, off_vx);
589 if (unlikely(!vh))
590 goto out;
591 }
dc01e7d3 592
dfd8645e
TH		 593 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
594
54bfd872 595 flags = vh->vx_flags;
dfd8645e
TH		 596
597 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
598 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 599 vh->vx_vni, &grc,
0ace2ca8
TH		 600 !!(vs->flags &
601 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 602
603 if (!vh)
604 goto out;
605 }
606
b7fe10e5
TH		 607 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
608
dc01e7d3
OG		 609 for (p = *head; p; p = p->next) {
610 if (!NAPI_GRO_CB(p)->same_flow)
611 continue;
612
613 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 614 if (vh->vx_flags != vh2->vx_flags ||
615 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 616 NAPI_GRO_CB(p)->same_flow = 0;
617 continue;
618 }
dc01e7d3
OG		 619 }
620
fcd91dd4 621 pp = call_gro_receive(eth_gro_receive, head, skb);
c194cf93 622 flush = 0;
dc01e7d3 623
dc01e7d3 624out:
26c4f7da 625 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 626 NAPI_GRO_CB(skb)->flush |= flush;
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
d342894c 639/* Add new entry to forwarding table -- assumes lock held */
640static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 641 const u8 *mac, union vxlan_addr *ip,
6681712d 642 __u16 state, __u16 flags,
3ad7a4b1
RP		 643 __be16 port, __be32 src_vni, __be32 vni,
644 __u32 ifindex, __u8 ndm_flags)
d342894c 645{
9e4b93f9 646 struct vxlan_rdst *rd = NULL;
d342894c 647 struct vxlan_fdb *f;
648 int notify = 0;
17b46365 649 int rc;
d342894c 650
3ad7a4b1 651 f = __vxlan_find_mac(vxlan, mac, src_vni);
d342894c 652 if (f) {
653 if (flags & NLM_F_EXCL) {
654 netdev_dbg(vxlan->dev,
655 "lost race to create %pM\n", mac);
656 return -EEXIST;
657 }
658 if (f->state != state) {
659 f->state = state;
660 f->updated = jiffies;
661 notify = 1;
662 }
ae884082
DS		 663 if (f->flags != ndm_flags) {
664 f->flags = ndm_flags;
665 f->updated = jiffies;
666 notify = 1;
667 }
906dc186
TR		 668 if ((flags & NLM_F_REPLACE)) {
669 /* Only change unicasts */
670 if (!(is_multicast_ether_addr(f->eth_addr) ||
671 is_zero_ether_addr(f->eth_addr))) {
60840429 672 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 673 ifindex);
906dc186
TR		 674 } else
675 return -EOPNOTSUPP;
676 }
6681712d 677 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 678 (is_multicast_ether_addr(f->eth_addr) ||
679 is_zero_ether_addr(f->eth_addr))) {
17b46365 680 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
6681712d
DS		 681
682 if (rc < 0)
683 return rc;
684 notify |= rc;
685 }
d342894c 686 } else {
687 if (!(flags & NLM_F_CREATE))
688 return -ENOENT;
689
0dfbdf41
TG		 690 if (vxlan->cfg.addrmax &&
691 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 692 return -ENOSPC;
693
906dc186
TR		 694 /* Disallow replace to add a multicast entry */
695 if ((flags & NLM_F_REPLACE) &&
696 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
697 return -EOPNOTSUPP;
698
e4c7ed41 699 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 700 f = kmalloc(sizeof(*f), GFP_ATOMIC);
701 if (!f)
702 return -ENOMEM;
703
704 notify = 1;
d342894c 705 f->state = state;
ae884082 706 f->flags = ndm_flags;
d342894c 707 f->updated = f->used = jiffies;
3ad7a4b1 708 f->vni = src_vni;
3e61aa8f 709 INIT_LIST_HEAD(&f->remotes);
d342894c 710 memcpy(f->eth_addr, mac, ETH_ALEN);
711
17b46365
HY		 712 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
713 if (rc < 0) {
714 kfree(f);
715 return rc;
716 }
3e61aa8f 717
d342894c 718 ++vxlan->addrcnt;
719 hlist_add_head_rcu(&f->hlist,
3ad7a4b1 720 vxlan_fdb_head(vxlan, mac, src_vni));
d342894c 721 }
722
9e4b93f9
ND		 723 if (notify) {
724 if (rd == NULL)
725 rd = first_remote_rtnl(f);
726 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
727 }
d342894c 728
729 return 0;
730}
731
6706c82e 732static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 733{
734 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 735 struct vxlan_rdst *rd, *nd;
6681712d 736
0c1d70af
PA		 737 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
738 dst_cache_destroy(&rd->dst_cache);
6681712d 739 kfree(rd);
0c1d70af 740 }
6681712d
DS		 741 kfree(f);
742}
743
d342894c 744static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
745{
746 netdev_dbg(vxlan->dev,
747 "delete %pM\n", f->eth_addr);
748
749 --vxlan->addrcnt;
9e4b93f9 750 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 751
752 hlist_del_rcu(&f->hlist);
6681712d 753 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 754}
755
35cf2845
LR		 756static void vxlan_dst_free(struct rcu_head *head)
757{
758 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
759
760 dst_cache_destroy(&rd->dst_cache);
761 kfree(rd);
762}
763
764static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
765 struct vxlan_rdst *rd)
766{
767 list_del_rcu(&rd->list);
768 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
769 call_rcu(&rd->rcu, vxlan_dst_free);
770}
771
f0b074be 772static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
3ad7a4b1
RP		 773 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
774 __be32 *vni, u32 *ifindex)
d342894c 775{
6681712d 776 struct net *net = dev_net(vxlan->dev);
e4c7ed41 777 int err;
d342894c 778
f0b074be 779 if (tb[NDA_DST]) {
e4c7ed41
CW		 780 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
781 if (err)
782 return err;
f0b074be 783 } else {
e4c7ed41
CW		 784 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
785 if (remote->sa.sa_family == AF_INET) {
786 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
787 ip->sa.sa_family = AF_INET;
788#if IS_ENABLED(CONFIG_IPV6)
789 } else {
790 ip->sin6.sin6_addr = in6addr_any;
791 ip->sa.sa_family = AF_INET6;
792#endif
793 }
f0b074be 794 }
d342894c 795
6681712d 796 if (tb[NDA_PORT]) {
73cf3317 797 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 798 return -EINVAL;
f0b074be
MR		 799 *port = nla_get_be16(tb[NDA_PORT]);
800 } else {
0dfbdf41 801 *port = vxlan->cfg.dst_port;
f0b074be 802 }
6681712d
DS		 803
804 if (tb[NDA_VNI]) {
805 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
806 return -EINVAL;
54bfd872 807 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 808 } else {
809 *vni = vxlan->default_dst.remote_vni;
810 }
6681712d 811
3ad7a4b1
RP		 812 if (tb[NDA_SRC_VNI]) {
813 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
814 return -EINVAL;
815 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
816 } else {
817 *src_vni = vxlan->default_dst.remote_vni;
818 }
819
6681712d 820 if (tb[NDA_IFINDEX]) {
5abb0029 821 struct net_device *tdev;
6681712d
DS		 822
823 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
824 return -EINVAL;
f0b074be 825 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 826 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 827 if (!tdev)
6681712d 828 return -EADDRNOTAVAIL;
f0b074be
MR		 829 } else {
830 *ifindex = 0;
831 }
832
833 return 0;
834}
835
836/* Add static entry (via netlink) */
837static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
838 struct net_device *dev,
f6f6424b 839 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 840{
841 struct vxlan_dev *vxlan = netdev_priv(dev);
842 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 843 union vxlan_addr ip;
f0b074be 844 __be16 port;
3ad7a4b1 845 __be32 src_vni, vni;
54bfd872 846 u32 ifindex;
f0b074be
MR		 847 int err;
848
849 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
850 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
851 ndm->ndm_state);
852 return -EINVAL;
853 }
854
855 if (tb[NDA_DST] == NULL)
856 return -EINVAL;
857
3ad7a4b1 858 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
f0b074be
MR		 859 if (err)
860 return err;
6681712d 861
5933a7bb
MR		 862 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
863 return -EAFNOSUPPORT;
864
d342894c 865 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 866 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
3ad7a4b1 867 port, src_vni, vni, ifindex, ndm->ndm_flags);
d342894c 868 spin_unlock_bh(&vxlan->hash_lock);
869
870 return err;
871}
872
3ad7a4b1
RP		 873static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
874 const unsigned char *addr, union vxlan_addr ip,
875 __be16 port, __be32 src_vni, u32 vni, u32 ifindex,
876 u16 vid)
d342894c 877{
d342894c 878 struct vxlan_fdb *f;
bc7892ba 879 struct vxlan_rdst *rd = NULL;
3ad7a4b1 880 int err = -ENOENT;
bc7892ba 881
3ad7a4b1 882 f = vxlan_find_mac(vxlan, addr, src_vni);
bc7892ba 883 if (!f)
3ad7a4b1 884 return err;
bc7892ba 885
e4c7ed41
CW		 886 if (!vxlan_addr_any(&ip)) {
887 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 888 if (!rd)
889 goto out;
890 }
891
bc7892ba
MR		 892 /* remove a destination if it's not the only one on the list,
893 * otherwise destroy the fdb entry
894 */
895 if (rd && !list_is_singular(&f->remotes)) {
35cf2845 896 vxlan_fdb_dst_destroy(vxlan, f, rd);
bc7892ba 897 goto out;
d342894c 898 }
bc7892ba
MR		 899
900 vxlan_fdb_destroy(vxlan, f);
901
902out:
3ad7a4b1
RP		 903 return 0;
904}
905
906/* Delete entry (via netlink) */
907static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
908 struct net_device *dev,
909 const unsigned char *addr, u16 vid)
910{
911 struct vxlan_dev *vxlan = netdev_priv(dev);
912 union vxlan_addr ip;
913 __be32 src_vni, vni;
914 __be16 port;
915 u32 ifindex;
916 int err;
917
918 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
919 if (err)
920 return err;
921
922 spin_lock_bh(&vxlan->hash_lock);
923 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
924 vid);
d342894c 925 spin_unlock_bh(&vxlan->hash_lock);
926
927 return err;
928}
929
930/* Dump forwarding table */
931static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 932 struct net_device *dev,
d297653d 933 struct net_device *filter_dev, int *idx)
d342894c 934{
935 struct vxlan_dev *vxlan = netdev_priv(dev);
936 unsigned int h;
d297653d 937 int err = 0;
d342894c 938
939 for (h = 0; h < FDB_HASH_SIZE; ++h) {
940 struct vxlan_fdb *f;
d342894c 941
b67bfe0d 942 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 943 struct vxlan_rdst *rd;
6681712d 944
3e61aa8f 945 list_for_each_entry_rcu(rd, &f->remotes, list) {
d297653d 946 if (*idx < cb->args[2])
07a51cd3
AW		 947 goto skip;
948
6681712d
DS		 949 err = vxlan_fdb_info(skb, vxlan, f,
950 NETLINK_CB(cb->skb).portid,
951 cb->nlh->nlmsg_seq,
952 RTM_NEWNEIGH,
953 NLM_F_MULTI, rd);
d297653d 954 if (err < 0)
3e61aa8f 955 goto out;
3e61aa8f 956skip:
d297653d 957 *idx += 1;
07a51cd3 958 }
d342894c 959 }
960 }
3e61aa8f 961out:
d297653d 962 return err;
d342894c 963}
964
965/* Watch incoming packets to learn mapping between Ethernet address
966 * and Tunnel endpoint.
c4b49512 967 * Return true if packet is bogus and should be dropped.
d342894c 968 */
26a41ae6 969static bool vxlan_snoop(struct net_device *dev,
3ad7a4b1 970 union vxlan_addr *src_ip, const u8 *src_mac,
87613de9 971 u32 src_ifindex, __be32 vni)
d342894c 972{
973 struct vxlan_dev *vxlan = netdev_priv(dev);
974 struct vxlan_fdb *f;
87613de9
MS		 975 u32 ifindex = 0;
976
977#if IS_ENABLED(CONFIG_IPV6)
978 if (src_ip->sa.sa_family == AF_INET6 &&
979 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
980 ifindex = src_ifindex;
981#endif
d342894c 982
3ad7a4b1 983 f = vxlan_find_mac(vxlan, src_mac, vni);
d342894c 984 if (likely(f)) {
5ca5461c 985 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 986
87613de9
MS		 987 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
988 rdst->remote_ifindex == ifindex))
26a41ae6 989 return false;
990
991 /* Don't migrate static entries, drop packets */
e0090a9e 992 if (f->state & (NUD_PERMANENT | NUD_NOARP))
26a41ae6 993 return true;
d342894c 994
995 if (net_ratelimit())
996 netdev_info(dev,
e4c7ed41 997 "%pM migrated from %pIS to %pIS\n",
a4870f79 998 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 999
e4c7ed41 1000 rdst->remote_ip = *src_ip;
d342894c 1001 f->updated = jiffies;
9e4b93f9 1002 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 1003 } else {
1004 /* learned new entry */
1005 spin_lock(&vxlan->hash_lock);
3bf74b1a 1006
1007 /* close off race between vxlan_flush and incoming packets */
1008 if (netif_running(dev))
1009 vxlan_fdb_create(vxlan, src_mac, src_ip,
1010 NUD_REACHABLE,
1011 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 1012 vxlan->cfg.dst_port,
3ad7a4b1 1013 vni,
3bf74b1a 1014 vxlan->default_dst.remote_vni,
87613de9 1015 ifindex, NTF_SELF);
d342894c 1016 spin_unlock(&vxlan->hash_lock);
1017 }
26a41ae6 1018
1019 return false;
d342894c 1020}
1021
d342894c 1022/* See if multicast group is already in use by other ID */
95ab0991 1023static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1024{
553675fb 1025 struct vxlan_dev *vxlan;
c6fcc4fc 1026 struct vxlan_sock *sock4;
4053ab1b
AB		 1027#if IS_ENABLED(CONFIG_IPV6)
1028 struct vxlan_sock *sock6;
1029#endif
b1be00a6 1030 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1031
c6fcc4fc 1032 sock4 = rtnl_dereference(dev->vn4_sock);
1033
95ab0991
G		 1034 /* The vxlan_sock is only used by dev, leaving group has
1035 * no effect on other vxlan devices.
1036 */
c6fcc4fc 1037 if (family == AF_INET && sock4 && atomic_read(&sock4->refcnt) == 1)
95ab0991 1038 return false;
b1be00a6 1039#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1040 sock6 = rtnl_dereference(dev->vn6_sock);
1041 if (family == AF_INET6 && sock6 && atomic_read(&sock6->refcnt) == 1)
b1be00a6
JB		 1042 return false;
1043#endif
95ab0991 1044
553675fb 1045 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1046 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1047 continue;
d342894c 1048
c6fcc4fc 1049 if (family == AF_INET &&
1050 rtnl_dereference(vxlan->vn4_sock) != sock4)
95ab0991 1051 continue;
b1be00a6 1052#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1053 if (family == AF_INET6 &&
1054 rtnl_dereference(vxlan->vn6_sock) != sock6)
b1be00a6
JB		 1055 continue;
1056#endif
95ab0991
G		 1057
1058 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1059 &dev->default_dst.remote_ip))
1060 continue;
1061
1062 if (vxlan->default_dst.remote_ifindex !=
1063 dev->default_dst.remote_ifindex)
1064 continue;
1065
1066 return true;
553675fb 1067 }
d342894c 1068
1069 return false;
1070}
1071
544a773a 1072static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
7c47cedf 1073{
b1be00a6 1074 struct vxlan_net *vn;
012a5729 1075
b1be00a6 1076 if (!vs)
544a773a 1077 return false;
7c47cedf 1078 if (!atomic_dec_and_test(&vs->refcnt))
544a773a 1079 return false;
d342894c 1080
b1be00a6 1081 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1082 spin_lock(&vn->sock_lock);
7c47cedf 1083 hlist_del_rcu(&vs->hlist);
e7b3db5e 1084 udp_tunnel_notify_del_rx_port(vs->sock,
b9adcd69
AD		 1085 (vs->flags & VXLAN_F_GPE) ?
1086 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 1087 UDP_TUNNEL_TYPE_VXLAN);
1c51a915
SH		 1088 spin_unlock(&vn->sock_lock);
1089
544a773a 1090 return true;
d342894c 1091}
1092
b1be00a6
JB		 1093static void vxlan_sock_release(struct vxlan_dev *vxlan)
1094{
c6fcc4fc 1095 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
b1be00a6 1096#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1097 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1098
57d88182 1099 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
544a773a
HFS		 1100#endif
1101
57d88182 1102 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
544a773a
HFS		 1103 synchronize_net();
1104
a53cb29b
MB		 1105 vxlan_vs_del_dev(vxlan);
1106
c6fcc4fc 1107 if (__vxlan_sock_release_prep(sock4)) {
1108 udp_tunnel_sock_release(sock4->sock);
1109 kfree(sock4);
544a773a
HFS		 1110 }
1111
1112#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1113 if (__vxlan_sock_release_prep(sock6)) {
1114 udp_tunnel_sock_release(sock6->sock);
1115 kfree(sock6);
544a773a 1116 }
b1be00a6
JB		 1117#endif
1118}
1119
56ef9c90 1120/* Update multicast group membership when first VNI on
c4b49512 1121 * multicast address is brought up
7c47cedf 1122 */
56ef9c90 1123static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1124{
b1be00a6 1125 struct sock *sk;
e4c7ed41
CW		 1126 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1127 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1128 int ret = -EINVAL;
d342894c 1129
e4c7ed41 1130 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1131 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1132 struct ip_mreqn mreq = {
1133 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1134 .imr_ifindex = ifindex,
1135 };
1136
c6fcc4fc 1137 sk = sock4->sock->sk;
b1be00a6 1138 lock_sock(sk);
56ef9c90 1139 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1140 release_sock(sk);
e4c7ed41
CW		 1141#if IS_ENABLED(CONFIG_IPV6)
1142 } else {
c6fcc4fc 1143 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1144
1145 sk = sock6->sock->sk;
b1be00a6 1146 lock_sock(sk);
56ef9c90
MRL		 1147 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1148 &ip->sin6.sin6_addr);
b1be00a6 1149 release_sock(sk);
e4c7ed41
CW		 1150#endif
1151 }
3fc2de2f 1152
56ef9c90 1153 return ret;
3fc2de2f 1154}
1155
1156/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1157static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1158{
b1be00a6 1159 struct sock *sk;
e4c7ed41
CW		 1160 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1161 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1162 int ret = -EINVAL;
3fc2de2f 1163
e4c7ed41 1164 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1165 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1166 struct ip_mreqn mreq = {
1167 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1168 .imr_ifindex = ifindex,
1169 };
1170
c6fcc4fc 1171 sk = sock4->sock->sk;
b1be00a6 1172 lock_sock(sk);
56ef9c90 1173 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1174 release_sock(sk);
e4c7ed41
CW		 1175#if IS_ENABLED(CONFIG_IPV6)
1176 } else {
c6fcc4fc 1177 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1178
1179 sk = sock6->sock->sk;
b1be00a6 1180 lock_sock(sk);
56ef9c90
MRL		 1181 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1182 &ip->sin6.sin6_addr);
b1be00a6 1183 release_sock(sk);
e4c7ed41
CW		 1184#endif
1185 }
d342894c 1186
56ef9c90 1187 return ret;
d342894c 1188}
1189
f14ecebb
JB		 1190static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1191 struct sk_buff *skb, u32 vxflags)
dfd8645e 1192{
7d34fa75 1193 size_t start, offset;
dfd8645e 1194
f14ecebb
JB		 1195 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1196 goto out;
b7fe10e5 1197
f14ecebb
JB		 1198 start = vxlan_rco_start(unparsed->vx_vni);
1199 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1200
7d34fa75 1201 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1202 return false;
dfd8645e 1203
be5cfeab
JB		 1204 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1205 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1206out:
1207 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1208 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1209 return true;
dfd8645e
TH		 1210}
1211
f14ecebb 1212static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1213 struct sk_buff *skb, u32 vxflags,
10a5af23 1214 struct vxlan_metadata *md)
3288af08 1215{
f14ecebb 1216 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1217 struct metadata_dst *tun_dst;
f14ecebb
JB		 1218
1219 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1220 goto out;
3288af08 1221
3288af08
JB		 1222 md->gbp = ntohs(gbp->policy_id);
1223
10a5af23 1224 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1225 if (tun_dst) {
3288af08 1226 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1227 tun_dst->u.tun_info.options_len = sizeof(*md);
1228 }
3288af08
JB		 1229 if (gbp->dont_learn)
1230 md->gbp |= VXLAN_GBP_DONT_LEARN;
1231
1232 if (gbp->policy_applied)
1233 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1234
64f87d36
JB		 1235 /* In flow-based mode, GBP is carried in dst_metadata */
1236 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1237 skb->mark = md->gbp;
f14ecebb
JB		 1238out:
1239 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1240}
1241
e1e5314d 1242static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
61618eea 1243 __be16 *protocol,
e1e5314d
JB		 1244 struct sk_buff *skb, u32 vxflags)
1245{
1246 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1247
1248 /* Need to have Next Protocol set for interfaces in GPE mode. */
1249 if (!gpe->np_applied)
1250 return false;
1251 /* "The initial version is 0. If a receiver does not support the
1252 * version indicated it MUST drop the packet.
1253 */
1254 if (gpe->version != 0)
1255 return false;
1256 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1257 * processing MUST occur." However, we don't implement OAM
1258 * processing, thus drop the packet.
1259 */
1260 if (gpe->oam_flag)
1261 return false;
1262
1263 switch (gpe->next_protocol) {
1264 case VXLAN_GPE_NP_IPV4:
1265 *protocol = htons(ETH_P_IP);
1266 break;
1267 case VXLAN_GPE_NP_IPV6:
1268 *protocol = htons(ETH_P_IPV6);
1269 break;
1270 case VXLAN_GPE_NP_ETHERNET:
1271 *protocol = htons(ETH_P_TEB);
1272 break;
1273 default:
1274 return false;
1275 }
1276
1277 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1278 return true;
1279}
1280
1ab016e2
JB		 1281static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1282 struct vxlan_sock *vs,
3ad7a4b1 1283 struct sk_buff *skb, __be32 vni)
614732ea 1284{
614732ea 1285 union vxlan_addr saddr;
87613de9 1286 u32 ifindex = skb->dev->ifindex;
614732ea 1287
614732ea 1288 skb_reset_mac_header(skb);
614732ea
TG		 1289 skb->protocol = eth_type_trans(skb, vxlan->dev);
1290 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1291
1292 /* Ignore packet loops (and multicast echo) */
1293 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1294 return false;
614732ea 1295
760c6805 1296 /* Get address from the outer IP header */
ce212d0f 1297 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1298 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1299 saddr.sa.sa_family = AF_INET;
1300#if IS_ENABLED(CONFIG_IPV6)
1301 } else {
1ab016e2 1302 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1303 saddr.sa.sa_family = AF_INET6;
1304#endif
1305 }
1306
dc5321d7 1307 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
87613de9 1308 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1ab016e2
JB		 1309 return false;
1310
1311 return true;
1312}
1313
760c6805
JB		 1314static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1315 struct sk_buff *skb)
1316{
1317 int err = 0;
1318
1319 if (vxlan_get_sk_family(vs) == AF_INET)
1320 err = IP_ECN_decapsulate(oiph, skb);
1321#if IS_ENABLED(CONFIG_IPV6)
1322 else
1323 err = IP6_ECN_decapsulate(oiph, skb);
1324#endif
1325
1326 if (unlikely(err) && log_ecn_error) {
1327 if (vxlan_get_sk_family(vs) == AF_INET)
1328 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1329 &((struct iphdr *)oiph)->saddr,
1330 ((struct iphdr *)oiph)->tos);
1331 else
1332 net_info_ratelimited("non-ECT from %pI6\n",
1333 &((struct ipv6hdr *)oiph)->saddr);
1334 }
1335 return err <= 1;
1336}
1337
d342894c 1338/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1339static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1340{
f2d1968e 1341 struct pcpu_sw_netstats *stats;
c9e78efb 1342 struct vxlan_dev *vxlan;
5cfccc5a 1343 struct vxlan_sock *vs;
f14ecebb 1344 struct vxlanhdr unparsed;
ee122c79
TG		 1345 struct vxlan_metadata _md;
1346 struct vxlan_metadata *md = &_md;
61618eea 1347 __be16 protocol = htons(ETH_P_TEB);
e1e5314d 1348 bool raw_proto = false;
f2d1968e 1349 void *oiph;
3ad7a4b1 1350 __be32 vni = 0;
d342894c 1351
e1e5314d 1352 /* Need UDP and VXLAN header to be present */
7ce04758 1353 if (!pskb_may_pull(skb, VXLAN_HLEN))
e5aed006 1354 goto drop;
d342894c 1355
f14ecebb 1356 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1357 /* VNI flag always required to be set */
1358 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1359 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1360 ntohl(vxlan_hdr(skb)->vx_flags),
1361 ntohl(vxlan_hdr(skb)->vx_vni));
1362 /* Return non vxlan pkt */
e5aed006 1363 goto drop;
d342894c 1364 }
288b01c8
JB		 1365 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1366 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1367
559835ea 1368 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1369 if (!vs)
d342894c 1370 goto drop;
d342894c 1371
3ad7a4b1
RP		 1372 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1373
49f810f0 1374 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
c9e78efb
JB		 1375 if (!vxlan)
1376 goto drop;
1377
e1e5314d
JB		 1378 /* For backwards compatibility, only allow reserved fields to be
1379 * used by VXLAN extensions if explicitly requested.
1380 */
1381 if (vs->flags & VXLAN_F_GPE) {
1382 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1383 goto drop;
1384 raw_proto = true;
1385 }
1386
1387 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1388 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1389 goto drop;
c9e78efb 1390
ee122c79 1391 if (vxlan_collect_metadata(vs)) {
10a5af23 1392 struct metadata_dst *tun_dst;
07dabf20 1393
c29a70d2 1394 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
d817f432 1395 key32_to_tunnel_id(vni), sizeof(*md));
c29a70d2 1396
ee122c79
TG		 1397 if (!tun_dst)
1398 goto drop;
1399
0f1b7354 1400 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 1401
1402 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 1403 } else {
1404 memset(md, 0, sizeof(*md));
1405 }
1406
f14ecebb
JB		 1407 if (vs->flags & VXLAN_F_REMCSUM_RX)
1408 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1409 goto drop;
1410 if (vs->flags & VXLAN_F_GBP)
10a5af23 1411 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB		 1412 /* Note that GBP and GPE can never be active together. This is
1413 * ensured in vxlan_dev_configure.
1414 */
3511494c 1415
f14ecebb 1416 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1417 /* If there are any unprocessed flags remaining treat
1418 * this as a malformed packet. This behavior diverges from
1419 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1420 * in reserved fields are to be ignored. The approach here
c4b49512 1421 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1422 * is more robust and provides a little more security in
1423 * adding extensions to VXLAN.
1424 */
288b01c8 1425 goto drop;
3bf39475
TH		 1426 }
1427
e1e5314d 1428 if (!raw_proto) {
3ad7a4b1 1429 if (!vxlan_set_mac(vxlan, vs, skb, vni))
e1e5314d
JB		 1430 goto drop;
1431 } else {
8be0cfa4 1432 skb_reset_mac_header(skb);
e1e5314d
JB		 1433 skb->dev = vxlan->dev;
1434 skb->pkt_type = PACKET_HOST;
1435 }
f2d1968e 1436
f2d1968e
JB		 1437 oiph = skb_network_header(skb);
1438 skb_reset_network_header(skb);
1439
1440 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1441 ++vxlan->dev->stats.rx_frame_errors;
1442 ++vxlan->dev->stats.rx_errors;
1443 goto drop;
1444 }
1445
1446 stats = this_cpu_ptr(vxlan->dev->tstats);
1447 u64_stats_update_begin(&stats->syncp);
1448 stats->rx_packets++;
1449 stats->rx_bytes += skb->len;
1450 u64_stats_update_end(&stats->syncp);
1451
1452 gro_cells_receive(&vxlan->gro_cells, skb);
5cfccc5a
PS		 1453 return 0;
1454
1455drop:
288b01c8
JB		 1456 /* Consume bad packet */
1457 kfree_skb(skb);
1458 return 0;
5cfccc5a
PS		 1459}
1460
3ad7a4b1 1461static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
e4f67add
DS		 1462{
1463 struct vxlan_dev *vxlan = netdev_priv(dev);
1464 struct arphdr *parp;
1465 u8 *arpptr, *sha;
1466 __be32 sip, tip;
1467 struct neighbour *n;
1468
1469 if (dev->flags & IFF_NOARP)
1470 goto out;
1471
1472 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1473 dev->stats.tx_dropped++;
1474 goto out;
1475 }
1476 parp = arp_hdr(skb);
1477
1478 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1479 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1480 parp->ar_pro != htons(ETH_P_IP) ||
1481 parp->ar_op != htons(ARPOP_REQUEST) ||
1482 parp->ar_hln != dev->addr_len ||
1483 parp->ar_pln != 4)
1484 goto out;
1485 arpptr = (u8 *)parp + sizeof(struct arphdr);
1486 sha = arpptr;
1487 arpptr += dev->addr_len; /* sha */
1488 memcpy(&sip, arpptr, sizeof(sip));
1489 arpptr += sizeof(sip);
1490 arpptr += dev->addr_len; /* tha */
1491 memcpy(&tip, arpptr, sizeof(tip));
1492
1493 if (ipv4_is_loopback(tip) ||
1494 ipv4_is_multicast(tip))
1495 goto out;
1496
1497 n = neigh_lookup(&arp_tbl, &tip, dev);
1498
1499 if (n) {
e4f67add
DS		 1500 struct vxlan_fdb *f;
1501 struct sk_buff *reply;
1502
1503 if (!(n->nud_state & NUD_CONNECTED)) {
1504 neigh_release(n);
1505 goto out;
1506 }
1507
3ad7a4b1 1508 f = vxlan_find_mac(vxlan, n->ha, vni);
e4c7ed41 1509 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1510 /* bridge-local neighbor */
1511 neigh_release(n);
1512 goto out;
1513 }
1514
1515 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1516 n->ha, sha);
1517
1518 neigh_release(n);
1519
7346135d
DS		 1520 if (reply == NULL)
1521 goto out;
1522
e4f67add
DS		 1523 skb_reset_mac_header(reply);
1524 __skb_pull(reply, skb_network_offset(reply));
1525 reply->ip_summed = CHECKSUM_UNNECESSARY;
1526 reply->pkt_type = PACKET_HOST;
1527
1528 if (netif_rx_ni(reply) == NET_RX_DROP)
1529 dev->stats.rx_dropped++;
dc5321d7 1530 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
e4c7ed41
CW		 1531 union vxlan_addr ipa = {
1532 .sin.sin_addr.s_addr = tip,
a45e92a5 1533 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1534 };
1535
1536 vxlan_ip_miss(dev, &ipa);
1537 }
e4f67add
DS		 1538out:
1539 consume_skb(skb);
1540 return NETDEV_TX_OK;
1541}
1542
f564f45c 1543#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1544static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1545 struct neighbour *n, bool isrouter)
1546{
1547 struct net_device *dev = request->dev;
1548 struct sk_buff *reply;
1549 struct nd_msg *ns, *na;
1550 struct ipv6hdr *pip6;
1551 u8 *daddr;
1552 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1553 int ns_olen;
1554 int i, len;
1555
f1fb08f6 1556 if (dev == NULL || !pskb_may_pull(request, request->len))
4b29dba9
DS		 1557 return NULL;
1558
1559 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1560 sizeof(*na) + na_olen + dev->needed_tailroom;
1561 reply = alloc_skb(len, GFP_ATOMIC);
1562 if (reply == NULL)
1563 return NULL;
1564
1565 reply->protocol = htons(ETH_P_IPV6);
1566 reply->dev = dev;
1567 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1568 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1569 skb_reset_mac_header(reply);
4b29dba9 1570
f1fb08f6 1571 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
4b29dba9
DS		 1572
1573 daddr = eth_hdr(request)->h_source;
f1fb08f6
VB		 1574 ns_olen = request->len - skb_network_offset(request) -
1575 sizeof(struct ipv6hdr) - sizeof(*ns);
4b29dba9
DS		 1576 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1577 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1578 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1579 break;
1580 }
1581 }
1582
1583 /* Ethernet header */
1584 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1585 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1586 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1587 reply->protocol = htons(ETH_P_IPV6);
1588
1589 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1590 skb_reset_network_header(reply);
4b29dba9
DS		 1591 skb_put(reply, sizeof(struct ipv6hdr));
1592
1593 /* IPv6 header */
1594
1595 pip6 = ipv6_hdr(reply);
1596 memset(pip6, 0, sizeof(struct ipv6hdr));
1597 pip6->version = 6;
1598 pip6->priority = ipv6_hdr(request)->priority;
1599 pip6->nexthdr = IPPROTO_ICMPV6;
1600 pip6->hop_limit = 255;
1601 pip6->daddr = ipv6_hdr(request)->saddr;
1602 pip6->saddr = *(struct in6_addr *)n->primary_key;
1603
1604 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1605 skb_reset_transport_header(reply);
4b29dba9 1606
4b29dba9 1607 /* Neighbor Advertisement */
b080db58 1608 na = skb_put_zero(reply, sizeof(*na) + na_olen);
4b29dba9
DS		 1609 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1610 na->icmph.icmp6_router = isrouter;
1611 na->icmph.icmp6_override = 1;
1612 na->icmph.icmp6_solicited = 1;
1613 na->target = ns->target;
1614 ether_addr_copy(&na->opt[2], n->ha);
1615 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1616 na->opt[1] = na_olen >> 3;
1617
1618 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1619 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1620 csum_partial(na, sizeof(*na)+na_olen, 0));
1621
1622 pip6->payload_len = htons(sizeof(*na)+na_olen);
1623
1624 skb_push(reply, sizeof(struct ipv6hdr));
1625
1626 reply->ip_summed = CHECKSUM_UNNECESSARY;
1627
1628 return reply;
1629}
1630
3ad7a4b1 1631static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
f564f45c
CW		 1632{
1633 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1634 struct nd_msg *msg;
f564f45c 1635 const struct ipv6hdr *iphdr;
8dcd81a9 1636 const struct in6_addr *daddr;
4b29dba9
DS		 1637 struct neighbour *n;
1638 struct inet6_dev *in6_dev;
f564f45c
CW		 1639
1640 in6_dev = __in6_dev_get(dev);
1641 if (!in6_dev)
1642 goto out;
1643
f1fb08f6
VB		 1644 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + sizeof(struct nd_msg)))
1645 goto out;
1646
f564f45c 1647 iphdr = ipv6_hdr(skb);
f564f45c
CW		 1648 daddr = &iphdr->daddr;
1649
f1fb08f6 1650 msg = (struct nd_msg *)(iphdr + 1);
f564f45c
CW		 1651 if (msg->icmph.icmp6_code != 0 ||
1652 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1653 goto out;
1654
4b29dba9
DS		 1655 if (ipv6_addr_loopback(daddr) ||
1656 ipv6_addr_is_multicast(&msg->target))
1657 goto out;
1658
1659 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1660
1661 if (n) {
1662 struct vxlan_fdb *f;
4b29dba9 1663 struct sk_buff *reply;
f564f45c
CW		 1664
1665 if (!(n->nud_state & NUD_CONNECTED)) {
1666 neigh_release(n);
1667 goto out;
1668 }
1669
3ad7a4b1 1670 f = vxlan_find_mac(vxlan, n->ha, vni);
f564f45c
CW		 1671 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1672 /* bridge-local neighbor */
1673 neigh_release(n);
1674 goto out;
1675 }
1676
4b29dba9
DS		 1677 reply = vxlan_na_create(skb, n,
1678 !!(f ? f->flags & NTF_ROUTER : 0));
1679
f564f45c 1680 neigh_release(n);
4b29dba9
DS		 1681
1682 if (reply == NULL)
1683 goto out;
1684
1685 if (netif_rx_ni(reply) == NET_RX_DROP)
1686 dev->stats.rx_dropped++;
1687
dc5321d7 1688 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1689 union vxlan_addr ipa = {
1690 .sin6.sin6_addr = msg->target,
a45e92a5 1691 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1692 };
1693
f564f45c
CW		 1694 vxlan_ip_miss(dev, &ipa);
1695 }
1696
1697out:
1698 consume_skb(skb);
1699 return NETDEV_TX_OK;
1700}
1701#endif
1702
e4f67add
DS		 1703static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1704{
1705 struct vxlan_dev *vxlan = netdev_priv(dev);
1706 struct neighbour *n;
e4f67add
DS		 1707
1708 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1709 return false;
1710
1711 n = NULL;
1712 switch (ntohs(eth_hdr(skb)->h_proto)) {
1713 case ETH_P_IP:
e15a00aa
CW		 1714 {
1715 struct iphdr *pip;
1716
e4f67add
DS		 1717 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1718 return false;
1719 pip = ip_hdr(skb);
1720 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
dc5321d7 1721 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e4c7ed41
CW		 1722 union vxlan_addr ipa = {
1723 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1724 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1725 };
1726
1727 vxlan_ip_miss(dev, &ipa);
1728 return false;
1729 }
1730
e4f67add 1731 break;
e15a00aa
CW		 1732 }
1733#if IS_ENABLED(CONFIG_IPV6)
1734 case ETH_P_IPV6:
1735 {
1736 struct ipv6hdr *pip6;
1737
1738 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1739 return false;
1740 pip6 = ipv6_hdr(skb);
1741 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
dc5321d7 1742 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
e15a00aa
CW		 1743 union vxlan_addr ipa = {
1744 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1745 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1746 };
1747
1748 vxlan_ip_miss(dev, &ipa);
1749 return false;
1750 }
1751
1752 break;
1753 }
1754#endif
e4f67add
DS		 1755 default:
1756 return false;
1757 }
1758
1759 if (n) {
1760 bool diff;
1761
7367d0b5 1762 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1763 if (diff) {
1764 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1765 dev->addr_len);
1766 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1767 }
1768 neigh_release(n);
1769 return diff;
e4c7ed41
CW		 1770 }
1771
e4f67add
DS		 1772 return false;
1773}
1774
af33c1ad 1775static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1776 struct vxlan_metadata *md)
1777{
1778 struct vxlanhdr_gbp *gbp;
1779
db79a621
TG		 1780 if (!md->gbp)
1781 return;
1782
3511494c 1783 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1784 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 1785
1786 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1787 gbp->dont_learn = 1;
1788
1789 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1790 gbp->policy_applied = 1;
1791
1792 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1793}
1794
e1e5314d
JB		 1795static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1796 __be16 protocol)
1797{
1798 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1799
1800 gpe->np_applied = 1;
1801
1802 switch (protocol) {
1803 case htons(ETH_P_IP):
1804 gpe->next_protocol = VXLAN_GPE_NP_IPV4;
1805 return 0;
1806 case htons(ETH_P_IPV6):
1807 gpe->next_protocol = VXLAN_GPE_NP_IPV6;
1808 return 0;
1809 case htons(ETH_P_TEB):
1810 gpe->next_protocol = VXLAN_GPE_NP_ETHERNET;
1811 return 0;
1812 }
1813 return -EPFNOSUPPORT;
1814}
1815
f491e56d
JB		 1816static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1817 int iphdr_len, __be32 vni,
1818 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1819 bool udp_sum)
e4c7ed41 1820{
e4c7ed41 1821 struct vxlanhdr *vxh;
e4c7ed41
CW		 1822 int min_headroom;
1823 int err;
dfd8645e 1824 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 1825 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 1826
af33c1ad 1827 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1828 skb->ip_summed == CHECKSUM_PARTIAL) {
1829 int csum_start = skb_checksum_start_offset(skb);
1830
1831 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1832 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1833 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1834 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1835 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1836 }
e4c7ed41 1837
e4c7ed41 1838 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
4a4f86cc 1839 + VXLAN_HLEN + iphdr_len;
649c5b8b
PS		 1840
1841 /* Need space for new headers (invalidates iph ptr) */
1842 err = skb_cow_head(skb, min_headroom);
e1e5314d 1843 if (unlikely(err))
c46b7897 1844 return err;
649c5b8b 1845
aed069df
AD		 1846 err = iptunnel_handle_offloads(skb, type);
1847 if (err)
c46b7897 1848 return err;
b736a623 1849
d58ff351 1850 vxh = __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 1851 vxh->vx_flags = VXLAN_HF_VNI;
1852 vxh->vx_vni = vxlan_vni_field(vni);
49560532 1853
dfd8645e 1854 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 1855 unsigned int start;
dfd8645e 1856
54bfd872
JB		 1857 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1858 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1859 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 1860
1861 if (!skb_is_gso(skb)) {
1862 skb->ip_summed = CHECKSUM_NONE;
1863 skb->encapsulation = 0;
1864 }
1865 }
1866
af33c1ad
TH		 1867 if (vxflags & VXLAN_F_GBP)
1868 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB		 1869 if (vxflags & VXLAN_F_GPE) {
1870 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1871 if (err < 0)
c46b7897 1872 return err;
e1e5314d
JB		 1873 inner_protocol = skb->protocol;
1874 }
3511494c 1875
e1e5314d 1876 skb_set_inner_protocol(skb, inner_protocol);
039f5062 1877 return 0;
49560532 1878}
49560532 1879
655c3de1 1880static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
1881 struct vxlan_sock *sock4,
1a8496ba 1882 struct sk_buff *skb, int oif, u8 tos,
4ecb1d83 1883 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
0c1d70af 1884 struct dst_cache *dst_cache,
db3c6139 1885 const struct ip_tunnel_info *info)
1a8496ba 1886{
db3c6139 1887 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 1888 struct rtable *rt = NULL;
1889 struct flowi4 fl4;
1890
655c3de1 1891 if (!sock4)
1892 return ERR_PTR(-EIO);
1893
db3c6139
DB		 1894 if (tos && !info)
1895 use_cache = false;
1896 if (use_cache) {
0c1d70af
PA		 1897 rt = dst_cache_get_ip4(dst_cache, saddr);
1898 if (rt)
1899 return rt;
1900 }
1901
1a8496ba
JB		 1902 memset(&fl4, 0, sizeof(fl4));
1903 fl4.flowi4_oif = oif;
1904 fl4.flowi4_tos = RT_TOS(tos);
1905 fl4.flowi4_mark = skb->mark;
1906 fl4.flowi4_proto = IPPROTO_UDP;
1907 fl4.daddr = daddr;
272d96a5 1908 fl4.saddr = *saddr;
4ecb1d83
MP		 1909 fl4.fl4_dport = dport;
1910 fl4.fl4_sport = sport;
1a8496ba
JB		 1911
1912 rt = ip_route_output_key(vxlan->net, &fl4);
655c3de1 1913 if (likely(!IS_ERR(rt))) {
1914 if (rt->dst.dev == dev) {
1915 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
1916 ip_rt_put(rt);
1917 return ERR_PTR(-ELOOP);
1918 }
1919
1a8496ba 1920 *saddr = fl4.saddr;
0c1d70af
PA		 1921 if (use_cache)
1922 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
655c3de1 1923 } else {
1924 netdev_dbg(dev, "no route to %pI4\n", &daddr);
1925 return ERR_PTR(-ENETUNREACH);
0c1d70af 1926 }
1a8496ba
JB		 1927 return rt;
1928}
1929
e5d4b29f
JB		 1930#if IS_ENABLED(CONFIG_IPV6)
1931static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
655c3de1 1932 struct net_device *dev,
03dc52a8 1933 struct vxlan_sock *sock6,
1400615d 1934 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 1935 __be32 label,
e5d4b29f 1936 const struct in6_addr *daddr,
0c1d70af 1937 struct in6_addr *saddr,
4ecb1d83 1938 __be16 dport, __be16 sport,
db3c6139
DB		 1939 struct dst_cache *dst_cache,
1940 const struct ip_tunnel_info *info)
e5d4b29f 1941{
db3c6139 1942 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 1943 struct dst_entry *ndst;
1944 struct flowi6 fl6;
1945 int err;
1946
c6fcc4fc 1947 if (!sock6)
1948 return ERR_PTR(-EIO);
1949
1400615d
DB		 1950 if (tos && !info)
1951 use_cache = false;
db3c6139 1952 if (use_cache) {
0c1d70af
PA		 1953 ndst = dst_cache_get_ip6(dst_cache, saddr);
1954 if (ndst)
1955 return ndst;
1956 }
1957
e5d4b29f
JB		 1958 memset(&fl6, 0, sizeof(fl6));
1959 fl6.flowi6_oif = oif;
1960 fl6.daddr = *daddr;
272d96a5 1961 fl6.saddr = *saddr;
eaa93bf4 1962 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB		 1963 fl6.flowi6_mark = skb->mark;
1964 fl6.flowi6_proto = IPPROTO_UDP;
4ecb1d83
MP		 1965 fl6.fl6_dport = dport;
1966 fl6.fl6_sport = sport;
e5d4b29f
JB		 1967
1968 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
c6fcc4fc 1969 sock6->sock->sk,
e5d4b29f 1970 &ndst, &fl6);
655c3de1 1971 if (unlikely(err < 0)) {
1972 netdev_dbg(dev, "no route to %pI6\n", daddr);
1973 return ERR_PTR(-ENETUNREACH);
1974 }
1975
1976 if (unlikely(ndst->dev == dev)) {
1977 netdev_dbg(dev, "circular route to %pI6\n", daddr);
1978 dst_release(ndst);
1979 return ERR_PTR(-ELOOP);
1980 }
e5d4b29f
JB		 1981
1982 *saddr = fl6.saddr;
db3c6139 1983 if (use_cache)
0c1d70af 1984 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 1985 return ndst;
1986}
1987#endif
1988
9dcc71e1
SS		 1989/* Bypass encapsulation if the destination is local */
1990static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
3ad7a4b1 1991 struct vxlan_dev *dst_vxlan, __be32 vni)
9dcc71e1 1992{
8f84985f 1993 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1994 union vxlan_addr loopback;
1995 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1996 struct net_device *dev = skb->dev;
1997 int len = skb->len;
9dcc71e1 1998
8f84985f
LR		 1999 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2000 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 2001 skb->pkt_type = PACKET_HOST;
2002 skb->encapsulation = 0;
2003 skb->dev = dst_vxlan->dev;
2004 __skb_pull(skb, skb_network_offset(skb));
2005
e4c7ed41
CW		 2006 if (remote_ip->sa.sa_family == AF_INET) {
2007 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2008 loopback.sa.sa_family = AF_INET;
2009#if IS_ENABLED(CONFIG_IPV6)
2010 } else {
2011 loopback.sin6.sin6_addr = in6addr_loopback;
2012 loopback.sa.sa_family = AF_INET6;
2013#endif
2014 }
2015
dc5321d7 2016 if (dst_vxlan->cfg.flags & VXLAN_F_LEARN)
87613de9
MS		 2017 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source, 0,
2018 vni);
9dcc71e1
SS		 2019
2020 u64_stats_update_begin(&tx_stats->syncp);
2021 tx_stats->tx_packets++;
ce6502a8 2022 tx_stats->tx_bytes += len;
9dcc71e1
SS		 2023 u64_stats_update_end(&tx_stats->syncp);
2024
2025 if (netif_rx(skb) == NET_RX_SUCCESS) {
2026 u64_stats_update_begin(&rx_stats->syncp);
2027 rx_stats->rx_packets++;
ce6502a8 2028 rx_stats->rx_bytes += len;
9dcc71e1
SS		 2029 u64_stats_update_end(&rx_stats->syncp);
2030 } else {
ce6502a8 2031 dev->stats.rx_dropped++;
9dcc71e1
SS		 2032 }
2033}
2034
fee1fad7 2035static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
49f810f0
MS		 2036 struct vxlan_dev *vxlan,
2037 union vxlan_addr *daddr,
2038 __be16 dst_port, int dst_ifindex, __be32 vni,
2039 struct dst_entry *dst,
fee1fad7 2040 u32 rt_flags)
2041{
2042#if IS_ENABLED(CONFIG_IPV6)
2043 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2044 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2045 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2046 */
2047 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2048#endif
2049 /* Bypass encapsulation if the destination is local */
2050 if (rt_flags & RTCF_LOCAL &&
2051 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2052 struct vxlan_dev *dst_vxlan;
2053
2054 dst_release(dst);
49f810f0 2055 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
fee1fad7 2056 daddr->sa.sa_family, dst_port,
dc5321d7 2057 vxlan->cfg.flags);
fee1fad7 2058 if (!dst_vxlan) {
2059 dev->stats.tx_errors++;
2060 kfree_skb(skb);
2061
2062 return -ENOENT;
2063 }
3ad7a4b1 2064 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
fee1fad7 2065 return 1;
2066 }
2067
2068 return 0;
2069}
2070
4ad16930 2071static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
3ad7a4b1
RP		 2072 __be32 default_vni, struct vxlan_rdst *rdst,
2073 bool did_rsc)
d342894c 2074{
d71785ff 2075 struct dst_cache *dst_cache;
3093fbe7 2076 struct ip_tunnel_info *info;
d342894c 2077 struct vxlan_dev *vxlan = netdev_priv(dev);
0770b53b 2078 const struct iphdr *old_iph = ip_hdr(skb);
e4c7ed41 2079 union vxlan_addr *dst;
272d96a5 2080 union vxlan_addr remote_ip, local_ip;
ee122c79
TG		 2081 struct vxlan_metadata _md;
2082 struct vxlan_metadata *md = &_md;
e4c7ed41 2083 __be16 src_port = 0, dst_port;
655c3de1 2084 struct dst_entry *ndst = NULL;
e7f70af1 2085 __be32 vni, label;
d342894c 2086 __u8 tos, ttl;
49f810f0 2087 int ifindex;
0e6fbc5b 2088 int err;
dc5321d7 2089 u32 flags = vxlan->cfg.flags;
b4ed5cad 2090 bool udp_sum = false;
f491e56d 2091 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 2092
61adedf3 2093 info = skb_tunnel_info(skb);
3093fbe7 2094
ee122c79 2095 if (rdst) {
0770b53b 2096 dst = &rdst->remote_ip;
2097 if (vxlan_addr_any(dst)) {
2098 if (did_rsc) {
2099 /* short-circuited back to local bridge */
3ad7a4b1 2100 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
0770b53b 2101 return;
2102 }
2103 goto drop;
2104 }
2105
0dfbdf41 2106 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
3ad7a4b1 2107 vni = (rdst->remote_vni) ? : default_vni;
49f810f0 2108 ifindex = rdst->remote_ifindex;
1158632b 2109 local_ip = vxlan->cfg.saddr;
d71785ff 2110 dst_cache = &rdst->dst_cache;
0770b53b 2111 md->gbp = skb->mark;
2112 ttl = vxlan->cfg.ttl;
2113 if (!ttl && vxlan_addr_multicast(dst))
2114 ttl = 1;
2115
2116 tos = vxlan->cfg.tos;
2117 if (tos == 1)
2118 tos = ip_tunnel_get_dsfield(old_iph, skb);
2119
2120 if (dst->sa.sa_family == AF_INET)
2121 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2122 else
2123 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2124 label = vxlan->cfg.label;
ee122c79
TG		 2125 } else {
2126 if (!info) {
2127 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2128 dev->name);
2129 goto drop;
2130 }
b1be00a6 2131 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
272d96a5 2132 if (remote_ip.sa.sa_family == AF_INET) {
a725e514 2133 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
272d96a5 2134 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2135 } else {
a725e514 2136 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
272d96a5 2137 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2138 }
ee122c79 2139 dst = &remote_ip;
0770b53b 2140 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2141 vni = tunnel_id_to_key32(info->key.tun_id);
49f810f0 2142 ifindex = 0;
d71785ff 2143 dst_cache = &info->dst_cache;
0770b53b 2144 if (info->options_len)
2145 md = ip_tunnel_info_opts(info);
a725e514
JB		 2146 ttl = info->key.ttl;
2147 tos = info->key.tos;
e7f70af1 2148 label = info->key.label;
b4ed5cad 2149 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514 2150 }
0770b53b 2151 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2152 vxlan->cfg.port_max, true);
a725e514 2153
56de859e 2154 rcu_read_lock();
e4c7ed41 2155 if (dst->sa.sa_family == AF_INET) {
c6fcc4fc 2156 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
c46b7897 2157 struct rtable *rt;
0770b53b 2158 __be16 df = 0;
c6fcc4fc 2159
49f810f0 2160 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
272d96a5 2161 dst->sin.sin_addr.s_addr,
1158632b 2162 &local_ip.sin.sin_addr.s_addr,
4ecb1d83 2163 dst_port, src_port,
d71785ff 2164 dst_cache, info);
8ebd115b
DM		 2165 if (IS_ERR(rt)) {
2166 err = PTR_ERR(rt);
c46b7897 2167 goto tx_error;
8ebd115b 2168 }
e4c7ed41
CW		 2169
2170 /* Bypass encapsulation if the destination is local */
fee1fad7 2171 if (!info) {
2172 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2173 dst_port, ifindex, vni,
2174 &rt->dst, rt->rt_flags);
fee1fad7 2175 if (err)
56de859e 2176 goto out_unlock;
fee1fad7 2177 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
6ceb31ca 2178 df = htons(IP_DF);
fee1fad7 2179 }
6ceb31ca 2180
c46b7897 2181 ndst = &rt->dst;
e4c7ed41
CW		 2182 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2183 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
c46b7897 2184 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
54bfd872 2185 vni, md, flags, udp_sum);
f491e56d 2186 if (err < 0)
c46b7897 2187 goto tx_error;
f491e56d 2188
1158632b 2189 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
f491e56d
JB		 2190 dst->sin.sin_addr.s_addr, tos, ttl, df,
2191 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2192#if IS_ENABLED(CONFIG_IPV6)
2193 } else {
c6fcc4fc 2194 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
e4c7ed41 2195
49f810f0 2196 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
272d96a5 2197 label, &dst->sin6.sin6_addr,
1158632b 2198 &local_ip.sin6.sin6_addr,
4ecb1d83 2199 dst_port, src_port,
db3c6139 2200 dst_cache, info);
e5d4b29f 2201 if (IS_ERR(ndst)) {
8ebd115b 2202 err = PTR_ERR(ndst);
c46b7897 2203 ndst = NULL;
9dcc71e1 2204 goto tx_error;
e4c7ed41 2205 }
655c3de1 2206
fee1fad7 2207 if (!info) {
2208 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
49560532 2209
fee1fad7 2210 err = encap_bypass_if_local(skb, dev, vxlan, dst,
49f810f0
MS		 2211 dst_port, ifindex, vni,
2212 ndst, rt6i_flags);
fee1fad7 2213 if (err)
56de859e 2214 goto out_unlock;
fee1fad7 2215 }
35e2d115 2216
1400615d 2217 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2218 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2219 skb_scrub_packet(skb, xnet);
2220 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2221 vni, md, flags, udp_sum);
c46b7897 2222 if (err < 0)
2223 goto tx_error;
2224
0770b53b 2225 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
1158632b 2226 &local_ip.sin6.sin6_addr,
272d96a5 2227 &dst->sin6.sin6_addr, tos, ttl,
e7f70af1 2228 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2229#endif
2230 }
56de859e
JK		 2231out_unlock:
2232 rcu_read_unlock();
4ad16930 2233 return;
d342894c 2234
2235drop:
2236 dev->stats.tx_dropped++;
c46b7897 2237 dev_kfree_skb(skb);
2238 return;
d342894c 2239
2240tx_error:
56de859e 2241 rcu_read_unlock();
655c3de1 2242 if (err == -ELOOP)
2243 dev->stats.collisions++;
2244 else if (err == -ENETUNREACH)
2245 dev->stats.tx_carrier_errors++;
c46b7897 2246 dst_release(ndst);
d342894c 2247 dev->stats.tx_errors++;
c46b7897 2248 kfree_skb(skb);
d342894c 2249}
2250
6681712d
DS		 2251/* Transmit local packets over Vxlan
2252 *
2253 * Outer IP header inherits ECN and DF from inner header.
2254 * Outer UDP destination is the VXLAN assigned port.
2255 * source port is based on hash of flow
2256 */
2257static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2258{
2259 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2260 const struct ip_tunnel_info *info;
6681712d
DS		 2261 struct ethhdr *eth;
2262 bool did_rsc = false;
8f646c92 2263 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2264 struct vxlan_fdb *f;
3ad7a4b1 2265 __be32 vni = 0;
6681712d 2266
61adedf3 2267 info = skb_tunnel_info(skb);
3093fbe7 2268
6681712d 2269 skb_reset_mac_header(skb);
6681712d 2270
dc5321d7 2271 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 2272 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2273 info->mode & IP_TUNNEL_INFO_TX) {
2274 vni = tunnel_id_to_key32(info->key.tun_id);
2275 } else {
2276 if (info && info->mode & IP_TUNNEL_INFO_TX)
2277 vxlan_xmit_one(skb, dev, vni, NULL, false);
2278 else
2279 kfree_skb(skb);
2280 return NETDEV_TX_OK;
2281 }
47e5d1b0
JB		 2282 }
2283
dc5321d7 2284 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
47e5d1b0 2285 eth = eth_hdr(skb);
f564f45c 2286 if (ntohs(eth->h_proto) == ETH_P_ARP)
3ad7a4b1 2287 return arp_reduce(dev, skb, vni);
f564f45c 2288#if IS_ENABLED(CONFIG_IPV6)
f1fb08f6
VB		 2289 else if (ntohs(eth->h_proto) == ETH_P_IPV6) {
2290 struct ipv6hdr *hdr, _hdr;
2291 if ((hdr = skb_header_pointer(skb,
2292 skb_network_offset(skb),
2293 sizeof(_hdr), &_hdr)) &&
2294 hdr->nexthdr == IPPROTO_ICMPV6)
2295 return neigh_reduce(dev, skb, vni);
f564f45c
CW		 2296 }
2297#endif
2298 }
6681712d 2299
47e5d1b0 2300 eth = eth_hdr(skb);
3ad7a4b1 2301 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2302 did_rsc = false;
2303
dc5321d7 2304 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2305 (ntohs(eth->h_proto) == ETH_P_IP ||
2306 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2307 did_rsc = route_shortcircuit(dev, skb);
2308 if (did_rsc)
3ad7a4b1 2309 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2310 }
2311
6681712d 2312 if (f == NULL) {
3ad7a4b1 2313 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae 2314 if (f == NULL) {
dc5321d7 2315 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
afbd8bae
MR		 2316 !is_multicast_ether_addr(eth->h_dest))
2317 vxlan_fdb_miss(vxlan, eth->h_dest);
2318
2319 dev->stats.tx_dropped++;
8f646c92 2320 kfree_skb(skb);
afbd8bae
MR		 2321 return NETDEV_TX_OK;
2322 }
2323 }
6681712d 2324
afbd8bae
MR		 2325 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2326 struct sk_buff *skb1;
6681712d 2327
8f646c92
ED		 2328 if (!fdst) {
2329 fdst = rdst;
2330 continue;
2331 }
afbd8bae
MR		 2332 skb1 = skb_clone(skb, GFP_ATOMIC);
2333 if (skb1)
3ad7a4b1 2334 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
6681712d
DS		 2335 }
2336
8f646c92 2337 if (fdst)
3ad7a4b1 2338 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
8f646c92
ED		 2339 else
2340 kfree_skb(skb);
4ad16930 2341 return NETDEV_TX_OK;
6681712d
DS		 2342}
2343
d342894c 2344/* Walk the forwarding table and purge stale entries */
2345static void vxlan_cleanup(unsigned long arg)
2346{
2347 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2348 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2349 unsigned int h;
2350
2351 if (!netif_running(vxlan->dev))
2352 return;
2353
d342894c 2354 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2355 struct hlist_node *p, *n;
14e1d0fa
SD		 2356
2357 spin_lock_bh(&vxlan->hash_lock);
d342894c 2358 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2359 struct vxlan_fdb *f
2360 = container_of(p, struct vxlan_fdb, hlist);
2361 unsigned long timeout;
2362
efb5f68f 2363 if (f->state & (NUD_PERMANENT | NUD_NOARP))
d342894c 2364 continue;
2365
def499c9
RP		 2366 if (f->flags & NTF_EXT_LEARNED)
2367 continue;
2368
0dfbdf41 2369 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2370 if (time_before_eq(timeout, jiffies)) {
2371 netdev_dbg(vxlan->dev,
2372 "garbage collect %pM\n",
2373 f->eth_addr);
2374 f->state = NUD_STALE;
2375 vxlan_fdb_destroy(vxlan, f);
2376 } else if (time_before(timeout, next_timer))
2377 next_timer = timeout;
2378 }
14e1d0fa 2379 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2380 }
d342894c 2381
2382 mod_timer(&vxlan->age_timer, next_timer);
2383}
2384
a53cb29b
MB		 2385static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2386{
2387 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2388
2389 spin_lock(&vn->sock_lock);
2390 hlist_del_init_rcu(&vxlan->hlist);
2391 spin_unlock(&vn->sock_lock);
2392}
2393
9c2e24e1
PS		 2394static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2395{
56ef9c90 2396 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2397 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2398
56ef9c90 2399 spin_lock(&vn->sock_lock);
9c2e24e1 2400 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2401 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2402}
2403
d342894c 2404/* Setup stats when device is created */
2405static int vxlan_init(struct net_device *dev)
2406{
1c213bd2 2407 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2408 if (!dev->tstats)
d342894c 2409 return -ENOMEM;
2410
2411 return 0;
2412}
2413
3ad7a4b1 2414static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
afbd8bae
MR		 2415{
2416 struct vxlan_fdb *f;
2417
2418 spin_lock_bh(&vxlan->hash_lock);
3ad7a4b1 2419 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae
MR		 2420 if (f)
2421 vxlan_fdb_destroy(vxlan, f);
2422 spin_unlock_bh(&vxlan->hash_lock);
2423}
2424
ebf4063e
SH		 2425static void vxlan_uninit(struct net_device *dev)
2426{
2427 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2428
3ad7a4b1 2429 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
afbd8bae 2430
ebf4063e
SH		 2431 free_percpu(dev->tstats);
2432}
2433
d342894c 2434/* Start ageing timer and join group when device is brought up */
2435static int vxlan_open(struct net_device *dev)
2436{
2437 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2438 int ret;
56ef9c90 2439
205f356d
JB		 2440 ret = vxlan_sock_add(vxlan);
2441 if (ret < 0)
2442 return ret;
d342894c 2443
79d4a94f 2444 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2445 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2446 if (ret == -EADDRINUSE)
2447 ret = 0;
56ef9c90 2448 if (ret) {
205f356d 2449 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2450 return ret;
2451 }
d342894c 2452 }
2453
0dfbdf41 2454 if (vxlan->cfg.age_interval)
d342894c 2455 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2456
56ef9c90 2457 return ret;
d342894c 2458}
2459
2460/* Purge the forwarding table */
8b3f9337 2461static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
d342894c 2462{
31fec5aa 2463 unsigned int h;
d342894c 2464
2465 spin_lock_bh(&vxlan->hash_lock);
2466 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2467 struct hlist_node *p, *n;
2468 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2469 struct vxlan_fdb *f
2470 = container_of(p, struct vxlan_fdb, hlist);
8b3f9337
RP		 2471 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2472 continue;
afbd8bae
MR		 2473 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2474 if (!is_zero_ether_addr(f->eth_addr))
2475 vxlan_fdb_destroy(vxlan, f);
d342894c 2476 }
2477 }
2478 spin_unlock_bh(&vxlan->hash_lock);
2479}
2480
2481/* Cleanup timer and forwarding table on shutdown */
2482static int vxlan_stop(struct net_device *dev)
2483{
2484 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2485 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2486 int ret = 0;
d342894c 2487
24c0e683 2488 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2489 !vxlan_group_used(vn, vxlan))
56ef9c90 2490 ret = vxlan_igmp_leave(vxlan);
d342894c 2491
2492 del_timer_sync(&vxlan->age_timer);
2493
8b3f9337 2494 vxlan_flush(vxlan, false);
205f356d 2495 vxlan_sock_release(vxlan);
d342894c 2496
56ef9c90 2497 return ret;
d342894c 2498}
2499
d342894c 2500/* Stub, nothing needs to be done. */
2501static void vxlan_set_multicast_list(struct net_device *dev)
2502{
2503}
2504
91572088 2505static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
345010b5 2506{
91572088
JW		 2507 struct vxlan_dev *vxlan = netdev_priv(dev);
2508 struct vxlan_rdst *dst = &vxlan->default_dst;
2509 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2510 dst->remote_ifindex);
ce44a4ae 2511 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
345010b5 2512
91572088
JW		 2513 /* This check is different than dev->max_mtu, because it looks at
2514 * the lowerdev->mtu, rather than the static dev->max_mtu
2515 */
2516 if (lowerdev) {
2517 int max_mtu = lowerdev->mtu -
2518 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2519 if (new_mtu > max_mtu)
72564b59 2520 return -EINVAL;
72564b59
DW		 2521 }
2522
345010b5
DB		 2523 dev->mtu = new_mtu;
2524 return 0;
2525}
2526
fc4099f1
PS		 2527static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2528{
2529 struct vxlan_dev *vxlan = netdev_priv(dev);
2530 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2531 __be16 sport, dport;
2532
2533 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2534 vxlan->cfg.port_max, true);
2535 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2536
239e944f 2537 if (ip_tunnel_info_af(info) == AF_INET) {
c6fcc4fc 2538 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
1a8496ba
JB		 2539 struct rtable *rt;
2540
655c3de1 2541 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
1a8496ba 2542 info->key.u.ipv4.dst,
22f0708a
PA		 2543 &info->key.u.ipv4.src, dport, sport,
2544 &info->dst_cache, info);
1a8496ba
JB		 2545 if (IS_ERR(rt))
2546 return PTR_ERR(rt);
2547 ip_rt_put(rt);
239e944f
JB		 2548 } else {
2549#if IS_ENABLED(CONFIG_IPV6)
03dc52a8 2550 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
239e944f
JB		 2551 struct dst_entry *ndst;
2552
655c3de1 2553 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
e7f70af1 2554 info->key.label, &info->key.u.ipv6.dst,
22f0708a
PA		 2555 &info->key.u.ipv6.src, dport, sport,
2556 &info->dst_cache, info);
239e944f
JB		 2557 if (IS_ERR(ndst))
2558 return PTR_ERR(ndst);
2559 dst_release(ndst);
239e944f
JB		 2560#else /* !CONFIG_IPV6 */
2561 return -EPFNOSUPPORT;
2562#endif
2563 }
1a8496ba
JB		 2564 info->key.tp_src = sport;
2565 info->key.tp_dst = dport;
239e944f 2566 return 0;
fc4099f1
PS		 2567}
2568
0c867c9b 2569static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 2570 .ndo_init = vxlan_init,
ebf4063e 2571 .ndo_uninit = vxlan_uninit,
d342894c 2572 .ndo_open = vxlan_open,
2573 .ndo_stop = vxlan_stop,
2574 .ndo_start_xmit = vxlan_xmit,
e8171045 2575 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2576 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2577 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2578 .ndo_validate_addr = eth_validate_addr,
2579 .ndo_set_mac_address = eth_mac_addr,
2580 .ndo_fdb_add = vxlan_fdb_add,
2581 .ndo_fdb_del = vxlan_fdb_delete,
2582 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2583 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2584};
2585
e1e5314d
JB		 2586static const struct net_device_ops vxlan_netdev_raw_ops = {
2587 .ndo_init = vxlan_init,
2588 .ndo_uninit = vxlan_uninit,
2589 .ndo_open = vxlan_open,
2590 .ndo_stop = vxlan_stop,
2591 .ndo_start_xmit = vxlan_xmit,
2592 .ndo_get_stats64 = ip_tunnel_get_stats64,
2593 .ndo_change_mtu = vxlan_change_mtu,
2594 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2595};
2596
d342894c 2597/* Info for udev, that this is a virtual tunnel endpoint */
2598static struct device_type vxlan_type = {
2599 .name = "vxlan",
2600};
2601
e5de25dc 2602/* Calls the ndo_udp_tunnel_add of the caller in order to
35e42379 2603 * supply the listening VXLAN udp ports. Callers are expected
e5de25dc 2604 * to implement the ndo_udp_tunnel_add.
53cf5275 2605 */
b7aade15 2606static void vxlan_push_rx_ports(struct net_device *dev)
53cf5275
JG		 2607{
2608 struct vxlan_sock *vs;
2609 struct net *net = dev_net(dev);
2610 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
35e42379 2611 unsigned int i;
53cf5275
JG		 2612
2613 spin_lock(&vn->sock_lock);
2614 for (i = 0; i < PORT_HASH_SIZE; ++i) {
e7b3db5e
AD		 2615 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist)
2616 udp_tunnel_push_rx_port(dev, vs->sock,
b9adcd69
AD		 2617 (vs->flags & VXLAN_F_GPE) ?
2618 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 2619 UDP_TUNNEL_TYPE_VXLAN);
53cf5275
JG		 2620 }
2621 spin_unlock(&vn->sock_lock);
2622}
53cf5275 2623
d342894c 2624/* Initialize the device structure. */
2625static void vxlan_setup(struct net_device *dev)
2626{
2627 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2628 unsigned int h;
d342894c 2629
65226ef8
JB		 2630 eth_hw_addr_random(dev);
2631 ether_setup(dev);
2632
cf124db5 2633 dev->needs_free_netdev = true;
d342894c 2634 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2635
d342894c 2636 dev->features |= NETIF_F_LLTX;
d6727fe3 2637 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2638 dev->features |= NETIF_F_RXCSUM;
05c0db08 2639 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2640
1eaa8178 2641 dev->vlan_features = dev->features;
0afb1666 2642 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2643 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
02875878 2644 netif_keep_dst(dev);
0c867c9b 2645 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 2646
a985343b
MS		 2647 /* MTU range: 68 - 65535 */
2648 dev->min_mtu = ETH_MIN_MTU;
2649 dev->max_mtu = ETH_MAX_MTU;
2650
553675fb 2651 INIT_LIST_HEAD(&vxlan->next);
d342894c 2652 spin_lock_init(&vxlan->hash_lock);
2653
2654 init_timer_deferrable(&vxlan->age_timer);
2655 vxlan->age_timer.function = vxlan_cleanup;
2656 vxlan->age_timer.data = (unsigned long) vxlan;
2657
2658 vxlan->dev = dev;
a985343b 2659 vxlan->net = dev_net(dev);
d342894c 2660
58ce31cc
TH		 2661 gro_cells_init(&vxlan->gro_cells, dev);
2662
d342894c 2663 for (h = 0; h < FDB_HASH_SIZE; ++h)
2664 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2665}
2666
0c867c9b
JB		 2667static void vxlan_ether_setup(struct net_device *dev)
2668{
0c867c9b
JB		 2669 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2670 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2671 dev->netdev_ops = &vxlan_netdev_ether_ops;
2672}
2673
e1e5314d
JB		 2674static void vxlan_raw_setup(struct net_device *dev)
2675{
65226ef8 2676 dev->header_ops = NULL;
e1e5314d
JB		 2677 dev->type = ARPHRD_NONE;
2678 dev->hard_header_len = 0;
2679 dev->addr_len = 0;
e1e5314d
JB		 2680 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2681 dev->netdev_ops = &vxlan_netdev_raw_ops;
2682}
2683
d342894c 2684static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2685 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2686 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2687 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2688 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2689 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2690 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2691 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2692 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 2693 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 2694 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2695 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2696 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2697 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2698 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2699 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2700 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2701 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2702 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2703 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2704 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2705 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2706 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2707 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2708 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2709 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 2710 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 2711 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2712};
2713
a8b8a889
MS		 2714static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
2715 struct netlink_ext_ack *extack)
d342894c 2716{
2717 if (tb[IFLA_ADDRESS]) {
2718 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2719 pr_debug("invalid link address (not ethernet)\n");
2720 return -EINVAL;
2721 }
2722
2723 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2724 pr_debug("invalid all zero ethernet address\n");
2725 return -EADDRNOTAVAIL;
2726 }
2727 }
2728
a985343b 2729 if (tb[IFLA_MTU]) {
019b13ae 2730 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
a985343b
MS		 2731
2732 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU)
2733 return -EINVAL;
2734 }
2735
d342894c 2736 if (!data)
2737 return -EINVAL;
2738
2739 if (data[IFLA_VXLAN_ID]) {
a985343b
MS		 2740 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2741
4e37d691 2742 if (id >= VXLAN_N_VID)
d342894c 2743 return -ERANGE;
2744 }
2745
05f47d69 2746 if (data[IFLA_VXLAN_PORT_RANGE]) {
2747 const struct ifla_vxlan_port_range *p
2748 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2749
2750 if (ntohs(p->high) < ntohs(p->low)) {
2751 pr_debug("port range %u .. %u not valid\n",
2752 ntohs(p->low), ntohs(p->high));
2753 return -EINVAL;
2754 }
2755 }
2756
d342894c 2757 return 0;
2758}
2759
1b13c97f
YB		 2760static void vxlan_get_drvinfo(struct net_device *netdev,
2761 struct ethtool_drvinfo *drvinfo)
2762{
2763 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2764 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2765}
2766
2767static const struct ethtool_ops vxlan_ethtool_ops = {
2768 .get_drvinfo = vxlan_get_drvinfo,
2769 .get_link = ethtool_op_get_link,
2770};
2771
3ee64f39
TH		 2772static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2773 __be16 port, u32 flags)
553675fb 2774{
e4c7ed41 2775 struct socket *sock;
3ee64f39
TH		 2776 struct udp_port_cfg udp_conf;
2777 int err;
e4c7ed41 2778
3ee64f39 2779 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2780
3ee64f39
TH		 2781 if (ipv6) {
2782 udp_conf.family = AF_INET6;
3ee64f39 2783 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2784 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2785 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2786 } else {
2787 udp_conf.family = AF_INET;
e4c7ed41 2788 }
e4c7ed41 2789
3ee64f39 2790 udp_conf.local_udp_port = port;
553675fb 2791
3ee64f39
TH		 2792 /* Open UDP socket */
2793 err = udp_sock_create(net, &udp_conf, &sock);
2794 if (err < 0)
2795 return ERR_PTR(err);
e4c7ed41 2796
39deb2c7 2797 return sock;
e4c7ed41
CW		 2798}
2799
2800/* Create new listen socket if needed */
b1be00a6
JB		 2801static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2802 __be16 port, u32 flags)
e4c7ed41
CW		 2803{
2804 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2805 struct vxlan_sock *vs;
2806 struct socket *sock;
e4c7ed41 2807 unsigned int h;
acbf74a7 2808 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2809
dc01e7d3 2810 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2811 if (!vs)
2812 return ERR_PTR(-ENOMEM);
2813
2814 for (h = 0; h < VNI_HASH_SIZE; ++h)
2815 INIT_HLIST_HEAD(&vs->vni_list[h]);
2816
3ee64f39 2817 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2818 if (IS_ERR(sock)) {
553675fb 2819 kfree(vs);
e50fddc8 2820 return ERR_CAST(sock);
553675fb 2821 }
e4c7ed41
CW		 2822
2823 vs->sock = sock;
9c2e24e1 2824 atomic_set(&vs->refcnt, 1);
af33c1ad 2825 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2826
9c2e24e1
PS		 2827 spin_lock(&vn->sock_lock);
2828 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
e7b3db5e 2829 udp_tunnel_notify_add_rx_port(sock,
b9adcd69
AD		 2830 (vs->flags & VXLAN_F_GPE) ?
2831 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 2832 UDP_TUNNEL_TYPE_VXLAN);
9c2e24e1 2833 spin_unlock(&vn->sock_lock);
553675fb 2834
2835 /* Mark socket as an encapsulation socket. */
5602c48c 2836 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ		 2837 tunnel_cfg.sk_user_data = vs;
2838 tunnel_cfg.encap_type = 1;
f2d1968e 2839 tunnel_cfg.encap_rcv = vxlan_rcv;
acbf74a7 2840 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH		 2841 tunnel_cfg.gro_receive = vxlan_gro_receive;
2842 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ		 2843
2844 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2845
9c2e24e1
PS		 2846 return vs;
2847}
2848
b1be00a6 2849static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2850{
205f356d
JB		 2851 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2852 struct vxlan_sock *vs = NULL;
9c2e24e1 2853
205f356d 2854 if (!vxlan->cfg.no_share) {
56ef9c90 2855 spin_lock(&vn->sock_lock);
205f356d 2856 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
dc5321d7 2857 vxlan->cfg.dst_port, vxlan->cfg.flags);
205f356d 2858 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
56ef9c90 2859 spin_unlock(&vn->sock_lock);
205f356d 2860 return -EBUSY;
56ef9c90
MRL		 2861 }
2862 spin_unlock(&vn->sock_lock);
2863 }
205f356d 2864 if (!vs)
b1be00a6 2865 vs = vxlan_socket_create(vxlan->net, ipv6,
dc5321d7 2866 vxlan->cfg.dst_port, vxlan->cfg.flags);
205f356d
JB		 2867 if (IS_ERR(vs))
2868 return PTR_ERR(vs);
b1be00a6
JB		 2869#if IS_ENABLED(CONFIG_IPV6)
2870 if (ipv6)
c6fcc4fc 2871 rcu_assign_pointer(vxlan->vn6_sock, vs);
b1be00a6
JB		 2872 else
2873#endif
c6fcc4fc 2874 rcu_assign_pointer(vxlan->vn4_sock, vs);
205f356d
JB		 2875 vxlan_vs_add_dev(vs, vxlan);
2876 return 0;
553675fb 2877}
2878
b1be00a6
JB		 2879static int vxlan_sock_add(struct vxlan_dev *vxlan)
2880{
dc5321d7
MS		 2881 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
2882 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
d074bf96 2883 bool ipv4 = !ipv6 || metadata;
b1be00a6
JB		 2884 int ret = 0;
2885
c6fcc4fc 2886 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
b1be00a6 2887#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 2888 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
d074bf96 2889 if (ipv6) {
b1be00a6 2890 ret = __vxlan_sock_add(vxlan, true);
d074bf96
JB		 2891 if (ret < 0 && ret != -EAFNOSUPPORT)
2892 ipv4 = false;
2893 }
b1be00a6 2894#endif
d074bf96 2895 if (ipv4)
b1be00a6
JB		 2896 ret = __vxlan_sock_add(vxlan, false);
2897 if (ret < 0)
2898 vxlan_sock_release(vxlan);
2899 return ret;
2900}
2901
a985343b
MS		 2902static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
2903 struct net_device **lower,
2904 struct vxlan_dev *old)
d342894c 2905{
33564bbb 2906 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
a985343b 2907 struct vxlan_dev *tmp;
e4c7ed41 2908 bool use_ipv6 = false;
d342894c 2909
a985343b
MS		 2910 if (conf->flags & VXLAN_F_GPE) {
2911 /* For now, allow GPE only together with
2912 * COLLECT_METADATA. This can be relaxed later; in such
2913 * case, the other side of the PtP link will have to be
2914 * provided.
2915 */
2916 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
2917 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
2918 return -EINVAL;
3555621d 2919 }
e1e5314d 2920 }
0c867c9b 2921
ce44a4ae
MS		 2922 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
2923 /* Unless IPv6 is explicitly requested, assume IPv4 */
a985343b 2924 conf->remote_ip.sa.sa_family = AF_INET;
ce44a4ae
MS		 2925 conf->saddr.sa.sa_family = AF_INET;
2926 } else if (!conf->remote_ip.sa.sa_family) {
2927 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
2928 } else if (!conf->saddr.sa.sa_family) {
2929 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
2930 }
2931
2932 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family)
2933 return -EINVAL;
e4c7ed41 2934
0f22a3c6
MS		 2935 if (vxlan_addr_multicast(&conf->saddr))
2936 return -EINVAL;
2937
ce44a4ae 2938 if (conf->saddr.sa.sa_family == AF_INET6) {
057ba29b
JB		 2939 if (!IS_ENABLED(CONFIG_IPV6))
2940 return -EPFNOSUPPORT;
e4c7ed41 2941 use_ipv6 = true;
a985343b 2942 conf->flags |= VXLAN_F_IPV6;
0f22a3c6
MS		 2943
2944 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
2945 int local_type =
2946 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
2947 int remote_type =
2948 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
2949
2950 if (local_type & IPV6_ADDR_LINKLOCAL) {
2951 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
2952 (remote_type != IPV6_ADDR_ANY))
2953 return -EINVAL;
2954
2955 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
2956 } else {
2957 if (remote_type ==
2958 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL))
2959 return -EINVAL;
2960
2961 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
2962 }
2963 }
057ba29b 2964 }
d342894c 2965
a985343b 2966 if (conf->label && !use_ipv6)
e7f70af1 2967 return -EINVAL;
e7f70af1 2968
a985343b
MS		 2969 if (conf->remote_ifindex) {
2970 struct net_device *lowerdev;
34e02aa1 2971
a985343b
MS		 2972 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
2973 if (!lowerdev)
34e02aa1 2974 return -ENODEV;
d342894c 2975
e4c7ed41
CW		 2976#if IS_ENABLED(CONFIG_IPV6)
2977 if (use_ipv6) {
2978 struct inet6_dev *idev = __in6_dev_get(lowerdev);
a985343b 2979 if (idev && idev->cnf.disable_ipv6)
e4c7ed41 2980 return -EPERM;
e4c7ed41
CW		 2981 }
2982#endif
2983
a985343b
MS		 2984 *lower = lowerdev;
2985 } else {
2986 if (vxlan_addr_multicast(&conf->remote_ip))
2987 return -EINVAL;
1ba56fb4 2988
0f22a3c6
MS		 2989#if IS_ENABLED(CONFIG_IPV6)
2990 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL)
2991 return -EINVAL;
2992#endif
2993
a985343b 2994 *lower = NULL;
9dc2ad10 2995 }
d342894c 2996
a985343b
MS		 2997 if (!conf->dst_port) {
2998 if (conf->flags & VXLAN_F_GPE)
2999 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3000 else
3001 conf->dst_port = htons(vxlan_port);
d6acfeb1
FM		 3002 }
3003
a985343b
MS		 3004 if (!conf->age_interval)
3005 conf->age_interval = FDB_AGE_DEFAULT;
91572088 3006
a985343b
MS		 3007 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3008 if (tmp == old)
3009 continue;
91572088 3010
49f810f0
MS		 3011 if (tmp->cfg.vni != conf->vni)
3012 continue;
3013 if (tmp->cfg.dst_port != conf->dst_port)
3014 continue;
3015 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
ce44a4ae 3016 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
49f810f0
MS		 3017 continue;
3018
3019 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3020 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3021 continue;
3022
3023 return -EEXIST;
a985343b 3024 }
91572088 3025
a985343b
MS		 3026 return 0;
3027}
3028
3029static void vxlan_config_apply(struct net_device *dev,
3030 struct vxlan_config *conf,
3031 struct net_device *lowerdev, bool changelink)
3032{
3033 struct vxlan_dev *vxlan = netdev_priv(dev);
3034 struct vxlan_rdst *dst = &vxlan->default_dst;
3035 unsigned short needed_headroom = ETH_HLEN;
3036 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3037 int max_mtu = ETH_MAX_MTU;
3038
3039 if (!changelink) {
3040 if (conf->flags & VXLAN_F_GPE)
3041 vxlan_raw_setup(dev);
3042 else
3043 vxlan_ether_setup(dev);
3044
3045 if (conf->mtu)
3046 dev->mtu = conf->mtu;
3047 }
3048
3049 dst->remote_vni = conf->vni;
3050
3051 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3052
3053 if (lowerdev) {
3054 dst->remote_ifindex = conf->remote_ifindex;
3055
3056 dev->gso_max_size = lowerdev->gso_max_size;
3057 dev->gso_max_segs = lowerdev->gso_max_segs;
91572088 3058
a985343b 3059 needed_headroom = lowerdev->hard_header_len;
91572088 3060
a985343b
MS		 3061 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3062 VXLAN_HEADROOM);
7e059158
DW		 3063 }
3064
a985343b
MS		 3065 if (dev->mtu > max_mtu)
3066 dev->mtu = max_mtu;
3067
b1be00a6
JB		 3068 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3069 needed_headroom += VXLAN6_HEADROOM;
3070 else
3071 needed_headroom += VXLAN_HEADROOM;
3072 dev->needed_headroom = needed_headroom;
3073
0dfbdf41 3074 memcpy(&vxlan->cfg, conf, sizeof(*conf));
a985343b 3075}
0dfbdf41 3076
a985343b
MS		 3077static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3078 struct vxlan_config *conf,
3079 bool changelink)
3080{
3081 struct vxlan_dev *vxlan = netdev_priv(dev);
3082 struct net_device *lowerdev;
3083 int ret;
0dfbdf41 3084
a985343b
MS		 3085 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan);
3086 if (ret)
3087 return ret;
8bcdc4f3 3088
a985343b 3089 vxlan_config_apply(dev, conf, lowerdev, changelink);
0dfbdf41 3090
0dfbdf41
TG		 3091 return 0;
3092}
3093
c80498e3
ND		 3094static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3095 struct vxlan_config *conf)
3096{
3097 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3098 struct vxlan_dev *vxlan = netdev_priv(dev);
3099 int err;
3100
3101 err = vxlan_dev_configure(net, dev, conf, false);
3102 if (err)
3103 return err;
3104
3105 dev->ethtool_ops = &vxlan_ethtool_ops;
3106
3107 /* create an fdb entry for a valid default destination */
3108 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
3109 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3110 &vxlan->default_dst.remote_ip,
3111 NUD_REACHABLE | NUD_PERMANENT,
3112 NLM_F_EXCL | NLM_F_CREATE,
3113 vxlan->cfg.dst_port,
3114 vxlan->default_dst.remote_vni,
3115 vxlan->default_dst.remote_vni,
3116 vxlan->default_dst.remote_ifindex,
3117 NTF_SELF);
3118 if (err)
3119 return err;
3120 }
3121
3122 err = register_netdevice(dev);
3123 if (err) {
3124 vxlan_fdb_delete_default(vxlan, vxlan->default_dst.remote_vni);
3125 return err;
3126 }
3127
3128 list_add(&vxlan->next, &vn->vxlan_list);
3129 return 0;
3130}
3131
8bcdc4f3
RP		 3132static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3133 struct net_device *dev, struct vxlan_config *conf,
3134 bool changelink)
0dfbdf41 3135{
8bcdc4f3 3136 struct vxlan_dev *vxlan = netdev_priv(dev);
0dfbdf41 3137
8bcdc4f3 3138 memset(conf, 0, sizeof(*conf));
e277de5f 3139
8bcdc4f3
RP		 3140 /* if changelink operation, start with old existing cfg */
3141 if (changelink)
3142 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3143
3144 if (data[IFLA_VXLAN_ID]) {
3145 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3146
3147 if (changelink && (vni != conf->vni))
3148 return -EOPNOTSUPP;
3149 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3150 }
0dfbdf41
TG		 3151
3152 if (data[IFLA_VXLAN_GROUP]) {
ce44a4ae
MS		 3153 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET))
3154 return -EOPNOTSUPP;
3155
8bcdc4f3 3156 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
ce44a4ae 3157 conf->remote_ip.sa.sa_family = AF_INET;
0dfbdf41
TG		 3158 } else if (data[IFLA_VXLAN_GROUP6]) {
3159 if (!IS_ENABLED(CONFIG_IPV6))
3160 return -EPFNOSUPPORT;
3161
ce44a4ae
MS		 3162 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6))
3163 return -EOPNOTSUPP;
3164
8bcdc4f3
RP		 3165 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3166 conf->remote_ip.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3167 }
3168
3169 if (data[IFLA_VXLAN_LOCAL]) {
ce44a4ae
MS		 3170 if (changelink && (conf->saddr.sa.sa_family != AF_INET))
3171 return -EOPNOTSUPP;
3172
8bcdc4f3
RP		 3173 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3174 conf->saddr.sa.sa_family = AF_INET;
0dfbdf41
TG		 3175 } else if (data[IFLA_VXLAN_LOCAL6]) {
3176 if (!IS_ENABLED(CONFIG_IPV6))
3177 return -EPFNOSUPPORT;
3178
ce44a4ae
MS		 3179 if (changelink && (conf->saddr.sa.sa_family != AF_INET6))
3180 return -EOPNOTSUPP;
3181
0dfbdf41 3182 /* TODO: respect scope id */
8bcdc4f3
RP		 3183 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3184 conf->saddr.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3185 }
3186
3187 if (data[IFLA_VXLAN_LINK])
8bcdc4f3 3188 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
0dfbdf41 3189
d342894c 3190 if (data[IFLA_VXLAN_TOS])
8bcdc4f3 3191 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 3192
afb97186 3193 if (data[IFLA_VXLAN_TTL])
8bcdc4f3 3194 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 3195
e7f70af1 3196 if (data[IFLA_VXLAN_LABEL])
8bcdc4f3 3197 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
e7f70af1
DB		 3198 IPV6_FLOWLABEL_MASK;
3199
8bcdc4f3 3200 if (data[IFLA_VXLAN_LEARNING]) {
dc5321d7 3201 if (nla_get_u8(data[IFLA_VXLAN_LEARNING]))
8bcdc4f3 3202 conf->flags |= VXLAN_F_LEARN;
dc5321d7 3203 else
8bcdc4f3 3204 conf->flags &= ~VXLAN_F_LEARN;
8bcdc4f3
RP		 3205 } else if (!changelink) {
3206 /* default to learn on a new device */
3207 conf->flags |= VXLAN_F_LEARN;
3208 }
d342894c 3209
8bcdc4f3
RP		 3210 if (data[IFLA_VXLAN_AGEING]) {
3211 if (changelink)
3212 return -EOPNOTSUPP;
3213 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3214 }
d342894c 3215
8bcdc4f3
RP		 3216 if (data[IFLA_VXLAN_PROXY]) {
3217 if (changelink)
3218 return -EOPNOTSUPP;
3219 if (nla_get_u8(data[IFLA_VXLAN_PROXY]))
3220 conf->flags |= VXLAN_F_PROXY;
3221 }
e4f67add 3222
8bcdc4f3
RP		 3223 if (data[IFLA_VXLAN_RSC]) {
3224 if (changelink)
3225 return -EOPNOTSUPP;
3226 if (nla_get_u8(data[IFLA_VXLAN_RSC]))
3227 conf->flags |= VXLAN_F_RSC;
3228 }
e4f67add 3229
8bcdc4f3
RP		 3230 if (data[IFLA_VXLAN_L2MISS]) {
3231 if (changelink)
3232 return -EOPNOTSUPP;
3233 if (nla_get_u8(data[IFLA_VXLAN_L2MISS]))
3234 conf->flags |= VXLAN_F_L2MISS;
3235 }
e4f67add 3236
8bcdc4f3
RP		 3237 if (data[IFLA_VXLAN_L3MISS]) {
3238 if (changelink)
3239 return -EOPNOTSUPP;
3240 if (nla_get_u8(data[IFLA_VXLAN_L3MISS]))
3241 conf->flags |= VXLAN_F_L3MISS;
3242 }
e4f67add 3243
8bcdc4f3
RP		 3244 if (data[IFLA_VXLAN_LIMIT]) {
3245 if (changelink)
3246 return -EOPNOTSUPP;
3247 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3248 }
d342894c 3249
8bcdc4f3
RP		 3250 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3251 if (changelink)
3252 return -EOPNOTSUPP;
3253 if (nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3254 conf->flags |= VXLAN_F_COLLECT_METADATA;
3255 }
f8a9b1bc 3256
05f47d69 3257 if (data[IFLA_VXLAN_PORT_RANGE]) {
8bcdc4f3
RP		 3258 if (!changelink) {
3259 const struct ifla_vxlan_port_range *p
3260 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3261 conf->port_min = ntohs(p->low);
3262 conf->port_max = ntohs(p->high);
3263 } else {
3264 return -EOPNOTSUPP;
3265 }
05f47d69 3266 }
3267
8bcdc4f3
RP		 3268 if (data[IFLA_VXLAN_PORT]) {
3269 if (changelink)
3270 return -EOPNOTSUPP;
3271 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3272 }
823aa873 3273
8bcdc4f3
RP		 3274 if (data[IFLA_VXLAN_UDP_CSUM]) {
3275 if (changelink)
3276 return -EOPNOTSUPP;
3277 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3278 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3279 }
359a0ea9 3280
8bcdc4f3
RP		 3281 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3282 if (changelink)
3283 return -EOPNOTSUPP;
3284 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
3285 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
3286 }
359a0ea9 3287
8bcdc4f3
RP		 3288 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3289 if (changelink)
3290 return -EOPNOTSUPP;
3291 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
3292 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
3293 }
359a0ea9 3294
8bcdc4f3
RP		 3295 if (data[IFLA_VXLAN_REMCSUM_TX]) {
3296 if (changelink)
3297 return -EOPNOTSUPP;
3298 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
3299 conf->flags |= VXLAN_F_REMCSUM_TX;
3300 }
dfd8645e 3301
8bcdc4f3
RP		 3302 if (data[IFLA_VXLAN_REMCSUM_RX]) {
3303 if (changelink)
3304 return -EOPNOTSUPP;
3305 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3306 conf->flags |= VXLAN_F_REMCSUM_RX;
3307 }
3308
3309 if (data[IFLA_VXLAN_GBP]) {
3310 if (changelink)
3311 return -EOPNOTSUPP;
3312 conf->flags |= VXLAN_F_GBP;
3313 }
3314
3315 if (data[IFLA_VXLAN_GPE]) {
3316 if (changelink)
3317 return -EOPNOTSUPP;
3318 conf->flags |= VXLAN_F_GPE;
3319 }
3320
3321 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3322 if (changelink)
3323 return -EOPNOTSUPP;
3324 conf->flags |= VXLAN_F_REMCSUM_NOPARTIAL;
3325 }
3326
3327 if (tb[IFLA_MTU]) {
3328 if (changelink)
3329 return -EOPNOTSUPP;
3330 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3331 }
3332
3333 return 0;
3334}
3335
3336static int vxlan_newlink(struct net *src_net, struct net_device *dev,
7a3f4a18
MS		 3337 struct nlattr *tb[], struct nlattr *data[],
3338 struct netlink_ext_ack *extack)
8bcdc4f3 3339{
8bcdc4f3
RP		 3340 struct vxlan_config conf;
3341 int err;
3342
3343 err = vxlan_nl2conf(tb, data, dev, &conf, false);
3344 if (err)
3345 return err;
3346
c80498e3 3347 return __vxlan_dev_create(src_net, dev, &conf);
8bcdc4f3 3348}
dfd8645e 3349
8bcdc4f3 3350static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
ad744b22
MS		 3351 struct nlattr *data[],
3352 struct netlink_ext_ack *extack)
8bcdc4f3
RP		 3353{
3354 struct vxlan_dev *vxlan = netdev_priv(dev);
3355 struct vxlan_rdst *dst = &vxlan->default_dst;
3356 struct vxlan_rdst old_dst;
3357 struct vxlan_config conf;
3358 int err;
3359
3360 err = vxlan_nl2conf(tb, data,
3361 dev, &conf, true);
3362 if (err)
3363 return err;
3511494c 3364
8bcdc4f3 3365 memcpy(&old_dst, dst, sizeof(struct vxlan_rdst));
e1e5314d 3366
8bcdc4f3
RP		 3367 err = vxlan_dev_configure(vxlan->net, dev, &conf, true);
3368 if (err)
3369 return err;
0ace2ca8 3370
8bcdc4f3
RP		 3371 /* handle default dst entry */
3372 if (!vxlan_addr_equal(&dst->remote_ip, &old_dst.remote_ip)) {
3373 spin_lock_bh(&vxlan->hash_lock);
3374 if (!vxlan_addr_any(&old_dst.remote_ip))
3375 __vxlan_fdb_delete(vxlan, all_zeros_mac,
3376 old_dst.remote_ip,
3377 vxlan->cfg.dst_port,
3378 old_dst.remote_vni,
3379 old_dst.remote_vni,
3380 old_dst.remote_ifindex, 0);
3381
3382 if (!vxlan_addr_any(&dst->remote_ip)) {
3383 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3384 &dst->remote_ip,
3385 NUD_REACHABLE | NUD_PERMANENT,
3386 NLM_F_CREATE | NLM_F_APPEND,
3387 vxlan->cfg.dst_port,
3388 dst->remote_vni,
3389 dst->remote_vni,
3390 dst->remote_ifindex,
3391 NTF_SELF);
3392 if (err) {
3393 spin_unlock_bh(&vxlan->hash_lock);
3394 return err;
3395 }
3396 }
3397 spin_unlock_bh(&vxlan->hash_lock);
3398 }
ce577668 3399
8bcdc4f3 3400 return 0;
d342894c 3401}
3402
3403static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3404{
3405 struct vxlan_dev *vxlan = netdev_priv(dev);
3406
8b3f9337
RP		 3407 vxlan_flush(vxlan, true);
3408
58ce31cc 3409 gro_cells_destroy(&vxlan->gro_cells);
553675fb 3410 list_del(&vxlan->next);
d342894c 3411 unregister_netdevice_queue(dev, head);
3412}
3413
3414static size_t vxlan_get_size(const struct net_device *dev)
3415{
3416
3417 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3418 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3419 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3420 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3421 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3422 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
e7f70af1 3423 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 3424 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 3425 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3426 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3427 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3428 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3429 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3430 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3431 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3432 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3433 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3434 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3435 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3436 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3437 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3438 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3439 0;
3440}
3441
3442static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3443{
3444 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3445 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3446 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3447 .low = htons(vxlan->cfg.port_min),
3448 .high = htons(vxlan->cfg.port_max),
05f47d69 3449 };
d342894c 3450
54bfd872 3451 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3452 goto nla_put_failure;
3453
e4c7ed41
CW		 3454 if (!vxlan_addr_any(&dst->remote_ip)) {
3455 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3456 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3457 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3458 goto nla_put_failure;
3459#if IS_ENABLED(CONFIG_IPV6)
3460 } else {
930345ea
JB		 3461 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3462 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3463 goto nla_put_failure;
3464#endif
3465 }
3466 }
d342894c 3467
c7995c43 3468 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3469 goto nla_put_failure;
3470
0dfbdf41
TG		 3471 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3472 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3473 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3474 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3475 goto nla_put_failure;
3476#if IS_ENABLED(CONFIG_IPV6)
3477 } else {
930345ea 3478 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3479 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3480 goto nla_put_failure;
3481#endif
3482 }
3483 }
d342894c 3484
0dfbdf41
TG		 3485 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3486 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e7f70af1 3487 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add 3488 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
dc5321d7 3489 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
e4f67add 3490 nla_put_u8(skb, IFLA_VXLAN_PROXY,
dc5321d7
MS		 3491 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
3492 nla_put_u8(skb, IFLA_VXLAN_RSC,
3493 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
e4f67add 3494 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
dc5321d7 3495 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
e4f67add 3496 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
dc5321d7 3497 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
da8b43c0 3498 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
dc5321d7 3499 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3500 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3501 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3502 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3503 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
dc5321d7 3504 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9 3505 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
dc5321d7 3506 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
359a0ea9 3507 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dc5321d7 3508 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
dfd8645e 3509 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
dc5321d7 3510 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
dfd8645e 3511 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
dc5321d7 3512 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
d342894c 3513 goto nla_put_failure;
3514
05f47d69 3515 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3516 goto nla_put_failure;
3517
dc5321d7 3518 if (vxlan->cfg.flags & VXLAN_F_GBP &&
3511494c
TG		 3519 nla_put_flag(skb, IFLA_VXLAN_GBP))
3520 goto nla_put_failure;
3521
dc5321d7 3522 if (vxlan->cfg.flags & VXLAN_F_GPE &&
e1e5314d
JB		 3523 nla_put_flag(skb, IFLA_VXLAN_GPE))
3524 goto nla_put_failure;
3525
dc5321d7 3526 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
0ace2ca8
TH		 3527 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3528 goto nla_put_failure;
3529
d342894c 3530 return 0;
3531
3532nla_put_failure:
3533 return -EMSGSIZE;
3534}
3535
1728d4fa
ND		 3536static struct net *vxlan_get_link_net(const struct net_device *dev)
3537{
3538 struct vxlan_dev *vxlan = netdev_priv(dev);
3539
3540 return vxlan->net;
3541}
3542
d342894c 3543static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3544 .kind = "vxlan",
3545 .maxtype = IFLA_VXLAN_MAX,
3546 .policy = vxlan_policy,
3547 .priv_size = sizeof(struct vxlan_dev),
3548 .setup = vxlan_setup,
3549 .validate = vxlan_validate,
3550 .newlink = vxlan_newlink,
8bcdc4f3 3551 .changelink = vxlan_changelink,
d342894c 3552 .dellink = vxlan_dellink,
3553 .get_size = vxlan_get_size,
3554 .fill_info = vxlan_fill_info,
1728d4fa 3555 .get_link_net = vxlan_get_link_net,
d342894c 3556};
3557
cf5da330
ND		 3558struct net_device *vxlan_dev_create(struct net *net, const char *name,
3559 u8 name_assign_type,
3560 struct vxlan_config *conf)
3561{
3562 struct nlattr *tb[IFLA_MAX + 1];
3563 struct net_device *dev;
3564 int err;
3565
3566 memset(&tb, 0, sizeof(tb));
3567
3568 dev = rtnl_create_link(net, name, name_assign_type,
3569 &vxlan_link_ops, tb);
3570 if (IS_ERR(dev))
3571 return dev;
3572
c80498e3 3573 err = __vxlan_dev_create(net, dev, conf);
cf5da330
ND		 3574 if (err < 0) {
3575 free_netdev(dev);
3576 return ERR_PTR(err);
3577 }
3578
3579 err = rtnl_configure_link(dev, NULL);
3580 if (err < 0) {
3581 LIST_HEAD(list_kill);
3582
3583 vxlan_dellink(dev, &list_kill);
3584 unregister_netdevice_many(&list_kill);
3585 return ERR_PTR(err);
3586 }
3587
3588 return dev;
3589}
3590EXPORT_SYMBOL_GPL(vxlan_dev_create);
3591
acaf4e70
DB		 3592static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3593 struct net_device *dev)
3594{
3595 struct vxlan_dev *vxlan, *next;
3596 LIST_HEAD(list_kill);
3597
3598 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3599 struct vxlan_rdst *dst = &vxlan->default_dst;
3600
3601 /* In case we created vxlan device with carrier
3602 * and we loose the carrier due to module unload
3603 * we also need to remove vxlan device. In other
3604 * cases, it's not necessary and remote_ifindex
3605 * is 0 here, so no matches.
3606 */
3607 if (dst->remote_ifindex == dev->ifindex)
3608 vxlan_dellink(vxlan->dev, &list_kill);
3609 }
3610
3611 unregister_netdevice_many(&list_kill);
3612}
3613
b7aade15
HFS		 3614static int vxlan_netdevice_event(struct notifier_block *unused,
3615 unsigned long event, void *ptr)
acaf4e70
DB		 3616{
3617 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3618 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3619
783c1463 3620 if (event == NETDEV_UNREGISTER)
acaf4e70 3621 vxlan_handle_lowerdev_unregister(vn, dev);
7c46a640 3622 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
b7aade15 3623 vxlan_push_rx_ports(dev);
acaf4e70
DB		 3624
3625 return NOTIFY_DONE;
3626}
3627
3628static struct notifier_block vxlan_notifier_block __read_mostly = {
b7aade15 3629 .notifier_call = vxlan_netdevice_event,
acaf4e70
DB		 3630};
3631
d342894c 3632static __net_init int vxlan_init_net(struct net *net)
3633{
3634 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3635 unsigned int h;
d342894c 3636
553675fb 3637 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3638 spin_lock_init(&vn->sock_lock);
d342894c 3639
553675fb 3640 for (h = 0; h < PORT_HASH_SIZE; ++h)
3641 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3642
3643 return 0;
3644}
3645
f01ec1c0
ND		 3646static void __net_exit vxlan_exit_net(struct net *net)
3647{
3648 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3649 struct vxlan_dev *vxlan, *next;
3650 struct net_device *dev, *aux;
3651 LIST_HEAD(list);
3652
3653 rtnl_lock();
3654 for_each_netdev_safe(net, dev, aux)
3655 if (dev->rtnl_link_ops == &vxlan_link_ops)
3656 unregister_netdevice_queue(dev, &list);
3657
3658 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3659 /* If vxlan->dev is in the same netns, it has already been added
3660 * to the list by the previous loop.
3661 */
58ce31cc
TH		 3662 if (!net_eq(dev_net(vxlan->dev), net)) {
3663 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3664 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3665 }
f01ec1c0
ND		 3666 }
3667
3668 unregister_netdevice_many(&list);
3669 rtnl_unlock();
3670}
3671
d342894c 3672static struct pernet_operations vxlan_net_ops = {
3673 .init = vxlan_init_net,
f01ec1c0 3674 .exit = vxlan_exit_net,
d342894c 3675 .id = &vxlan_net_id,
3676 .size = sizeof(struct vxlan_net),
3677};
3678
3679static int __init vxlan_init_module(void)
3680{
3681 int rc;
3682
3683 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3684
783c1463 3685 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3686 if (rc)
3687 goto out1;
3688
acaf4e70 3689 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3690 if (rc)
3691 goto out2;
3692
acaf4e70
DB		 3693 rc = rtnl_link_register(&vxlan_link_ops);
3694 if (rc)
3695 goto out3;
d342894c 3696
acaf4e70
DB		 3697 return 0;
3698out3:
3699 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3700out2:
783c1463 3701 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3702out1:
3703 return rc;
3704}
7332a13b 3705late_initcall(vxlan_init_module);
d342894c 3706
3707static void __exit vxlan_cleanup_module(void)
3708{
b7153984 3709 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3710 unregister_netdevice_notifier(&vxlan_notifier_block);
783c1463
DB		 3711 unregister_pernet_subsys(&vxlan_net_ops);
3712 /* rcu_barrier() is called by netns */
d342894c 3713}
3714module_exit(vxlan_cleanup_module);
3715
3716MODULE_LICENSE("GPL");
3717MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3718MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3719MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3720MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Artful kernel mirror