projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'wireless-drivers-next-for-davem-2017-06-12' of git://git.kernel.org/pub...
[linux-2.6-block.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
54bfd872 229static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni)
d342894c 230{
231 struct vxlan_dev *vxlan;
d342894c 232
b9167b2e
JB		 233 /* For flow based devices, map all packets to VNI 0 */
234 if (vs->flags & VXLAN_F_COLLECT_METADATA)
235 vni = 0;
236
54bfd872
JB		 237 hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) {
238 if (vxlan->default_dst.remote_vni == vni)
d342894c 239 return vxlan;
240 }
241
242 return NULL;
243}
244
5cfccc5a 245/* Look up VNI in a per net namespace table */
54bfd872 246static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni,
ac5132d1
TG		 247 sa_family_t family, __be16 port,
248 u32 flags)
5cfccc5a
PS		 249{
250 struct vxlan_sock *vs;
251
ac5132d1 252 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 253 if (!vs)
254 return NULL;
255
54bfd872 256 return vxlan_vs_find_vni(vs, vni);
5cfccc5a
PS		 257}
258
d342894c 259/* Fill in neighbour message in skbuff. */
260static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 261 const struct vxlan_fdb *fdb,
262 u32 portid, u32 seq, int type, unsigned int flags,
263 const struct vxlan_rdst *rdst)
d342894c 264{
265 unsigned long now = jiffies;
266 struct nda_cacheinfo ci;
267 struct nlmsghdr *nlh;
268 struct ndmsg *ndm;
e4f67add 269 bool send_ip, send_eth;
d342894c 270
271 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
272 if (nlh == NULL)
273 return -EMSGSIZE;
274
275 ndm = nlmsg_data(nlh);
276 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 277
278 send_eth = send_ip = true;
279
280 if (type == RTM_GETNEIGH) {
e4c7ed41 281 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add 282 send_eth = !is_zero_ether_addr(fdb->eth_addr);
8f48ba71 283 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
e4f67add
DS		 284 } else
285 ndm->ndm_family = AF_BRIDGE;
d342894c 286 ndm->ndm_state = fdb->state;
287 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 288 ndm->ndm_flags = fdb->flags;
545469f7 289 ndm->ndm_type = RTN_UNICAST;
d342894c 290
193523bf 291 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 292 nla_put_s32(skb, NDA_LINK_NETNSID,
38f507f1 293 peernet2id(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 294 goto nla_put_failure;
295
e4f67add 296 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 297 goto nla_put_failure;
298
e4c7ed41 299 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 300 goto nla_put_failure;
301
0dfbdf41 302 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 303 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
304 goto nla_put_failure;
c7995c43 305 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
54bfd872 306 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
6681712d 307 goto nla_put_failure;
3ad7a4b1
RP		 308 if ((vxlan->flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
309 nla_put_u32(skb, NDA_SRC_VNI,
310 be32_to_cpu(fdb->vni)))
311 goto nla_put_failure;
6681712d
DS		 312 if (rdst->remote_ifindex &&
313 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 314 goto nla_put_failure;
315
316 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
317 ci.ndm_confirmed = 0;
318 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
319 ci.ndm_refcnt = 0;
320
321 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
322 goto nla_put_failure;
323
053c095a
JB		 324 nlmsg_end(skb, nlh);
325 return 0;
d342894c 326
327nla_put_failure:
328 nlmsg_cancel(skb, nlh);
329 return -EMSGSIZE;
330}
331
332static inline size_t vxlan_nlmsg_size(void)
333{
334 return NLMSG_ALIGN(sizeof(struct ndmsg))
335 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 336 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 337 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 338 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
339 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 340 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 341 + nla_total_size(sizeof(struct nda_cacheinfo));
342}
343
9e4b93f9
ND		 344static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
345 struct vxlan_rdst *rd, int type)
d342894c 346{
347 struct net *net = dev_net(vxlan->dev);
348 struct sk_buff *skb;
349 int err = -ENOBUFS;
350
351 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
352 if (skb == NULL)
353 goto errout;
354
9e4b93f9 355 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 356 if (err < 0) {
357 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
358 WARN_ON(err == -EMSGSIZE);
359 kfree_skb(skb);
360 goto errout;
361 }
362
363 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
364 return;
365errout:
366 if (err < 0)
367 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
368}
369
e4c7ed41 370static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 371{
372 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 373 struct vxlan_fdb f = {
374 .state = NUD_STALE,
375 };
376 struct vxlan_rdst remote = {
e4c7ed41 377 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 378 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 379 };
3e61aa8f 380
9e4b93f9 381 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 382}
383
384static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
385{
bb3fd687
SH		 386 struct vxlan_fdb f = {
387 .state = NUD_STALE,
388 };
9e4b93f9 389 struct vxlan_rdst remote = { };
e4f67add 390
e4f67add
DS		 391 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
392
9e4b93f9 393 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 394}
395
d342894c 396/* Hash Ethernet address */
397static u32 eth_hash(const unsigned char *addr)
398{
399 u64 value = get_unaligned((u64 *)addr);
400
401 /* only want 6 bytes */
402#ifdef __BIG_ENDIAN
d342894c 403 value >>= 16;
321fb991 404#else
405 value <<= 16;
d342894c 406#endif
407 return hash_64(value, FDB_HASH_BITS);
408}
409
3ad7a4b1
RP		 410static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
411{
412 /* use 1 byte of OUI and 3 bytes of NIC */
413 u32 key = get_unaligned((u32 *)(addr + 2));
414
415 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
416}
417
d342894c 418/* Hash chain to use given mac address */
419static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
3ad7a4b1 420 const u8 *mac, __be32 vni)
d342894c 421{
3ad7a4b1
RP		 422 if (vxlan->flags & VXLAN_F_COLLECT_METADATA)
423 return &vxlan->fdb_head[eth_vni_hash(mac, vni)];
424 else
425 return &vxlan->fdb_head[eth_hash(mac)];
d342894c 426}
427
428/* Look up Ethernet address in forwarding table */
014be2c8 429static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 430 const u8 *mac, __be32 vni)
d342894c 431{
3ad7a4b1 432 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
d342894c 433 struct vxlan_fdb *f;
d342894c 434
b67bfe0d 435 hlist_for_each_entry_rcu(f, head, hlist) {
3ad7a4b1
RP		 436 if (ether_addr_equal(mac, f->eth_addr)) {
437 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
438 if (vni == f->vni)
439 return f;
440 } else {
441 return f;
442 }
443 }
d342894c 444 }
445
446 return NULL;
447}
448
014be2c8 449static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 450 const u8 *mac, __be32 vni)
014be2c8
SS		 451{
452 struct vxlan_fdb *f;
453
3ad7a4b1 454 f = __vxlan_find_mac(vxlan, mac, vni);
014be2c8
SS		 455 if (f)
456 f->used = jiffies;
457
458 return f;
459}
460
a5e7c10a
MR		 461/* caller should hold vxlan->hash_lock */
462static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 463 union vxlan_addr *ip, __be16 port,
54bfd872 464 __be32 vni, __u32 ifindex)
6681712d 465{
3e61aa8f 466 struct vxlan_rdst *rd;
6681712d 467
3e61aa8f 468 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 469 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 470 rd->remote_port == port &&
471 rd->remote_vni == vni &&
472 rd->remote_ifindex == ifindex)
a5e7c10a 473 return rd;
6681712d 474 }
3e61aa8f 475
a5e7c10a
MR		 476 return NULL;
477}
478
906dc186
TR		 479/* Replace destination of unicast mac */
480static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872
JB		 481 union vxlan_addr *ip, __be16 port, __be32 vni,
482 __u32 ifindex)
906dc186
TR		 483{
484 struct vxlan_rdst *rd;
485
486 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
487 if (rd)
488 return 0;
489
490 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
491 if (!rd)
492 return 0;
0c1d70af
PA		 493
494 dst_cache_reset(&rd->dst_cache);
e4c7ed41 495 rd->remote_ip = *ip;
906dc186
TR		 496 rd->remote_port = port;
497 rd->remote_vni = vni;
498 rd->remote_ifindex = ifindex;
499 return 1;
500}
501
a5e7c10a
MR		 502/* Add/update destinations for multicast */
503static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 504 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 505 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 506{
507 struct vxlan_rdst *rd;
508
509 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
510 if (rd)
511 return 0;
512
6681712d
DS		 513 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
514 if (rd == NULL)
515 return -ENOBUFS;
0c1d70af
PA		 516
517 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
518 kfree(rd);
519 return -ENOBUFS;
520 }
521
e4c7ed41 522 rd->remote_ip = *ip;
6681712d
DS		 523 rd->remote_port = port;
524 rd->remote_vni = vni;
525 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 526
527 list_add_tail_rcu(&rd->list, &f->remotes);
528
9e4b93f9 529 *rdp = rd;
6681712d
DS		 530 return 1;
531}
532
dfd8645e
TH		 533static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
534 unsigned int off,
535 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 536 __be32 vni_field,
537 struct gro_remcsum *grc,
0ace2ca8 538 bool nopartial)
dfd8645e 539{
b7fe10e5 540 size_t start, offset;
dfd8645e
TH		 541
542 if (skb->remcsum_offload)
b7fe10e5 543 return vh;
dfd8645e
TH		 544
545 if (!NAPI_GRO_CB(skb)->csum_valid)
546 return NULL;
547
54bfd872
JB		 548 start = vxlan_rco_start(vni_field);
549 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 550
b7fe10e5
TH		 551 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
552 start, offset, grc, nopartial);
dfd8645e
TH		 553
554 skb->remcsum_offload = 1;
555
556 return vh;
557}
558
5602c48c
TH		 559static struct sk_buff **vxlan_gro_receive(struct sock *sk,
560 struct sk_buff **head,
561 struct sk_buff *skb)
dc01e7d3
OG		 562{
563 struct sk_buff *p, **pp = NULL;
564 struct vxlanhdr *vh, *vh2;
9b174d88 565 unsigned int hlen, off_vx;
dc01e7d3 566 int flush = 1;
5602c48c 567 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
54bfd872 568 __be32 flags;
26c4f7da
TH		 569 struct gro_remcsum grc;
570
571 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 572
573 off_vx = skb_gro_offset(skb);
574 hlen = off_vx + sizeof(*vh);
575 vh = skb_gro_header_fast(skb, off_vx);
576 if (skb_gro_header_hard(skb, hlen)) {
577 vh = skb_gro_header_slow(skb, hlen, off_vx);
578 if (unlikely(!vh))
579 goto out;
580 }
dc01e7d3 581
dfd8645e
TH		 582 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
583
54bfd872 584 flags = vh->vx_flags;
dfd8645e
TH		 585
586 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
587 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 588 vh->vx_vni, &grc,
0ace2ca8
TH		 589 !!(vs->flags &
590 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 591
592 if (!vh)
593 goto out;
594 }
595
b7fe10e5
TH		 596 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
597
dc01e7d3
OG		 598 for (p = *head; p; p = p->next) {
599 if (!NAPI_GRO_CB(p)->same_flow)
600 continue;
601
602 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 603 if (vh->vx_flags != vh2->vx_flags ||
604 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 605 NAPI_GRO_CB(p)->same_flow = 0;
606 continue;
607 }
dc01e7d3
OG		 608 }
609
fcd91dd4 610 pp = call_gro_receive(eth_gro_receive, head, skb);
c194cf93 611 flush = 0;
dc01e7d3 612
dc01e7d3 613out:
26c4f7da 614 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 615 NAPI_GRO_CB(skb)->flush |= flush;
616
617 return pp;
618}
619
5602c48c 620static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
dc01e7d3 621{
229740c6
JR		 622 /* Sets 'skb->inner_mac_header' since we are always called with
623 * 'skb->encapsulation' set.
624 */
9b174d88 625 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 626}
627
d342894c 628/* Add new entry to forwarding table -- assumes lock held */
629static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 630 const u8 *mac, union vxlan_addr *ip,
6681712d 631 __u16 state, __u16 flags,
3ad7a4b1
RP		 632 __be16 port, __be32 src_vni, __be32 vni,
633 __u32 ifindex, __u8 ndm_flags)
d342894c 634{
9e4b93f9 635 struct vxlan_rdst *rd = NULL;
d342894c 636 struct vxlan_fdb *f;
637 int notify = 0;
17b46365 638 int rc;
d342894c 639
3ad7a4b1 640 f = __vxlan_find_mac(vxlan, mac, src_vni);
d342894c 641 if (f) {
642 if (flags & NLM_F_EXCL) {
643 netdev_dbg(vxlan->dev,
644 "lost race to create %pM\n", mac);
645 return -EEXIST;
646 }
647 if (f->state != state) {
648 f->state = state;
649 f->updated = jiffies;
650 notify = 1;
651 }
ae884082
DS		 652 if (f->flags != ndm_flags) {
653 f->flags = ndm_flags;
654 f->updated = jiffies;
655 notify = 1;
656 }
906dc186
TR		 657 if ((flags & NLM_F_REPLACE)) {
658 /* Only change unicasts */
659 if (!(is_multicast_ether_addr(f->eth_addr) ||
660 is_zero_ether_addr(f->eth_addr))) {
60840429 661 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 662 ifindex);
906dc186
TR		 663 } else
664 return -EOPNOTSUPP;
665 }
6681712d 666 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 667 (is_multicast_ether_addr(f->eth_addr) ||
668 is_zero_ether_addr(f->eth_addr))) {
17b46365 669 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
6681712d
DS		 670
671 if (rc < 0)
672 return rc;
673 notify |= rc;
674 }
d342894c 675 } else {
676 if (!(flags & NLM_F_CREATE))
677 return -ENOENT;
678
0dfbdf41
TG		 679 if (vxlan->cfg.addrmax &&
680 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 681 return -ENOSPC;
682
906dc186
TR		 683 /* Disallow replace to add a multicast entry */
684 if ((flags & NLM_F_REPLACE) &&
685 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
686 return -EOPNOTSUPP;
687
e4c7ed41 688 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 689 f = kmalloc(sizeof(*f), GFP_ATOMIC);
690 if (!f)
691 return -ENOMEM;
692
693 notify = 1;
d342894c 694 f->state = state;
ae884082 695 f->flags = ndm_flags;
d342894c 696 f->updated = f->used = jiffies;
3ad7a4b1 697 f->vni = src_vni;
3e61aa8f 698 INIT_LIST_HEAD(&f->remotes);
d342894c 699 memcpy(f->eth_addr, mac, ETH_ALEN);
700
17b46365
HY		 701 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
702 if (rc < 0) {
703 kfree(f);
704 return rc;
705 }
3e61aa8f 706
d342894c 707 ++vxlan->addrcnt;
708 hlist_add_head_rcu(&f->hlist,
3ad7a4b1 709 vxlan_fdb_head(vxlan, mac, src_vni));
d342894c 710 }
711
9e4b93f9
ND		 712 if (notify) {
713 if (rd == NULL)
714 rd = first_remote_rtnl(f);
715 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
716 }
d342894c 717
718 return 0;
719}
720
6706c82e 721static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 722{
723 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 724 struct vxlan_rdst *rd, *nd;
6681712d 725
0c1d70af
PA		 726 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
727 dst_cache_destroy(&rd->dst_cache);
6681712d 728 kfree(rd);
0c1d70af 729 }
6681712d
DS		 730 kfree(f);
731}
732
d342894c 733static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
734{
735 netdev_dbg(vxlan->dev,
736 "delete %pM\n", f->eth_addr);
737
738 --vxlan->addrcnt;
9e4b93f9 739 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 740
741 hlist_del_rcu(&f->hlist);
6681712d 742 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 743}
744
35cf2845
LR		 745static void vxlan_dst_free(struct rcu_head *head)
746{
747 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
748
749 dst_cache_destroy(&rd->dst_cache);
750 kfree(rd);
751}
752
753static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
754 struct vxlan_rdst *rd)
755{
756 list_del_rcu(&rd->list);
757 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
758 call_rcu(&rd->rcu, vxlan_dst_free);
759}
760
f0b074be 761static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
3ad7a4b1
RP		 762 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
763 __be32 *vni, u32 *ifindex)
d342894c 764{
6681712d 765 struct net *net = dev_net(vxlan->dev);
e4c7ed41 766 int err;
d342894c 767
f0b074be 768 if (tb[NDA_DST]) {
e4c7ed41
CW		 769 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
770 if (err)
771 return err;
f0b074be 772 } else {
e4c7ed41
CW		 773 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
774 if (remote->sa.sa_family == AF_INET) {
775 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
776 ip->sa.sa_family = AF_INET;
777#if IS_ENABLED(CONFIG_IPV6)
778 } else {
779 ip->sin6.sin6_addr = in6addr_any;
780 ip->sa.sa_family = AF_INET6;
781#endif
782 }
f0b074be 783 }
d342894c 784
6681712d 785 if (tb[NDA_PORT]) {
73cf3317 786 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 787 return -EINVAL;
f0b074be
MR		 788 *port = nla_get_be16(tb[NDA_PORT]);
789 } else {
0dfbdf41 790 *port = vxlan->cfg.dst_port;
f0b074be 791 }
6681712d
DS		 792
793 if (tb[NDA_VNI]) {
794 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
795 return -EINVAL;
54bfd872 796 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 797 } else {
798 *vni = vxlan->default_dst.remote_vni;
799 }
6681712d 800
3ad7a4b1
RP		 801 if (tb[NDA_SRC_VNI]) {
802 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
803 return -EINVAL;
804 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
805 } else {
806 *src_vni = vxlan->default_dst.remote_vni;
807 }
808
6681712d 809 if (tb[NDA_IFINDEX]) {
5abb0029 810 struct net_device *tdev;
6681712d
DS		 811
812 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
813 return -EINVAL;
f0b074be 814 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 815 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 816 if (!tdev)
6681712d 817 return -EADDRNOTAVAIL;
f0b074be
MR		 818 } else {
819 *ifindex = 0;
820 }
821
822 return 0;
823}
824
825/* Add static entry (via netlink) */
826static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
827 struct net_device *dev,
f6f6424b 828 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 829{
830 struct vxlan_dev *vxlan = netdev_priv(dev);
831 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 832 union vxlan_addr ip;
f0b074be 833 __be16 port;
3ad7a4b1 834 __be32 src_vni, vni;
54bfd872 835 u32 ifindex;
f0b074be
MR		 836 int err;
837
838 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
839 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
840 ndm->ndm_state);
841 return -EINVAL;
842 }
843
844 if (tb[NDA_DST] == NULL)
845 return -EINVAL;
846
3ad7a4b1 847 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
f0b074be
MR		 848 if (err)
849 return err;
6681712d 850
5933a7bb
MR		 851 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
852 return -EAFNOSUPPORT;
853
d342894c 854 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 855 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
3ad7a4b1 856 port, src_vni, vni, ifindex, ndm->ndm_flags);
d342894c 857 spin_unlock_bh(&vxlan->hash_lock);
858
859 return err;
860}
861
3ad7a4b1
RP		 862static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
863 const unsigned char *addr, union vxlan_addr ip,
864 __be16 port, __be32 src_vni, u32 vni, u32 ifindex,
865 u16 vid)
d342894c 866{
d342894c 867 struct vxlan_fdb *f;
bc7892ba 868 struct vxlan_rdst *rd = NULL;
3ad7a4b1 869 int err = -ENOENT;
bc7892ba 870
3ad7a4b1 871 f = vxlan_find_mac(vxlan, addr, src_vni);
bc7892ba 872 if (!f)
3ad7a4b1 873 return err;
bc7892ba 874
e4c7ed41
CW		 875 if (!vxlan_addr_any(&ip)) {
876 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 877 if (!rd)
878 goto out;
879 }
880
bc7892ba
MR		 881 /* remove a destination if it's not the only one on the list,
882 * otherwise destroy the fdb entry
883 */
884 if (rd && !list_is_singular(&f->remotes)) {
35cf2845 885 vxlan_fdb_dst_destroy(vxlan, f, rd);
bc7892ba 886 goto out;
d342894c 887 }
bc7892ba
MR		 888
889 vxlan_fdb_destroy(vxlan, f);
890
891out:
3ad7a4b1
RP		 892 return 0;
893}
894
895/* Delete entry (via netlink) */
896static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
897 struct net_device *dev,
898 const unsigned char *addr, u16 vid)
899{
900 struct vxlan_dev *vxlan = netdev_priv(dev);
901 union vxlan_addr ip;
902 __be32 src_vni, vni;
903 __be16 port;
904 u32 ifindex;
905 int err;
906
907 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
908 if (err)
909 return err;
910
911 spin_lock_bh(&vxlan->hash_lock);
912 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
913 vid);
d342894c 914 spin_unlock_bh(&vxlan->hash_lock);
915
916 return err;
917}
918
919/* Dump forwarding table */
920static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 921 struct net_device *dev,
d297653d 922 struct net_device *filter_dev, int *idx)
d342894c 923{
924 struct vxlan_dev *vxlan = netdev_priv(dev);
925 unsigned int h;
d297653d 926 int err = 0;
d342894c 927
928 for (h = 0; h < FDB_HASH_SIZE; ++h) {
929 struct vxlan_fdb *f;
d342894c 930
b67bfe0d 931 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 932 struct vxlan_rdst *rd;
6681712d 933
3e61aa8f 934 list_for_each_entry_rcu(rd, &f->remotes, list) {
d297653d 935 if (*idx < cb->args[2])
07a51cd3
AW		 936 goto skip;
937
6681712d
DS		 938 err = vxlan_fdb_info(skb, vxlan, f,
939 NETLINK_CB(cb->skb).portid,
940 cb->nlh->nlmsg_seq,
941 RTM_NEWNEIGH,
942 NLM_F_MULTI, rd);
d297653d 943 if (err < 0)
3e61aa8f 944 goto out;
3e61aa8f 945skip:
d297653d 946 *idx += 1;
07a51cd3 947 }
d342894c 948 }
949 }
3e61aa8f 950out:
d297653d 951 return err;
d342894c 952}
953
954/* Watch incoming packets to learn mapping between Ethernet address
955 * and Tunnel endpoint.
c4b49512 956 * Return true if packet is bogus and should be dropped.
d342894c 957 */
26a41ae6 958static bool vxlan_snoop(struct net_device *dev,
3ad7a4b1
RP		 959 union vxlan_addr *src_ip, const u8 *src_mac,
960 __be32 vni)
d342894c 961{
962 struct vxlan_dev *vxlan = netdev_priv(dev);
963 struct vxlan_fdb *f;
d342894c 964
3ad7a4b1 965 f = vxlan_find_mac(vxlan, src_mac, vni);
d342894c 966 if (likely(f)) {
5ca5461c 967 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 968
e4c7ed41 969 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 970 return false;
971
972 /* Don't migrate static entries, drop packets */
eb064c3b 973 if (f->state & NUD_NOARP)
26a41ae6 974 return true;
d342894c 975
976 if (net_ratelimit())
977 netdev_info(dev,
e4c7ed41 978 "%pM migrated from %pIS to %pIS\n",
a4870f79 979 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 980
e4c7ed41 981 rdst->remote_ip = *src_ip;
d342894c 982 f->updated = jiffies;
9e4b93f9 983 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 984 } else {
985 /* learned new entry */
986 spin_lock(&vxlan->hash_lock);
3bf74b1a 987
988 /* close off race between vxlan_flush and incoming packets */
989 if (netif_running(dev))
990 vxlan_fdb_create(vxlan, src_mac, src_ip,
991 NUD_REACHABLE,
992 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 993 vxlan->cfg.dst_port,
3ad7a4b1 994 vni,
3bf74b1a 995 vxlan->default_dst.remote_vni,
996 0, NTF_SELF);
d342894c 997 spin_unlock(&vxlan->hash_lock);
998 }
26a41ae6 999
1000 return false;
d342894c 1001}
1002
d342894c 1003/* See if multicast group is already in use by other ID */
95ab0991 1004static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1005{
553675fb 1006 struct vxlan_dev *vxlan;
c6fcc4fc 1007 struct vxlan_sock *sock4;
4053ab1b
AB		 1008#if IS_ENABLED(CONFIG_IPV6)
1009 struct vxlan_sock *sock6;
1010#endif
b1be00a6 1011 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1012
c6fcc4fc 1013 sock4 = rtnl_dereference(dev->vn4_sock);
1014
95ab0991
G		 1015 /* The vxlan_sock is only used by dev, leaving group has
1016 * no effect on other vxlan devices.
1017 */
c6fcc4fc 1018 if (family == AF_INET && sock4 && atomic_read(&sock4->refcnt) == 1)
95ab0991 1019 return false;
b1be00a6 1020#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1021 sock6 = rtnl_dereference(dev->vn6_sock);
1022 if (family == AF_INET6 && sock6 && atomic_read(&sock6->refcnt) == 1)
b1be00a6
JB		 1023 return false;
1024#endif
95ab0991 1025
553675fb 1026 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1027 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1028 continue;
d342894c 1029
c6fcc4fc 1030 if (family == AF_INET &&
1031 rtnl_dereference(vxlan->vn4_sock) != sock4)
95ab0991 1032 continue;
b1be00a6 1033#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1034 if (family == AF_INET6 &&
1035 rtnl_dereference(vxlan->vn6_sock) != sock6)
b1be00a6
JB		 1036 continue;
1037#endif
95ab0991
G		 1038
1039 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1040 &dev->default_dst.remote_ip))
1041 continue;
1042
1043 if (vxlan->default_dst.remote_ifindex !=
1044 dev->default_dst.remote_ifindex)
1045 continue;
1046
1047 return true;
553675fb 1048 }
d342894c 1049
1050 return false;
1051}
1052
544a773a 1053static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
7c47cedf 1054{
b1be00a6 1055 struct vxlan_net *vn;
012a5729 1056
b1be00a6 1057 if (!vs)
544a773a 1058 return false;
7c47cedf 1059 if (!atomic_dec_and_test(&vs->refcnt))
544a773a 1060 return false;
d342894c 1061
b1be00a6 1062 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1063 spin_lock(&vn->sock_lock);
7c47cedf 1064 hlist_del_rcu(&vs->hlist);
e7b3db5e 1065 udp_tunnel_notify_del_rx_port(vs->sock,
b9adcd69
AD		 1066 (vs->flags & VXLAN_F_GPE) ?
1067 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 1068 UDP_TUNNEL_TYPE_VXLAN);
1c51a915
SH		 1069 spin_unlock(&vn->sock_lock);
1070
544a773a 1071 return true;
d342894c 1072}
1073
b1be00a6
JB		 1074static void vxlan_sock_release(struct vxlan_dev *vxlan)
1075{
c6fcc4fc 1076 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
b1be00a6 1077#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1078 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1079
57d88182 1080 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
544a773a
HFS		 1081#endif
1082
57d88182 1083 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
544a773a
HFS		 1084 synchronize_net();
1085
a53cb29b
MB		 1086 vxlan_vs_del_dev(vxlan);
1087
c6fcc4fc 1088 if (__vxlan_sock_release_prep(sock4)) {
1089 udp_tunnel_sock_release(sock4->sock);
1090 kfree(sock4);
544a773a
HFS		 1091 }
1092
1093#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1094 if (__vxlan_sock_release_prep(sock6)) {
1095 udp_tunnel_sock_release(sock6->sock);
1096 kfree(sock6);
544a773a 1097 }
b1be00a6
JB		 1098#endif
1099}
1100
56ef9c90 1101/* Update multicast group membership when first VNI on
c4b49512 1102 * multicast address is brought up
7c47cedf 1103 */
56ef9c90 1104static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1105{
b1be00a6 1106 struct sock *sk;
e4c7ed41
CW		 1107 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1108 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1109 int ret = -EINVAL;
d342894c 1110
e4c7ed41 1111 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1112 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1113 struct ip_mreqn mreq = {
1114 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1115 .imr_ifindex = ifindex,
1116 };
1117
c6fcc4fc 1118 sk = sock4->sock->sk;
b1be00a6 1119 lock_sock(sk);
56ef9c90 1120 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1121 release_sock(sk);
e4c7ed41
CW		 1122#if IS_ENABLED(CONFIG_IPV6)
1123 } else {
c6fcc4fc 1124 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1125
1126 sk = sock6->sock->sk;
b1be00a6 1127 lock_sock(sk);
56ef9c90
MRL		 1128 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1129 &ip->sin6.sin6_addr);
b1be00a6 1130 release_sock(sk);
e4c7ed41
CW		 1131#endif
1132 }
3fc2de2f 1133
56ef9c90 1134 return ret;
3fc2de2f 1135}
1136
1137/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1138static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1139{
b1be00a6 1140 struct sock *sk;
e4c7ed41
CW		 1141 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1142 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1143 int ret = -EINVAL;
3fc2de2f 1144
e4c7ed41 1145 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1146 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1147 struct ip_mreqn mreq = {
1148 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1149 .imr_ifindex = ifindex,
1150 };
1151
c6fcc4fc 1152 sk = sock4->sock->sk;
b1be00a6 1153 lock_sock(sk);
56ef9c90 1154 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1155 release_sock(sk);
e4c7ed41
CW		 1156#if IS_ENABLED(CONFIG_IPV6)
1157 } else {
c6fcc4fc 1158 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1159
1160 sk = sock6->sock->sk;
b1be00a6 1161 lock_sock(sk);
56ef9c90
MRL		 1162 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1163 &ip->sin6.sin6_addr);
b1be00a6 1164 release_sock(sk);
e4c7ed41
CW		 1165#endif
1166 }
d342894c 1167
56ef9c90 1168 return ret;
d342894c 1169}
1170
f14ecebb
JB		 1171static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1172 struct sk_buff *skb, u32 vxflags)
dfd8645e 1173{
7d34fa75 1174 size_t start, offset;
dfd8645e 1175
f14ecebb
JB		 1176 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1177 goto out;
b7fe10e5 1178
f14ecebb
JB		 1179 start = vxlan_rco_start(unparsed->vx_vni);
1180 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1181
7d34fa75 1182 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1183 return false;
dfd8645e 1184
be5cfeab
JB		 1185 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1186 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1187out:
1188 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1189 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1190 return true;
dfd8645e
TH		 1191}
1192
f14ecebb 1193static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1194 struct sk_buff *skb, u32 vxflags,
10a5af23 1195 struct vxlan_metadata *md)
3288af08 1196{
f14ecebb 1197 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1198 struct metadata_dst *tun_dst;
f14ecebb
JB		 1199
1200 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1201 goto out;
3288af08 1202
3288af08
JB		 1203 md->gbp = ntohs(gbp->policy_id);
1204
10a5af23 1205 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1206 if (tun_dst) {
3288af08 1207 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1208 tun_dst->u.tun_info.options_len = sizeof(*md);
1209 }
3288af08
JB		 1210 if (gbp->dont_learn)
1211 md->gbp |= VXLAN_GBP_DONT_LEARN;
1212
1213 if (gbp->policy_applied)
1214 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1215
64f87d36
JB		 1216 /* In flow-based mode, GBP is carried in dst_metadata */
1217 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1218 skb->mark = md->gbp;
f14ecebb
JB		 1219out:
1220 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1221}
1222
e1e5314d 1223static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
61618eea 1224 __be16 *protocol,
e1e5314d
JB		 1225 struct sk_buff *skb, u32 vxflags)
1226{
1227 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1228
1229 /* Need to have Next Protocol set for interfaces in GPE mode. */
1230 if (!gpe->np_applied)
1231 return false;
1232 /* "The initial version is 0. If a receiver does not support the
1233 * version indicated it MUST drop the packet.
1234 */
1235 if (gpe->version != 0)
1236 return false;
1237 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1238 * processing MUST occur." However, we don't implement OAM
1239 * processing, thus drop the packet.
1240 */
1241 if (gpe->oam_flag)
1242 return false;
1243
1244 switch (gpe->next_protocol) {
1245 case VXLAN_GPE_NP_IPV4:
1246 *protocol = htons(ETH_P_IP);
1247 break;
1248 case VXLAN_GPE_NP_IPV6:
1249 *protocol = htons(ETH_P_IPV6);
1250 break;
1251 case VXLAN_GPE_NP_ETHERNET:
1252 *protocol = htons(ETH_P_TEB);
1253 break;
1254 default:
1255 return false;
1256 }
1257
1258 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1259 return true;
1260}
1261
1ab016e2
JB		 1262static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1263 struct vxlan_sock *vs,
3ad7a4b1 1264 struct sk_buff *skb, __be32 vni)
614732ea 1265{
614732ea 1266 union vxlan_addr saddr;
614732ea 1267
614732ea 1268 skb_reset_mac_header(skb);
614732ea
TG		 1269 skb->protocol = eth_type_trans(skb, vxlan->dev);
1270 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1271
1272 /* Ignore packet loops (and multicast echo) */
1273 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1274 return false;
614732ea 1275
760c6805 1276 /* Get address from the outer IP header */
ce212d0f 1277 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1278 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1279 saddr.sa.sa_family = AF_INET;
1280#if IS_ENABLED(CONFIG_IPV6)
1281 } else {
1ab016e2 1282 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1283 saddr.sa.sa_family = AF_INET6;
1284#endif
1285 }
1286
1ab016e2 1287 if ((vxlan->flags & VXLAN_F_LEARN) &&
3ad7a4b1 1288 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, vni))
1ab016e2
JB		 1289 return false;
1290
1291 return true;
1292}
1293
760c6805
JB		 1294static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1295 struct sk_buff *skb)
1296{
1297 int err = 0;
1298
1299 if (vxlan_get_sk_family(vs) == AF_INET)
1300 err = IP_ECN_decapsulate(oiph, skb);
1301#if IS_ENABLED(CONFIG_IPV6)
1302 else
1303 err = IP6_ECN_decapsulate(oiph, skb);
1304#endif
1305
1306 if (unlikely(err) && log_ecn_error) {
1307 if (vxlan_get_sk_family(vs) == AF_INET)
1308 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1309 &((struct iphdr *)oiph)->saddr,
1310 ((struct iphdr *)oiph)->tos);
1311 else
1312 net_info_ratelimited("non-ECT from %pI6\n",
1313 &((struct ipv6hdr *)oiph)->saddr);
1314 }
1315 return err <= 1;
1316}
1317
d342894c 1318/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1319static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1320{
f2d1968e 1321 struct pcpu_sw_netstats *stats;
c9e78efb 1322 struct vxlan_dev *vxlan;
5cfccc5a 1323 struct vxlan_sock *vs;
f14ecebb 1324 struct vxlanhdr unparsed;
ee122c79
TG		 1325 struct vxlan_metadata _md;
1326 struct vxlan_metadata *md = &_md;
61618eea 1327 __be16 protocol = htons(ETH_P_TEB);
e1e5314d 1328 bool raw_proto = false;
f2d1968e 1329 void *oiph;
3ad7a4b1 1330 __be32 vni = 0;
d342894c 1331
e1e5314d 1332 /* Need UDP and VXLAN header to be present */
7ce04758 1333 if (!pskb_may_pull(skb, VXLAN_HLEN))
e5aed006 1334 goto drop;
d342894c 1335
f14ecebb 1336 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1337 /* VNI flag always required to be set */
1338 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1339 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1340 ntohl(vxlan_hdr(skb)->vx_flags),
1341 ntohl(vxlan_hdr(skb)->vx_vni));
1342 /* Return non vxlan pkt */
e5aed006 1343 goto drop;
d342894c 1344 }
288b01c8
JB		 1345 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1346 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1347
559835ea 1348 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1349 if (!vs)
d342894c 1350 goto drop;
d342894c 1351
3ad7a4b1
RP		 1352 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1353
3ad7a4b1 1354 vxlan = vxlan_vs_find_vni(vs, vni);
c9e78efb
JB		 1355 if (!vxlan)
1356 goto drop;
1357
e1e5314d
JB		 1358 /* For backwards compatibility, only allow reserved fields to be
1359 * used by VXLAN extensions if explicitly requested.
1360 */
1361 if (vs->flags & VXLAN_F_GPE) {
1362 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1363 goto drop;
1364 raw_proto = true;
1365 }
1366
1367 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1368 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1369 goto drop;
c9e78efb 1370
ee122c79 1371 if (vxlan_collect_metadata(vs)) {
10a5af23 1372 struct metadata_dst *tun_dst;
07dabf20 1373
c29a70d2 1374 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
d817f432 1375 key32_to_tunnel_id(vni), sizeof(*md));
c29a70d2 1376
ee122c79
TG		 1377 if (!tun_dst)
1378 goto drop;
1379
0f1b7354 1380 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 1381
1382 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 1383 } else {
1384 memset(md, 0, sizeof(*md));
1385 }
1386
f14ecebb
JB		 1387 if (vs->flags & VXLAN_F_REMCSUM_RX)
1388 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1389 goto drop;
1390 if (vs->flags & VXLAN_F_GBP)
10a5af23 1391 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB		 1392 /* Note that GBP and GPE can never be active together. This is
1393 * ensured in vxlan_dev_configure.
1394 */
3511494c 1395
f14ecebb 1396 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1397 /* If there are any unprocessed flags remaining treat
1398 * this as a malformed packet. This behavior diverges from
1399 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1400 * in reserved fields are to be ignored. The approach here
c4b49512 1401 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1402 * is more robust and provides a little more security in
1403 * adding extensions to VXLAN.
1404 */
288b01c8 1405 goto drop;
3bf39475
TH		 1406 }
1407
e1e5314d 1408 if (!raw_proto) {
3ad7a4b1 1409 if (!vxlan_set_mac(vxlan, vs, skb, vni))
e1e5314d
JB		 1410 goto drop;
1411 } else {
8be0cfa4 1412 skb_reset_mac_header(skb);
e1e5314d
JB		 1413 skb->dev = vxlan->dev;
1414 skb->pkt_type = PACKET_HOST;
1415 }
f2d1968e 1416
f2d1968e
JB		 1417 oiph = skb_network_header(skb);
1418 skb_reset_network_header(skb);
1419
1420 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1421 ++vxlan->dev->stats.rx_frame_errors;
1422 ++vxlan->dev->stats.rx_errors;
1423 goto drop;
1424 }
1425
1426 stats = this_cpu_ptr(vxlan->dev->tstats);
1427 u64_stats_update_begin(&stats->syncp);
1428 stats->rx_packets++;
1429 stats->rx_bytes += skb->len;
1430 u64_stats_update_end(&stats->syncp);
1431
1432 gro_cells_receive(&vxlan->gro_cells, skb);
5cfccc5a
PS		 1433 return 0;
1434
1435drop:
288b01c8
JB		 1436 /* Consume bad packet */
1437 kfree_skb(skb);
1438 return 0;
5cfccc5a
PS		 1439}
1440
3ad7a4b1 1441static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
e4f67add
DS		 1442{
1443 struct vxlan_dev *vxlan = netdev_priv(dev);
1444 struct arphdr *parp;
1445 u8 *arpptr, *sha;
1446 __be32 sip, tip;
1447 struct neighbour *n;
1448
1449 if (dev->flags & IFF_NOARP)
1450 goto out;
1451
1452 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1453 dev->stats.tx_dropped++;
1454 goto out;
1455 }
1456 parp = arp_hdr(skb);
1457
1458 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1459 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1460 parp->ar_pro != htons(ETH_P_IP) ||
1461 parp->ar_op != htons(ARPOP_REQUEST) ||
1462 parp->ar_hln != dev->addr_len ||
1463 parp->ar_pln != 4)
1464 goto out;
1465 arpptr = (u8 *)parp + sizeof(struct arphdr);
1466 sha = arpptr;
1467 arpptr += dev->addr_len; /* sha */
1468 memcpy(&sip, arpptr, sizeof(sip));
1469 arpptr += sizeof(sip);
1470 arpptr += dev->addr_len; /* tha */
1471 memcpy(&tip, arpptr, sizeof(tip));
1472
1473 if (ipv4_is_loopback(tip) ||
1474 ipv4_is_multicast(tip))
1475 goto out;
1476
1477 n = neigh_lookup(&arp_tbl, &tip, dev);
1478
1479 if (n) {
e4f67add
DS		 1480 struct vxlan_fdb *f;
1481 struct sk_buff *reply;
1482
1483 if (!(n->nud_state & NUD_CONNECTED)) {
1484 neigh_release(n);
1485 goto out;
1486 }
1487
3ad7a4b1 1488 f = vxlan_find_mac(vxlan, n->ha, vni);
e4c7ed41 1489 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1490 /* bridge-local neighbor */
1491 neigh_release(n);
1492 goto out;
1493 }
1494
1495 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1496 n->ha, sha);
1497
1498 neigh_release(n);
1499
7346135d
DS		 1500 if (reply == NULL)
1501 goto out;
1502
e4f67add
DS		 1503 skb_reset_mac_header(reply);
1504 __skb_pull(reply, skb_network_offset(reply));
1505 reply->ip_summed = CHECKSUM_UNNECESSARY;
1506 reply->pkt_type = PACKET_HOST;
1507
1508 if (netif_rx_ni(reply) == NET_RX_DROP)
1509 dev->stats.rx_dropped++;
e4c7ed41
CW		 1510 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1511 union vxlan_addr ipa = {
1512 .sin.sin_addr.s_addr = tip,
a45e92a5 1513 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1514 };
1515
1516 vxlan_ip_miss(dev, &ipa);
1517 }
e4f67add
DS		 1518out:
1519 consume_skb(skb);
1520 return NETDEV_TX_OK;
1521}
1522
f564f45c 1523#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1524static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1525 struct neighbour *n, bool isrouter)
1526{
1527 struct net_device *dev = request->dev;
1528 struct sk_buff *reply;
1529 struct nd_msg *ns, *na;
1530 struct ipv6hdr *pip6;
1531 u8 *daddr;
1532 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1533 int ns_olen;
1534 int i, len;
1535
f1fb08f6 1536 if (dev == NULL || !pskb_may_pull(request, request->len))
4b29dba9
DS		 1537 return NULL;
1538
1539 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1540 sizeof(*na) + na_olen + dev->needed_tailroom;
1541 reply = alloc_skb(len, GFP_ATOMIC);
1542 if (reply == NULL)
1543 return NULL;
1544
1545 reply->protocol = htons(ETH_P_IPV6);
1546 reply->dev = dev;
1547 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1548 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1549 skb_reset_mac_header(reply);
4b29dba9 1550
f1fb08f6 1551 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
4b29dba9
DS		 1552
1553 daddr = eth_hdr(request)->h_source;
f1fb08f6
VB		 1554 ns_olen = request->len - skb_network_offset(request) -
1555 sizeof(struct ipv6hdr) - sizeof(*ns);
4b29dba9
DS		 1556 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1557 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1558 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1559 break;
1560 }
1561 }
1562
1563 /* Ethernet header */
1564 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1565 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1566 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1567 reply->protocol = htons(ETH_P_IPV6);
1568
1569 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1570 skb_reset_network_header(reply);
4b29dba9
DS		 1571 skb_put(reply, sizeof(struct ipv6hdr));
1572
1573 /* IPv6 header */
1574
1575 pip6 = ipv6_hdr(reply);
1576 memset(pip6, 0, sizeof(struct ipv6hdr));
1577 pip6->version = 6;
1578 pip6->priority = ipv6_hdr(request)->priority;
1579 pip6->nexthdr = IPPROTO_ICMPV6;
1580 pip6->hop_limit = 255;
1581 pip6->daddr = ipv6_hdr(request)->saddr;
1582 pip6->saddr = *(struct in6_addr *)n->primary_key;
1583
1584 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1585 skb_reset_transport_header(reply);
4b29dba9
DS		 1586
1587 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1588
1589 /* Neighbor Advertisement */
1590 memset(na, 0, sizeof(*na)+na_olen);
1591 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1592 na->icmph.icmp6_router = isrouter;
1593 na->icmph.icmp6_override = 1;
1594 na->icmph.icmp6_solicited = 1;
1595 na->target = ns->target;
1596 ether_addr_copy(&na->opt[2], n->ha);
1597 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1598 na->opt[1] = na_olen >> 3;
1599
1600 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1601 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1602 csum_partial(na, sizeof(*na)+na_olen, 0));
1603
1604 pip6->payload_len = htons(sizeof(*na)+na_olen);
1605
1606 skb_push(reply, sizeof(struct ipv6hdr));
1607
1608 reply->ip_summed = CHECKSUM_UNNECESSARY;
1609
1610 return reply;
1611}
1612
3ad7a4b1 1613static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
f564f45c
CW		 1614{
1615 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1616 struct nd_msg *msg;
f564f45c 1617 const struct ipv6hdr *iphdr;
8dcd81a9 1618 const struct in6_addr *daddr;
4b29dba9
DS		 1619 struct neighbour *n;
1620 struct inet6_dev *in6_dev;
f564f45c
CW		 1621
1622 in6_dev = __in6_dev_get(dev);
1623 if (!in6_dev)
1624 goto out;
1625
f1fb08f6
VB		 1626 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + sizeof(struct nd_msg)))
1627 goto out;
1628
f564f45c 1629 iphdr = ipv6_hdr(skb);
f564f45c
CW		 1630 daddr = &iphdr->daddr;
1631
f1fb08f6 1632 msg = (struct nd_msg *)(iphdr + 1);
f564f45c
CW		 1633 if (msg->icmph.icmp6_code != 0 ||
1634 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1635 goto out;
1636
4b29dba9
DS		 1637 if (ipv6_addr_loopback(daddr) ||
1638 ipv6_addr_is_multicast(&msg->target))
1639 goto out;
1640
1641 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1642
1643 if (n) {
1644 struct vxlan_fdb *f;
4b29dba9 1645 struct sk_buff *reply;
f564f45c
CW		 1646
1647 if (!(n->nud_state & NUD_CONNECTED)) {
1648 neigh_release(n);
1649 goto out;
1650 }
1651
3ad7a4b1 1652 f = vxlan_find_mac(vxlan, n->ha, vni);
f564f45c
CW		 1653 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1654 /* bridge-local neighbor */
1655 neigh_release(n);
1656 goto out;
1657 }
1658
4b29dba9
DS		 1659 reply = vxlan_na_create(skb, n,
1660 !!(f ? f->flags & NTF_ROUTER : 0));
1661
f564f45c 1662 neigh_release(n);
4b29dba9
DS		 1663
1664 if (reply == NULL)
1665 goto out;
1666
1667 if (netif_rx_ni(reply) == NET_RX_DROP)
1668 dev->stats.rx_dropped++;
1669
f564f45c 1670 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1671 union vxlan_addr ipa = {
1672 .sin6.sin6_addr = msg->target,
a45e92a5 1673 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1674 };
1675
f564f45c
CW		 1676 vxlan_ip_miss(dev, &ipa);
1677 }
1678
1679out:
1680 consume_skb(skb);
1681 return NETDEV_TX_OK;
1682}
1683#endif
1684
e4f67add
DS		 1685static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1686{
1687 struct vxlan_dev *vxlan = netdev_priv(dev);
1688 struct neighbour *n;
e4f67add
DS		 1689
1690 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1691 return false;
1692
1693 n = NULL;
1694 switch (ntohs(eth_hdr(skb)->h_proto)) {
1695 case ETH_P_IP:
e15a00aa
CW		 1696 {
1697 struct iphdr *pip;
1698
e4f67add
DS		 1699 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1700 return false;
1701 pip = ip_hdr(skb);
1702 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1703 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1704 union vxlan_addr ipa = {
1705 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1706 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1707 };
1708
1709 vxlan_ip_miss(dev, &ipa);
1710 return false;
1711 }
1712
e4f67add 1713 break;
e15a00aa
CW		 1714 }
1715#if IS_ENABLED(CONFIG_IPV6)
1716 case ETH_P_IPV6:
1717 {
1718 struct ipv6hdr *pip6;
1719
1720 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1721 return false;
1722 pip6 = ipv6_hdr(skb);
1723 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1724 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1725 union vxlan_addr ipa = {
1726 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1727 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1728 };
1729
1730 vxlan_ip_miss(dev, &ipa);
1731 return false;
1732 }
1733
1734 break;
1735 }
1736#endif
e4f67add
DS		 1737 default:
1738 return false;
1739 }
1740
1741 if (n) {
1742 bool diff;
1743
7367d0b5 1744 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1745 if (diff) {
1746 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1747 dev->addr_len);
1748 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1749 }
1750 neigh_release(n);
1751 return diff;
e4c7ed41
CW		 1752 }
1753
e4f67add
DS		 1754 return false;
1755}
1756
af33c1ad 1757static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1758 struct vxlan_metadata *md)
1759{
1760 struct vxlanhdr_gbp *gbp;
1761
db79a621
TG		 1762 if (!md->gbp)
1763 return;
1764
3511494c 1765 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1766 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 1767
1768 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1769 gbp->dont_learn = 1;
1770
1771 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1772 gbp->policy_applied = 1;
1773
1774 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1775}
1776
e1e5314d
JB		 1777static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1778 __be16 protocol)
1779{
1780 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1781
1782 gpe->np_applied = 1;
1783
1784 switch (protocol) {
1785 case htons(ETH_P_IP):
1786 gpe->next_protocol = VXLAN_GPE_NP_IPV4;
1787 return 0;
1788 case htons(ETH_P_IPV6):
1789 gpe->next_protocol = VXLAN_GPE_NP_IPV6;
1790 return 0;
1791 case htons(ETH_P_TEB):
1792 gpe->next_protocol = VXLAN_GPE_NP_ETHERNET;
1793 return 0;
1794 }
1795 return -EPFNOSUPPORT;
1796}
1797
f491e56d
JB		 1798static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1799 int iphdr_len, __be32 vni,
1800 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1801 bool udp_sum)
e4c7ed41 1802{
e4c7ed41 1803 struct vxlanhdr *vxh;
e4c7ed41
CW		 1804 int min_headroom;
1805 int err;
dfd8645e 1806 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 1807 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 1808
af33c1ad 1809 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1810 skb->ip_summed == CHECKSUM_PARTIAL) {
1811 int csum_start = skb_checksum_start_offset(skb);
1812
1813 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1814 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1815 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1816 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1817 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1818 }
e4c7ed41 1819
e4c7ed41 1820 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
4a4f86cc 1821 + VXLAN_HLEN + iphdr_len;
649c5b8b
PS		 1822
1823 /* Need space for new headers (invalidates iph ptr) */
1824 err = skb_cow_head(skb, min_headroom);
e1e5314d 1825 if (unlikely(err))
c46b7897 1826 return err;
649c5b8b 1827
aed069df
AD		 1828 err = iptunnel_handle_offloads(skb, type);
1829 if (err)
c46b7897 1830 return err;
b736a623 1831
49560532 1832 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 1833 vxh->vx_flags = VXLAN_HF_VNI;
1834 vxh->vx_vni = vxlan_vni_field(vni);
49560532 1835
dfd8645e 1836 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 1837 unsigned int start;
dfd8645e 1838
54bfd872
JB		 1839 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1840 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1841 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 1842
1843 if (!skb_is_gso(skb)) {
1844 skb->ip_summed = CHECKSUM_NONE;
1845 skb->encapsulation = 0;
1846 }
1847 }
1848
af33c1ad
TH		 1849 if (vxflags & VXLAN_F_GBP)
1850 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB		 1851 if (vxflags & VXLAN_F_GPE) {
1852 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1853 if (err < 0)
c46b7897 1854 return err;
e1e5314d
JB		 1855 inner_protocol = skb->protocol;
1856 }
3511494c 1857
e1e5314d 1858 skb_set_inner_protocol(skb, inner_protocol);
039f5062 1859 return 0;
49560532 1860}
49560532 1861
655c3de1 1862static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
1863 struct vxlan_sock *sock4,
1a8496ba 1864 struct sk_buff *skb, int oif, u8 tos,
4ecb1d83 1865 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
0c1d70af 1866 struct dst_cache *dst_cache,
db3c6139 1867 const struct ip_tunnel_info *info)
1a8496ba 1868{
db3c6139 1869 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 1870 struct rtable *rt = NULL;
1871 struct flowi4 fl4;
1872
655c3de1 1873 if (!sock4)
1874 return ERR_PTR(-EIO);
1875
db3c6139
DB		 1876 if (tos && !info)
1877 use_cache = false;
1878 if (use_cache) {
0c1d70af
PA		 1879 rt = dst_cache_get_ip4(dst_cache, saddr);
1880 if (rt)
1881 return rt;
1882 }
1883
1a8496ba
JB		 1884 memset(&fl4, 0, sizeof(fl4));
1885 fl4.flowi4_oif = oif;
1886 fl4.flowi4_tos = RT_TOS(tos);
1887 fl4.flowi4_mark = skb->mark;
1888 fl4.flowi4_proto = IPPROTO_UDP;
1889 fl4.daddr = daddr;
272d96a5 1890 fl4.saddr = *saddr;
4ecb1d83
MP		 1891 fl4.fl4_dport = dport;
1892 fl4.fl4_sport = sport;
1a8496ba
JB		 1893
1894 rt = ip_route_output_key(vxlan->net, &fl4);
655c3de1 1895 if (likely(!IS_ERR(rt))) {
1896 if (rt->dst.dev == dev) {
1897 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
1898 ip_rt_put(rt);
1899 return ERR_PTR(-ELOOP);
1900 }
1901
1a8496ba 1902 *saddr = fl4.saddr;
0c1d70af
PA		 1903 if (use_cache)
1904 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
655c3de1 1905 } else {
1906 netdev_dbg(dev, "no route to %pI4\n", &daddr);
1907 return ERR_PTR(-ENETUNREACH);
0c1d70af 1908 }
1a8496ba
JB		 1909 return rt;
1910}
1911
e5d4b29f
JB		 1912#if IS_ENABLED(CONFIG_IPV6)
1913static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
655c3de1 1914 struct net_device *dev,
03dc52a8 1915 struct vxlan_sock *sock6,
1400615d 1916 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 1917 __be32 label,
e5d4b29f 1918 const struct in6_addr *daddr,
0c1d70af 1919 struct in6_addr *saddr,
4ecb1d83 1920 __be16 dport, __be16 sport,
db3c6139
DB		 1921 struct dst_cache *dst_cache,
1922 const struct ip_tunnel_info *info)
e5d4b29f 1923{
db3c6139 1924 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 1925 struct dst_entry *ndst;
1926 struct flowi6 fl6;
1927 int err;
1928
c6fcc4fc 1929 if (!sock6)
1930 return ERR_PTR(-EIO);
1931
1400615d
DB		 1932 if (tos && !info)
1933 use_cache = false;
db3c6139 1934 if (use_cache) {
0c1d70af
PA		 1935 ndst = dst_cache_get_ip6(dst_cache, saddr);
1936 if (ndst)
1937 return ndst;
1938 }
1939
e5d4b29f
JB		 1940 memset(&fl6, 0, sizeof(fl6));
1941 fl6.flowi6_oif = oif;
1942 fl6.daddr = *daddr;
272d96a5 1943 fl6.saddr = *saddr;
eaa93bf4 1944 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB		 1945 fl6.flowi6_mark = skb->mark;
1946 fl6.flowi6_proto = IPPROTO_UDP;
4ecb1d83
MP		 1947 fl6.fl6_dport = dport;
1948 fl6.fl6_sport = sport;
e5d4b29f
JB		 1949
1950 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
c6fcc4fc 1951 sock6->sock->sk,
e5d4b29f 1952 &ndst, &fl6);
655c3de1 1953 if (unlikely(err < 0)) {
1954 netdev_dbg(dev, "no route to %pI6\n", daddr);
1955 return ERR_PTR(-ENETUNREACH);
1956 }
1957
1958 if (unlikely(ndst->dev == dev)) {
1959 netdev_dbg(dev, "circular route to %pI6\n", daddr);
1960 dst_release(ndst);
1961 return ERR_PTR(-ELOOP);
1962 }
e5d4b29f
JB		 1963
1964 *saddr = fl6.saddr;
db3c6139 1965 if (use_cache)
0c1d70af 1966 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 1967 return ndst;
1968}
1969#endif
1970
9dcc71e1
SS		 1971/* Bypass encapsulation if the destination is local */
1972static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
3ad7a4b1 1973 struct vxlan_dev *dst_vxlan, __be32 vni)
9dcc71e1 1974{
8f84985f 1975 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1976 union vxlan_addr loopback;
1977 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1978 struct net_device *dev = skb->dev;
1979 int len = skb->len;
9dcc71e1 1980
8f84985f
LR		 1981 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1982 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1983 skb->pkt_type = PACKET_HOST;
1984 skb->encapsulation = 0;
1985 skb->dev = dst_vxlan->dev;
1986 __skb_pull(skb, skb_network_offset(skb));
1987
e4c7ed41
CW		 1988 if (remote_ip->sa.sa_family == AF_INET) {
1989 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1990 loopback.sa.sa_family = AF_INET;
1991#if IS_ENABLED(CONFIG_IPV6)
1992 } else {
1993 loopback.sin6.sin6_addr = in6addr_loopback;
1994 loopback.sa.sa_family = AF_INET6;
1995#endif
1996 }
1997
9dcc71e1 1998 if (dst_vxlan->flags & VXLAN_F_LEARN)
3ad7a4b1 1999 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source, vni);
9dcc71e1
SS		 2000
2001 u64_stats_update_begin(&tx_stats->syncp);
2002 tx_stats->tx_packets++;
ce6502a8 2003 tx_stats->tx_bytes += len;
9dcc71e1
SS		 2004 u64_stats_update_end(&tx_stats->syncp);
2005
2006 if (netif_rx(skb) == NET_RX_SUCCESS) {
2007 u64_stats_update_begin(&rx_stats->syncp);
2008 rx_stats->rx_packets++;
ce6502a8 2009 rx_stats->rx_bytes += len;
9dcc71e1
SS		 2010 u64_stats_update_end(&rx_stats->syncp);
2011 } else {
ce6502a8 2012 dev->stats.rx_dropped++;
9dcc71e1
SS		 2013 }
2014}
2015
fee1fad7 2016static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2017 struct vxlan_dev *vxlan, union vxlan_addr *daddr,
1f6cc07e 2018 __be16 dst_port, __be32 vni, struct dst_entry *dst,
fee1fad7 2019 u32 rt_flags)
2020{
2021#if IS_ENABLED(CONFIG_IPV6)
2022 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2023 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2024 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2025 */
2026 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2027#endif
2028 /* Bypass encapsulation if the destination is local */
2029 if (rt_flags & RTCF_LOCAL &&
2030 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2031 struct vxlan_dev *dst_vxlan;
2032
2033 dst_release(dst);
2034 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
2035 daddr->sa.sa_family, dst_port,
2036 vxlan->flags);
2037 if (!dst_vxlan) {
2038 dev->stats.tx_errors++;
2039 kfree_skb(skb);
2040
2041 return -ENOENT;
2042 }
3ad7a4b1 2043 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
fee1fad7 2044 return 1;
2045 }
2046
2047 return 0;
2048}
2049
4ad16930 2050static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
3ad7a4b1
RP		 2051 __be32 default_vni, struct vxlan_rdst *rdst,
2052 bool did_rsc)
d342894c 2053{
d71785ff 2054 struct dst_cache *dst_cache;
3093fbe7 2055 struct ip_tunnel_info *info;
d342894c 2056 struct vxlan_dev *vxlan = netdev_priv(dev);
0770b53b 2057 const struct iphdr *old_iph = ip_hdr(skb);
e4c7ed41 2058 union vxlan_addr *dst;
272d96a5 2059 union vxlan_addr remote_ip, local_ip;
ee122c79
TG		 2060 struct vxlan_metadata _md;
2061 struct vxlan_metadata *md = &_md;
e4c7ed41 2062 __be16 src_port = 0, dst_port;
655c3de1 2063 struct dst_entry *ndst = NULL;
e7f70af1 2064 __be32 vni, label;
d342894c 2065 __u8 tos, ttl;
0e6fbc5b 2066 int err;
ee122c79 2067 u32 flags = vxlan->flags;
b4ed5cad 2068 bool udp_sum = false;
f491e56d 2069 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 2070
61adedf3 2071 info = skb_tunnel_info(skb);
3093fbe7 2072
ee122c79 2073 if (rdst) {
0770b53b 2074 dst = &rdst->remote_ip;
2075 if (vxlan_addr_any(dst)) {
2076 if (did_rsc) {
2077 /* short-circuited back to local bridge */
3ad7a4b1 2078 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
0770b53b 2079 return;
2080 }
2081 goto drop;
2082 }
2083
0dfbdf41 2084 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
3ad7a4b1 2085 vni = (rdst->remote_vni) ? : default_vni;
1158632b 2086 local_ip = vxlan->cfg.saddr;
d71785ff 2087 dst_cache = &rdst->dst_cache;
0770b53b 2088 md->gbp = skb->mark;
2089 ttl = vxlan->cfg.ttl;
2090 if (!ttl && vxlan_addr_multicast(dst))
2091 ttl = 1;
2092
2093 tos = vxlan->cfg.tos;
2094 if (tos == 1)
2095 tos = ip_tunnel_get_dsfield(old_iph, skb);
2096
2097 if (dst->sa.sa_family == AF_INET)
2098 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2099 else
2100 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2101 label = vxlan->cfg.label;
ee122c79
TG		 2102 } else {
2103 if (!info) {
2104 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2105 dev->name);
2106 goto drop;
2107 }
b1be00a6 2108 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
272d96a5 2109 if (remote_ip.sa.sa_family == AF_INET) {
a725e514 2110 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
272d96a5 2111 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2112 } else {
a725e514 2113 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
272d96a5 2114 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2115 }
ee122c79 2116 dst = &remote_ip;
0770b53b 2117 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2118 vni = tunnel_id_to_key32(info->key.tun_id);
d71785ff 2119 dst_cache = &info->dst_cache;
0770b53b 2120 if (info->options_len)
2121 md = ip_tunnel_info_opts(info);
a725e514
JB		 2122 ttl = info->key.ttl;
2123 tos = info->key.tos;
e7f70af1 2124 label = info->key.label;
b4ed5cad 2125 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514 2126 }
0770b53b 2127 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2128 vxlan->cfg.port_max, true);
a725e514 2129
56de859e 2130 rcu_read_lock();
e4c7ed41 2131 if (dst->sa.sa_family == AF_INET) {
c6fcc4fc 2132 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
c46b7897 2133 struct rtable *rt;
0770b53b 2134 __be16 df = 0;
c6fcc4fc 2135
655c3de1 2136 rt = vxlan_get_route(vxlan, dev, sock4, skb,
1a8496ba 2137 rdst ? rdst->remote_ifindex : 0, tos,
272d96a5 2138 dst->sin.sin_addr.s_addr,
1158632b 2139 &local_ip.sin.sin_addr.s_addr,
4ecb1d83 2140 dst_port, src_port,
d71785ff 2141 dst_cache, info);
8ebd115b
DM		 2142 if (IS_ERR(rt)) {
2143 err = PTR_ERR(rt);
c46b7897 2144 goto tx_error;
8ebd115b 2145 }
e4c7ed41
CW		 2146
2147 /* Bypass encapsulation if the destination is local */
fee1fad7 2148 if (!info) {
2149 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2150 dst_port, vni, &rt->dst,
2151 rt->rt_flags);
2152 if (err)
56de859e 2153 goto out_unlock;
fee1fad7 2154 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
6ceb31ca 2155 df = htons(IP_DF);
fee1fad7 2156 }
6ceb31ca 2157
c46b7897 2158 ndst = &rt->dst;
e4c7ed41
CW		 2159 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2160 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
c46b7897 2161 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
54bfd872 2162 vni, md, flags, udp_sum);
f491e56d 2163 if (err < 0)
c46b7897 2164 goto tx_error;
f491e56d 2165
1158632b 2166 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
f491e56d
JB		 2167 dst->sin.sin_addr.s_addr, tos, ttl, df,
2168 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2169#if IS_ENABLED(CONFIG_IPV6)
2170 } else {
c6fcc4fc 2171 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
e4c7ed41 2172
655c3de1 2173 ndst = vxlan6_get_route(vxlan, dev, sock6, skb,
1400615d 2174 rdst ? rdst->remote_ifindex : 0, tos,
272d96a5 2175 label, &dst->sin6.sin6_addr,
1158632b 2176 &local_ip.sin6.sin6_addr,
4ecb1d83 2177 dst_port, src_port,
db3c6139 2178 dst_cache, info);
e5d4b29f 2179 if (IS_ERR(ndst)) {
8ebd115b 2180 err = PTR_ERR(ndst);
c46b7897 2181 ndst = NULL;
9dcc71e1 2182 goto tx_error;
e4c7ed41 2183 }
655c3de1 2184
fee1fad7 2185 if (!info) {
2186 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
49560532 2187
fee1fad7 2188 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2189 dst_port, vni, ndst,
2190 rt6i_flags);
2191 if (err)
56de859e 2192 goto out_unlock;
fee1fad7 2193 }
35e2d115 2194
1400615d 2195 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2196 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2197 skb_scrub_packet(skb, xnet);
2198 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2199 vni, md, flags, udp_sum);
c46b7897 2200 if (err < 0)
2201 goto tx_error;
2202
0770b53b 2203 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
1158632b 2204 &local_ip.sin6.sin6_addr,
272d96a5 2205 &dst->sin6.sin6_addr, tos, ttl,
e7f70af1 2206 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2207#endif
2208 }
56de859e
JK		 2209out_unlock:
2210 rcu_read_unlock();
4ad16930 2211 return;
d342894c 2212
2213drop:
2214 dev->stats.tx_dropped++;
c46b7897 2215 dev_kfree_skb(skb);
2216 return;
d342894c 2217
2218tx_error:
56de859e 2219 rcu_read_unlock();
655c3de1 2220 if (err == -ELOOP)
2221 dev->stats.collisions++;
2222 else if (err == -ENETUNREACH)
2223 dev->stats.tx_carrier_errors++;
c46b7897 2224 dst_release(ndst);
d342894c 2225 dev->stats.tx_errors++;
c46b7897 2226 kfree_skb(skb);
d342894c 2227}
2228
6681712d
DS		 2229/* Transmit local packets over Vxlan
2230 *
2231 * Outer IP header inherits ECN and DF from inner header.
2232 * Outer UDP destination is the VXLAN assigned port.
2233 * source port is based on hash of flow
2234 */
2235static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2236{
2237 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2238 const struct ip_tunnel_info *info;
6681712d
DS		 2239 struct ethhdr *eth;
2240 bool did_rsc = false;
8f646c92 2241 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2242 struct vxlan_fdb *f;
3ad7a4b1 2243 __be32 vni = 0;
6681712d 2244
61adedf3 2245 info = skb_tunnel_info(skb);
3093fbe7 2246
6681712d 2247 skb_reset_mac_header(skb);
6681712d 2248
47e5d1b0 2249 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 2250 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2251 info->mode & IP_TUNNEL_INFO_TX) {
2252 vni = tunnel_id_to_key32(info->key.tun_id);
2253 } else {
2254 if (info && info->mode & IP_TUNNEL_INFO_TX)
2255 vxlan_xmit_one(skb, dev, vni, NULL, false);
2256 else
2257 kfree_skb(skb);
2258 return NETDEV_TX_OK;
2259 }
47e5d1b0
JB		 2260 }
2261
2262 if (vxlan->flags & VXLAN_F_PROXY) {
2263 eth = eth_hdr(skb);
f564f45c 2264 if (ntohs(eth->h_proto) == ETH_P_ARP)
3ad7a4b1 2265 return arp_reduce(dev, skb, vni);
f564f45c 2266#if IS_ENABLED(CONFIG_IPV6)
f1fb08f6
VB		 2267 else if (ntohs(eth->h_proto) == ETH_P_IPV6) {
2268 struct ipv6hdr *hdr, _hdr;
2269 if ((hdr = skb_header_pointer(skb,
2270 skb_network_offset(skb),
2271 sizeof(_hdr), &_hdr)) &&
2272 hdr->nexthdr == IPPROTO_ICMPV6)
2273 return neigh_reduce(dev, skb, vni);
f564f45c
CW		 2274 }
2275#endif
2276 }
6681712d 2277
47e5d1b0 2278 eth = eth_hdr(skb);
3ad7a4b1 2279 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2280 did_rsc = false;
2281
2282 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2283 (ntohs(eth->h_proto) == ETH_P_IP ||
2284 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2285 did_rsc = route_shortcircuit(dev, skb);
2286 if (did_rsc)
3ad7a4b1 2287 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2288 }
2289
6681712d 2290 if (f == NULL) {
3ad7a4b1 2291 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae
MR		 2292 if (f == NULL) {
2293 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2294 !is_multicast_ether_addr(eth->h_dest))
2295 vxlan_fdb_miss(vxlan, eth->h_dest);
2296
2297 dev->stats.tx_dropped++;
8f646c92 2298 kfree_skb(skb);
afbd8bae
MR		 2299 return NETDEV_TX_OK;
2300 }
2301 }
6681712d 2302
afbd8bae
MR		 2303 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2304 struct sk_buff *skb1;
6681712d 2305
8f646c92
ED		 2306 if (!fdst) {
2307 fdst = rdst;
2308 continue;
2309 }
afbd8bae
MR		 2310 skb1 = skb_clone(skb, GFP_ATOMIC);
2311 if (skb1)
3ad7a4b1 2312 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
6681712d
DS		 2313 }
2314
8f646c92 2315 if (fdst)
3ad7a4b1 2316 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
8f646c92
ED		 2317 else
2318 kfree_skb(skb);
4ad16930 2319 return NETDEV_TX_OK;
6681712d
DS		 2320}
2321
d342894c 2322/* Walk the forwarding table and purge stale entries */
2323static void vxlan_cleanup(unsigned long arg)
2324{
2325 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2326 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2327 unsigned int h;
2328
2329 if (!netif_running(vxlan->dev))
2330 return;
2331
d342894c 2332 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2333 struct hlist_node *p, *n;
14e1d0fa
SD		 2334
2335 spin_lock_bh(&vxlan->hash_lock);
d342894c 2336 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2337 struct vxlan_fdb *f
2338 = container_of(p, struct vxlan_fdb, hlist);
2339 unsigned long timeout;
2340
efb5f68f 2341 if (f->state & (NUD_PERMANENT | NUD_NOARP))
d342894c 2342 continue;
2343
def499c9
RP		 2344 if (f->flags & NTF_EXT_LEARNED)
2345 continue;
2346
0dfbdf41 2347 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2348 if (time_before_eq(timeout, jiffies)) {
2349 netdev_dbg(vxlan->dev,
2350 "garbage collect %pM\n",
2351 f->eth_addr);
2352 f->state = NUD_STALE;
2353 vxlan_fdb_destroy(vxlan, f);
2354 } else if (time_before(timeout, next_timer))
2355 next_timer = timeout;
2356 }
14e1d0fa 2357 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2358 }
d342894c 2359
2360 mod_timer(&vxlan->age_timer, next_timer);
2361}
2362
a53cb29b
MB		 2363static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2364{
2365 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2366
2367 spin_lock(&vn->sock_lock);
2368 hlist_del_init_rcu(&vxlan->hlist);
2369 spin_unlock(&vn->sock_lock);
2370}
2371
9c2e24e1
PS		 2372static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2373{
56ef9c90 2374 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2375 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2376
56ef9c90 2377 spin_lock(&vn->sock_lock);
9c2e24e1 2378 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2379 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2380}
2381
d342894c 2382/* Setup stats when device is created */
2383static int vxlan_init(struct net_device *dev)
2384{
1c213bd2 2385 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2386 if (!dev->tstats)
d342894c 2387 return -ENOMEM;
2388
2389 return 0;
2390}
2391
3ad7a4b1 2392static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
afbd8bae
MR		 2393{
2394 struct vxlan_fdb *f;
2395
2396 spin_lock_bh(&vxlan->hash_lock);
3ad7a4b1 2397 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae
MR		 2398 if (f)
2399 vxlan_fdb_destroy(vxlan, f);
2400 spin_unlock_bh(&vxlan->hash_lock);
2401}
2402
ebf4063e
SH		 2403static void vxlan_uninit(struct net_device *dev)
2404{
2405 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2406
3ad7a4b1 2407 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
afbd8bae 2408
ebf4063e
SH		 2409 free_percpu(dev->tstats);
2410}
2411
d342894c 2412/* Start ageing timer and join group when device is brought up */
2413static int vxlan_open(struct net_device *dev)
2414{
2415 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2416 int ret;
56ef9c90 2417
205f356d
JB		 2418 ret = vxlan_sock_add(vxlan);
2419 if (ret < 0)
2420 return ret;
d342894c 2421
79d4a94f 2422 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2423 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2424 if (ret == -EADDRINUSE)
2425 ret = 0;
56ef9c90 2426 if (ret) {
205f356d 2427 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2428 return ret;
2429 }
d342894c 2430 }
2431
0dfbdf41 2432 if (vxlan->cfg.age_interval)
d342894c 2433 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2434
56ef9c90 2435 return ret;
d342894c 2436}
2437
2438/* Purge the forwarding table */
8b3f9337 2439static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
d342894c 2440{
31fec5aa 2441 unsigned int h;
d342894c 2442
2443 spin_lock_bh(&vxlan->hash_lock);
2444 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2445 struct hlist_node *p, *n;
2446 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2447 struct vxlan_fdb *f
2448 = container_of(p, struct vxlan_fdb, hlist);
8b3f9337
RP		 2449 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2450 continue;
afbd8bae
MR		 2451 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2452 if (!is_zero_ether_addr(f->eth_addr))
2453 vxlan_fdb_destroy(vxlan, f);
d342894c 2454 }
2455 }
2456 spin_unlock_bh(&vxlan->hash_lock);
2457}
2458
2459/* Cleanup timer and forwarding table on shutdown */
2460static int vxlan_stop(struct net_device *dev)
2461{
2462 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2463 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2464 int ret = 0;
d342894c 2465
24c0e683 2466 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2467 !vxlan_group_used(vn, vxlan))
56ef9c90 2468 ret = vxlan_igmp_leave(vxlan);
d342894c 2469
2470 del_timer_sync(&vxlan->age_timer);
2471
8b3f9337 2472 vxlan_flush(vxlan, false);
205f356d 2473 vxlan_sock_release(vxlan);
d342894c 2474
56ef9c90 2475 return ret;
d342894c 2476}
2477
d342894c 2478/* Stub, nothing needs to be done. */
2479static void vxlan_set_multicast_list(struct net_device *dev)
2480{
2481}
2482
91572088 2483static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
345010b5 2484{
91572088
JW		 2485 struct vxlan_dev *vxlan = netdev_priv(dev);
2486 struct vxlan_rdst *dst = &vxlan->default_dst;
2487 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2488 dst->remote_ifindex);
2489 bool use_ipv6 = false;
345010b5
DB		 2490
2491 if (dst->remote_ip.sa.sa_family == AF_INET6)
91572088 2492 use_ipv6 = true;
345010b5 2493
91572088
JW		 2494 /* This check is different than dev->max_mtu, because it looks at
2495 * the lowerdev->mtu, rather than the static dev->max_mtu
2496 */
2497 if (lowerdev) {
2498 int max_mtu = lowerdev->mtu -
2499 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2500 if (new_mtu > max_mtu)
72564b59 2501 return -EINVAL;
72564b59
DW		 2502 }
2503
345010b5
DB		 2504 dev->mtu = new_mtu;
2505 return 0;
2506}
2507
fc4099f1
PS		 2508static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2509{
2510 struct vxlan_dev *vxlan = netdev_priv(dev);
2511 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2512 __be16 sport, dport;
2513
2514 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2515 vxlan->cfg.port_max, true);
2516 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2517
239e944f 2518 if (ip_tunnel_info_af(info) == AF_INET) {
c6fcc4fc 2519 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
1a8496ba
JB		 2520 struct rtable *rt;
2521
655c3de1 2522 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
1a8496ba 2523 info->key.u.ipv4.dst,
22f0708a
PA		 2524 &info->key.u.ipv4.src, dport, sport,
2525 &info->dst_cache, info);
1a8496ba
JB		 2526 if (IS_ERR(rt))
2527 return PTR_ERR(rt);
2528 ip_rt_put(rt);
239e944f
JB		 2529 } else {
2530#if IS_ENABLED(CONFIG_IPV6)
03dc52a8 2531 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
239e944f
JB		 2532 struct dst_entry *ndst;
2533
655c3de1 2534 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
e7f70af1 2535 info->key.label, &info->key.u.ipv6.dst,
22f0708a
PA		 2536 &info->key.u.ipv6.src, dport, sport,
2537 &info->dst_cache, info);
239e944f
JB		 2538 if (IS_ERR(ndst))
2539 return PTR_ERR(ndst);
2540 dst_release(ndst);
239e944f
JB		 2541#else /* !CONFIG_IPV6 */
2542 return -EPFNOSUPPORT;
2543#endif
2544 }
1a8496ba
JB		 2545 info->key.tp_src = sport;
2546 info->key.tp_dst = dport;
239e944f 2547 return 0;
fc4099f1
PS		 2548}
2549
0c867c9b 2550static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 2551 .ndo_init = vxlan_init,
ebf4063e 2552 .ndo_uninit = vxlan_uninit,
d342894c 2553 .ndo_open = vxlan_open,
2554 .ndo_stop = vxlan_stop,
2555 .ndo_start_xmit = vxlan_xmit,
e8171045 2556 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2557 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2558 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2559 .ndo_validate_addr = eth_validate_addr,
2560 .ndo_set_mac_address = eth_mac_addr,
2561 .ndo_fdb_add = vxlan_fdb_add,
2562 .ndo_fdb_del = vxlan_fdb_delete,
2563 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2564 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2565};
2566
e1e5314d
JB		 2567static const struct net_device_ops vxlan_netdev_raw_ops = {
2568 .ndo_init = vxlan_init,
2569 .ndo_uninit = vxlan_uninit,
2570 .ndo_open = vxlan_open,
2571 .ndo_stop = vxlan_stop,
2572 .ndo_start_xmit = vxlan_xmit,
2573 .ndo_get_stats64 = ip_tunnel_get_stats64,
2574 .ndo_change_mtu = vxlan_change_mtu,
2575 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2576};
2577
d342894c 2578/* Info for udev, that this is a virtual tunnel endpoint */
2579static struct device_type vxlan_type = {
2580 .name = "vxlan",
2581};
2582
e5de25dc 2583/* Calls the ndo_udp_tunnel_add of the caller in order to
35e42379 2584 * supply the listening VXLAN udp ports. Callers are expected
e5de25dc 2585 * to implement the ndo_udp_tunnel_add.
53cf5275 2586 */
b7aade15 2587static void vxlan_push_rx_ports(struct net_device *dev)
53cf5275
JG		 2588{
2589 struct vxlan_sock *vs;
2590 struct net *net = dev_net(dev);
2591 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
35e42379 2592 unsigned int i;
53cf5275
JG		 2593
2594 spin_lock(&vn->sock_lock);
2595 for (i = 0; i < PORT_HASH_SIZE; ++i) {
e7b3db5e
AD		 2596 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist)
2597 udp_tunnel_push_rx_port(dev, vs->sock,
b9adcd69
AD		 2598 (vs->flags & VXLAN_F_GPE) ?
2599 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 2600 UDP_TUNNEL_TYPE_VXLAN);
53cf5275
JG		 2601 }
2602 spin_unlock(&vn->sock_lock);
2603}
53cf5275 2604
d342894c 2605/* Initialize the device structure. */
2606static void vxlan_setup(struct net_device *dev)
2607{
2608 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2609 unsigned int h;
d342894c 2610
65226ef8
JB		 2611 eth_hw_addr_random(dev);
2612 ether_setup(dev);
2613
ebf4063e 2614 dev->destructor = free_netdev;
d342894c 2615 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2616
d342894c 2617 dev->features |= NETIF_F_LLTX;
d6727fe3 2618 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2619 dev->features |= NETIF_F_RXCSUM;
05c0db08 2620 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2621
1eaa8178 2622 dev->vlan_features = dev->features;
0afb1666 2623 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2624 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
02875878 2625 netif_keep_dst(dev);
0c867c9b 2626 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 2627
553675fb 2628 INIT_LIST_HEAD(&vxlan->next);
d342894c 2629 spin_lock_init(&vxlan->hash_lock);
2630
2631 init_timer_deferrable(&vxlan->age_timer);
2632 vxlan->age_timer.function = vxlan_cleanup;
2633 vxlan->age_timer.data = (unsigned long) vxlan;
2634
0dfbdf41 2635 vxlan->cfg.dst_port = htons(vxlan_port);
05f47d69 2636
d342894c 2637 vxlan->dev = dev;
2638
58ce31cc
TH		 2639 gro_cells_init(&vxlan->gro_cells, dev);
2640
d342894c 2641 for (h = 0; h < FDB_HASH_SIZE; ++h)
2642 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2643}
2644
0c867c9b
JB		 2645static void vxlan_ether_setup(struct net_device *dev)
2646{
0c867c9b
JB		 2647 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2648 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2649 dev->netdev_ops = &vxlan_netdev_ether_ops;
2650}
2651
e1e5314d
JB		 2652static void vxlan_raw_setup(struct net_device *dev)
2653{
65226ef8 2654 dev->header_ops = NULL;
e1e5314d
JB		 2655 dev->type = ARPHRD_NONE;
2656 dev->hard_header_len = 0;
2657 dev->addr_len = 0;
e1e5314d
JB		 2658 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2659 dev->netdev_ops = &vxlan_netdev_raw_ops;
2660}
2661
d342894c 2662static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2663 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2664 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2665 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2666 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2667 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2668 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2669 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2670 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 2671 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 2672 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2673 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2674 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2675 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2676 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2677 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2678 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2679 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2680 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2681 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2682 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2683 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2684 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2685 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2686 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2687 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 2688 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 2689 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2690};
2691
2692static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2693{
2694 if (tb[IFLA_ADDRESS]) {
2695 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2696 pr_debug("invalid link address (not ethernet)\n");
2697 return -EINVAL;
2698 }
2699
2700 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2701 pr_debug("invalid all zero ethernet address\n");
2702 return -EADDRNOTAVAIL;
2703 }
2704 }
2705
2706 if (!data)
2707 return -EINVAL;
2708
2709 if (data[IFLA_VXLAN_ID]) {
2710 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
4e37d691 2711 if (id >= VXLAN_N_VID)
d342894c 2712 return -ERANGE;
2713 }
2714
05f47d69 2715 if (data[IFLA_VXLAN_PORT_RANGE]) {
2716 const struct ifla_vxlan_port_range *p
2717 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2718
2719 if (ntohs(p->high) < ntohs(p->low)) {
2720 pr_debug("port range %u .. %u not valid\n",
2721 ntohs(p->low), ntohs(p->high));
2722 return -EINVAL;
2723 }
2724 }
2725
d342894c 2726 return 0;
2727}
2728
1b13c97f
YB		 2729static void vxlan_get_drvinfo(struct net_device *netdev,
2730 struct ethtool_drvinfo *drvinfo)
2731{
2732 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2733 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2734}
2735
2736static const struct ethtool_ops vxlan_ethtool_ops = {
2737 .get_drvinfo = vxlan_get_drvinfo,
2738 .get_link = ethtool_op_get_link,
2739};
2740
3ee64f39
TH		 2741static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2742 __be16 port, u32 flags)
553675fb 2743{
e4c7ed41 2744 struct socket *sock;
3ee64f39
TH		 2745 struct udp_port_cfg udp_conf;
2746 int err;
e4c7ed41 2747
3ee64f39 2748 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2749
3ee64f39
TH		 2750 if (ipv6) {
2751 udp_conf.family = AF_INET6;
3ee64f39 2752 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2753 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2754 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2755 } else {
2756 udp_conf.family = AF_INET;
e4c7ed41 2757 }
e4c7ed41 2758
3ee64f39 2759 udp_conf.local_udp_port = port;
553675fb 2760
3ee64f39
TH		 2761 /* Open UDP socket */
2762 err = udp_sock_create(net, &udp_conf, &sock);
2763 if (err < 0)
2764 return ERR_PTR(err);
e4c7ed41 2765
39deb2c7 2766 return sock;
e4c7ed41
CW		 2767}
2768
2769/* Create new listen socket if needed */
b1be00a6
JB		 2770static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2771 __be16 port, u32 flags)
e4c7ed41
CW		 2772{
2773 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2774 struct vxlan_sock *vs;
2775 struct socket *sock;
e4c7ed41 2776 unsigned int h;
acbf74a7 2777 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2778
dc01e7d3 2779 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2780 if (!vs)
2781 return ERR_PTR(-ENOMEM);
2782
2783 for (h = 0; h < VNI_HASH_SIZE; ++h)
2784 INIT_HLIST_HEAD(&vs->vni_list[h]);
2785
3ee64f39 2786 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2787 if (IS_ERR(sock)) {
553675fb 2788 kfree(vs);
e50fddc8 2789 return ERR_CAST(sock);
553675fb 2790 }
e4c7ed41
CW		 2791
2792 vs->sock = sock;
9c2e24e1 2793 atomic_set(&vs->refcnt, 1);
af33c1ad 2794 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2795
9c2e24e1
PS		 2796 spin_lock(&vn->sock_lock);
2797 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
e7b3db5e 2798 udp_tunnel_notify_add_rx_port(sock,
b9adcd69
AD		 2799 (vs->flags & VXLAN_F_GPE) ?
2800 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 2801 UDP_TUNNEL_TYPE_VXLAN);
9c2e24e1 2802 spin_unlock(&vn->sock_lock);
553675fb 2803
2804 /* Mark socket as an encapsulation socket. */
5602c48c 2805 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ		 2806 tunnel_cfg.sk_user_data = vs;
2807 tunnel_cfg.encap_type = 1;
f2d1968e 2808 tunnel_cfg.encap_rcv = vxlan_rcv;
acbf74a7 2809 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH		 2810 tunnel_cfg.gro_receive = vxlan_gro_receive;
2811 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ		 2812
2813 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2814
9c2e24e1
PS		 2815 return vs;
2816}
2817
b1be00a6 2818static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2819{
205f356d
JB		 2820 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2821 struct vxlan_sock *vs = NULL;
9c2e24e1 2822
205f356d 2823 if (!vxlan->cfg.no_share) {
56ef9c90 2824 spin_lock(&vn->sock_lock);
205f356d
JB		 2825 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2826 vxlan->cfg.dst_port, vxlan->flags);
2827 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
56ef9c90 2828 spin_unlock(&vn->sock_lock);
205f356d 2829 return -EBUSY;
56ef9c90
MRL		 2830 }
2831 spin_unlock(&vn->sock_lock);
2832 }
205f356d 2833 if (!vs)
b1be00a6
JB		 2834 vs = vxlan_socket_create(vxlan->net, ipv6,
2835 vxlan->cfg.dst_port, vxlan->flags);
205f356d
JB		 2836 if (IS_ERR(vs))
2837 return PTR_ERR(vs);
b1be00a6
JB		 2838#if IS_ENABLED(CONFIG_IPV6)
2839 if (ipv6)
c6fcc4fc 2840 rcu_assign_pointer(vxlan->vn6_sock, vs);
b1be00a6
JB		 2841 else
2842#endif
c6fcc4fc 2843 rcu_assign_pointer(vxlan->vn4_sock, vs);
205f356d
JB		 2844 vxlan_vs_add_dev(vs, vxlan);
2845 return 0;
553675fb 2846}
2847
b1be00a6
JB		 2848static int vxlan_sock_add(struct vxlan_dev *vxlan)
2849{
b1be00a6 2850 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
d074bf96
JB		 2851 bool ipv6 = vxlan->flags & VXLAN_F_IPV6 || metadata;
2852 bool ipv4 = !ipv6 || metadata;
b1be00a6
JB		 2853 int ret = 0;
2854
c6fcc4fc 2855 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
b1be00a6 2856#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 2857 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
d074bf96 2858 if (ipv6) {
b1be00a6 2859 ret = __vxlan_sock_add(vxlan, true);
d074bf96
JB		 2860 if (ret < 0 && ret != -EAFNOSUPPORT)
2861 ipv4 = false;
2862 }
b1be00a6 2863#endif
d074bf96 2864 if (ipv4)
b1be00a6
JB		 2865 ret = __vxlan_sock_add(vxlan, false);
2866 if (ret < 0)
2867 vxlan_sock_release(vxlan);
2868 return ret;
2869}
2870
0dfbdf41 2871static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
8bcdc4f3
RP		 2872 struct vxlan_config *conf,
2873 bool changelink)
d342894c 2874{
33564bbb 2875 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
07b9b37c 2876 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
c7995c43 2877 struct vxlan_rdst *dst = &vxlan->default_dst;
b1be00a6 2878 unsigned short needed_headroom = ETH_HLEN;
e4c7ed41 2879 bool use_ipv6 = false;
0dfbdf41 2880 __be16 default_port = vxlan->cfg.dst_port;
7e059158 2881 struct net_device *lowerdev = NULL;
d342894c 2882
8bcdc4f3
RP		 2883 if (!changelink) {
2884 if (conf->flags & VXLAN_F_GPE) {
2885 /* For now, allow GPE only together with
2886 * COLLECT_METADATA. This can be relaxed later; in such
2887 * case, the other side of the PtP link will have to be
2888 * provided.
2889 */
2890 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
2891 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
2892 pr_info("unsupported combination of extensions\n");
2893 return -EINVAL;
2894 }
2895 vxlan_raw_setup(dev);
2896 } else {
2897 vxlan_ether_setup(dev);
3555621d 2898 }
e1e5314d 2899
8bcdc4f3
RP		 2900 /* MTU range: 68 - 65535 */
2901 dev->min_mtu = ETH_MIN_MTU;
2902 dev->max_mtu = ETH_MAX_MTU;
2903 vxlan->net = src_net;
e1e5314d 2904 }
0c867c9b 2905
0dfbdf41 2906 dst->remote_vni = conf->vni;
d342894c 2907
0dfbdf41 2908 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
d342894c 2909
0dfbdf41
TG		 2910 /* Unless IPv6 is explicitly requested, assume IPv4 */
2911 if (!dst->remote_ip.sa.sa_family)
2912 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41 2913
0dfbdf41 2914 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
057ba29b
JB		 2915 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2916 if (!IS_ENABLED(CONFIG_IPV6))
2917 return -EPFNOSUPPORT;
e4c7ed41 2918 use_ipv6 = true;
b1be00a6 2919 vxlan->flags |= VXLAN_F_IPV6;
057ba29b 2920 }
d342894c 2921
e7f70af1
DB		 2922 if (conf->label && !use_ipv6) {
2923 pr_info("label only supported in use with IPv6\n");
2924 return -EINVAL;
2925 }
2926
8bcdc4f3
RP		 2927 if (conf->remote_ifindex &&
2928 conf->remote_ifindex != vxlan->cfg.remote_ifindex) {
7e059158 2929 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
0dfbdf41 2930 dst->remote_ifindex = conf->remote_ifindex;
34e02aa1 2931
2932 if (!lowerdev) {
8bcdc4f3
RP		 2933 pr_info("ifindex %d does not exist\n",
2934 dst->remote_ifindex);
34e02aa1 2935 return -ENODEV;
2936 }
d342894c 2937
e4c7ed41
CW		 2938#if IS_ENABLED(CONFIG_IPV6)
2939 if (use_ipv6) {
2940 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2941 if (idev && idev->cnf.disable_ipv6) {
2942 pr_info("IPv6 is disabled via sysctl\n");
2943 return -EPERM;
2944 }
e4c7ed41
CW		 2945 }
2946#endif
2947
0dfbdf41 2948 if (!conf->mtu)
91572088
JW		 2949 dev->mtu = lowerdev->mtu -
2950 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2951
b1be00a6 2952 needed_headroom = lowerdev->hard_header_len;
8bcdc4f3
RP		 2953 } else if (!conf->remote_ifindex &&
2954 vxlan_addr_multicast(&dst->remote_ip)) {
9b4cdd51
JB		 2955 pr_info("multicast destination requires interface to be specified\n");
2956 return -EINVAL;
9dc2ad10 2957 }
d342894c 2958
d6acfeb1
FM		 2959 if (lowerdev) {
2960 dev->gso_max_size = lowerdev->gso_max_size;
2961 dev->gso_max_segs = lowerdev->gso_max_segs;
2962 }
2963
7e059158 2964 if (conf->mtu) {
91572088
JW		 2965 int max_mtu = ETH_MAX_MTU;
2966
2967 if (lowerdev)
2968 max_mtu = lowerdev->mtu;
2969
2970 max_mtu -= (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2971
2972 if (conf->mtu < dev->min_mtu || conf->mtu > dev->max_mtu)
2973 return -EINVAL;
2974
2975 dev->mtu = conf->mtu;
2976
2977 if (conf->mtu > max_mtu)
2978 dev->mtu = max_mtu;
7e059158
DW		 2979 }
2980
b1be00a6
JB		 2981 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2982 needed_headroom += VXLAN6_HEADROOM;
2983 else
2984 needed_headroom += VXLAN_HEADROOM;
2985 dev->needed_headroom = needed_headroom;
2986
0dfbdf41 2987 memcpy(&vxlan->cfg, conf, sizeof(*conf));
e1e5314d
JB		 2988 if (!vxlan->cfg.dst_port) {
2989 if (conf->flags & VXLAN_F_GPE)
d5ff72d9 2990 vxlan->cfg.dst_port = htons(4790); /* IANA VXLAN-GPE port */
e1e5314d
JB		 2991 else
2992 vxlan->cfg.dst_port = default_port;
2993 }
0dfbdf41
TG		 2994 vxlan->flags |= conf->flags;
2995
2996 if (!vxlan->cfg.age_interval)
2997 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2998
8bcdc4f3
RP		 2999 if (changelink)
3000 return 0;
3001
07b9b37c
ND		 3002 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3003 if (tmp->cfg.vni == conf->vni &&
3004 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
3005 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
3006 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
3007 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
3555621d
JB		 3008 (vxlan->flags & VXLAN_F_RCV_FLAGS)) {
3009 pr_info("duplicate VNI %u\n", be32_to_cpu(conf->vni));
3010 return -EEXIST;
3011 }
07b9b37c 3012 }
0dfbdf41 3013
0dfbdf41
TG		 3014 return 0;
3015}
3016
c80498e3
ND		 3017static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3018 struct vxlan_config *conf)
3019{
3020 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3021 struct vxlan_dev *vxlan = netdev_priv(dev);
3022 int err;
3023
3024 err = vxlan_dev_configure(net, dev, conf, false);
3025 if (err)
3026 return err;
3027
3028 dev->ethtool_ops = &vxlan_ethtool_ops;
3029
3030 /* create an fdb entry for a valid default destination */
3031 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
3032 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3033 &vxlan->default_dst.remote_ip,
3034 NUD_REACHABLE | NUD_PERMANENT,
3035 NLM_F_EXCL | NLM_F_CREATE,
3036 vxlan->cfg.dst_port,
3037 vxlan->default_dst.remote_vni,
3038 vxlan->default_dst.remote_vni,
3039 vxlan->default_dst.remote_ifindex,
3040 NTF_SELF);
3041 if (err)
3042 return err;
3043 }
3044
3045 err = register_netdevice(dev);
3046 if (err) {
3047 vxlan_fdb_delete_default(vxlan, vxlan->default_dst.remote_vni);
3048 return err;
3049 }
3050
3051 list_add(&vxlan->next, &vn->vxlan_list);
3052 return 0;
3053}
3054
8bcdc4f3
RP		 3055static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3056 struct net_device *dev, struct vxlan_config *conf,
3057 bool changelink)
0dfbdf41 3058{
8bcdc4f3 3059 struct vxlan_dev *vxlan = netdev_priv(dev);
0dfbdf41 3060
8bcdc4f3 3061 memset(conf, 0, sizeof(*conf));
e277de5f 3062
8bcdc4f3
RP		 3063 /* if changelink operation, start with old existing cfg */
3064 if (changelink)
3065 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3066
3067 if (data[IFLA_VXLAN_ID]) {
3068 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3069
3070 if (changelink && (vni != conf->vni))
3071 return -EOPNOTSUPP;
3072 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3073 }
0dfbdf41
TG		 3074
3075 if (data[IFLA_VXLAN_GROUP]) {
8bcdc4f3 3076 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
0dfbdf41
TG		 3077 } else if (data[IFLA_VXLAN_GROUP6]) {
3078 if (!IS_ENABLED(CONFIG_IPV6))
3079 return -EPFNOSUPPORT;
3080
8bcdc4f3
RP		 3081 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3082 conf->remote_ip.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3083 }
3084
3085 if (data[IFLA_VXLAN_LOCAL]) {
8bcdc4f3
RP		 3086 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3087 conf->saddr.sa.sa_family = AF_INET;
0dfbdf41
TG		 3088 } else if (data[IFLA_VXLAN_LOCAL6]) {
3089 if (!IS_ENABLED(CONFIG_IPV6))
3090 return -EPFNOSUPPORT;
3091
3092 /* TODO: respect scope id */
8bcdc4f3
RP		 3093 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3094 conf->saddr.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3095 }
3096
3097 if (data[IFLA_VXLAN_LINK])
8bcdc4f3 3098 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
0dfbdf41 3099
d342894c 3100 if (data[IFLA_VXLAN_TOS])
8bcdc4f3 3101 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 3102
afb97186 3103 if (data[IFLA_VXLAN_TTL])
8bcdc4f3 3104 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 3105
e7f70af1 3106 if (data[IFLA_VXLAN_LABEL])
8bcdc4f3 3107 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
e7f70af1
DB		 3108 IPV6_FLOWLABEL_MASK;
3109
8bcdc4f3
RP		 3110 if (data[IFLA_VXLAN_LEARNING]) {
3111 if (nla_get_u8(data[IFLA_VXLAN_LEARNING])) {
3112 conf->flags |= VXLAN_F_LEARN;
3113 } else {
3114 conf->flags &= ~VXLAN_F_LEARN;
3115 vxlan->flags &= ~VXLAN_F_LEARN;
3116 }
3117 } else if (!changelink) {
3118 /* default to learn on a new device */
3119 conf->flags |= VXLAN_F_LEARN;
3120 }
d342894c 3121
8bcdc4f3
RP		 3122 if (data[IFLA_VXLAN_AGEING]) {
3123 if (changelink)
3124 return -EOPNOTSUPP;
3125 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3126 }
d342894c 3127
8bcdc4f3
RP		 3128 if (data[IFLA_VXLAN_PROXY]) {
3129 if (changelink)
3130 return -EOPNOTSUPP;
3131 if (nla_get_u8(data[IFLA_VXLAN_PROXY]))
3132 conf->flags |= VXLAN_F_PROXY;
3133 }
e4f67add 3134
8bcdc4f3
RP		 3135 if (data[IFLA_VXLAN_RSC]) {
3136 if (changelink)
3137 return -EOPNOTSUPP;
3138 if (nla_get_u8(data[IFLA_VXLAN_RSC]))
3139 conf->flags |= VXLAN_F_RSC;
3140 }
e4f67add 3141
8bcdc4f3
RP		 3142 if (data[IFLA_VXLAN_L2MISS]) {
3143 if (changelink)
3144 return -EOPNOTSUPP;
3145 if (nla_get_u8(data[IFLA_VXLAN_L2MISS]))
3146 conf->flags |= VXLAN_F_L2MISS;
3147 }
e4f67add 3148
8bcdc4f3
RP		 3149 if (data[IFLA_VXLAN_L3MISS]) {
3150 if (changelink)
3151 return -EOPNOTSUPP;
3152 if (nla_get_u8(data[IFLA_VXLAN_L3MISS]))
3153 conf->flags |= VXLAN_F_L3MISS;
3154 }
e4f67add 3155
8bcdc4f3
RP		 3156 if (data[IFLA_VXLAN_LIMIT]) {
3157 if (changelink)
3158 return -EOPNOTSUPP;
3159 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3160 }
d342894c 3161
8bcdc4f3
RP		 3162 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3163 if (changelink)
3164 return -EOPNOTSUPP;
3165 if (nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3166 conf->flags |= VXLAN_F_COLLECT_METADATA;
3167 }
f8a9b1bc 3168
05f47d69 3169 if (data[IFLA_VXLAN_PORT_RANGE]) {
8bcdc4f3
RP		 3170 if (!changelink) {
3171 const struct ifla_vxlan_port_range *p
3172 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3173 conf->port_min = ntohs(p->low);
3174 conf->port_max = ntohs(p->high);
3175 } else {
3176 return -EOPNOTSUPP;
3177 }
05f47d69 3178 }
3179
8bcdc4f3
RP		 3180 if (data[IFLA_VXLAN_PORT]) {
3181 if (changelink)
3182 return -EOPNOTSUPP;
3183 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3184 }
823aa873 3185
8bcdc4f3
RP		 3186 if (data[IFLA_VXLAN_UDP_CSUM]) {
3187 if (changelink)
3188 return -EOPNOTSUPP;
3189 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3190 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3191 }
359a0ea9 3192
8bcdc4f3
RP		 3193 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3194 if (changelink)
3195 return -EOPNOTSUPP;
3196 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
3197 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
3198 }
359a0ea9 3199
8bcdc4f3
RP		 3200 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3201 if (changelink)
3202 return -EOPNOTSUPP;
3203 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
3204 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
3205 }
359a0ea9 3206
8bcdc4f3
RP		 3207 if (data[IFLA_VXLAN_REMCSUM_TX]) {
3208 if (changelink)
3209 return -EOPNOTSUPP;
3210 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
3211 conf->flags |= VXLAN_F_REMCSUM_TX;
3212 }
dfd8645e 3213
8bcdc4f3
RP		 3214 if (data[IFLA_VXLAN_REMCSUM_RX]) {
3215 if (changelink)
3216 return -EOPNOTSUPP;
3217 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3218 conf->flags |= VXLAN_F_REMCSUM_RX;
3219 }
3220
3221 if (data[IFLA_VXLAN_GBP]) {
3222 if (changelink)
3223 return -EOPNOTSUPP;
3224 conf->flags |= VXLAN_F_GBP;
3225 }
3226
3227 if (data[IFLA_VXLAN_GPE]) {
3228 if (changelink)
3229 return -EOPNOTSUPP;
3230 conf->flags |= VXLAN_F_GPE;
3231 }
3232
3233 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3234 if (changelink)
3235 return -EOPNOTSUPP;
3236 conf->flags |= VXLAN_F_REMCSUM_NOPARTIAL;
3237 }
3238
3239 if (tb[IFLA_MTU]) {
3240 if (changelink)
3241 return -EOPNOTSUPP;
3242 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3243 }
3244
3245 return 0;
3246}
3247
3248static int vxlan_newlink(struct net *src_net, struct net_device *dev,
3249 struct nlattr *tb[], struct nlattr *data[])
3250{
8bcdc4f3
RP		 3251 struct vxlan_config conf;
3252 int err;
3253
3254 err = vxlan_nl2conf(tb, data, dev, &conf, false);
3255 if (err)
3256 return err;
3257
c80498e3 3258 return __vxlan_dev_create(src_net, dev, &conf);
8bcdc4f3 3259}
dfd8645e 3260
8bcdc4f3
RP		 3261static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
3262 struct nlattr *data[])
3263{
3264 struct vxlan_dev *vxlan = netdev_priv(dev);
3265 struct vxlan_rdst *dst = &vxlan->default_dst;
3266 struct vxlan_rdst old_dst;
3267 struct vxlan_config conf;
3268 int err;
3269
3270 err = vxlan_nl2conf(tb, data,
3271 dev, &conf, true);
3272 if (err)
3273 return err;
3511494c 3274
8bcdc4f3 3275 memcpy(&old_dst, dst, sizeof(struct vxlan_rdst));
e1e5314d 3276
8bcdc4f3
RP		 3277 err = vxlan_dev_configure(vxlan->net, dev, &conf, true);
3278 if (err)
3279 return err;
0ace2ca8 3280
8bcdc4f3
RP		 3281 /* handle default dst entry */
3282 if (!vxlan_addr_equal(&dst->remote_ip, &old_dst.remote_ip)) {
3283 spin_lock_bh(&vxlan->hash_lock);
3284 if (!vxlan_addr_any(&old_dst.remote_ip))
3285 __vxlan_fdb_delete(vxlan, all_zeros_mac,
3286 old_dst.remote_ip,
3287 vxlan->cfg.dst_port,
3288 old_dst.remote_vni,
3289 old_dst.remote_vni,
3290 old_dst.remote_ifindex, 0);
3291
3292 if (!vxlan_addr_any(&dst->remote_ip)) {
3293 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3294 &dst->remote_ip,
3295 NUD_REACHABLE | NUD_PERMANENT,
3296 NLM_F_CREATE | NLM_F_APPEND,
3297 vxlan->cfg.dst_port,
3298 dst->remote_vni,
3299 dst->remote_vni,
3300 dst->remote_ifindex,
3301 NTF_SELF);
3302 if (err) {
3303 spin_unlock_bh(&vxlan->hash_lock);
3304 return err;
3305 }
3306 }
3307 spin_unlock_bh(&vxlan->hash_lock);
3308 }
ce577668 3309
8bcdc4f3 3310 return 0;
d342894c 3311}
3312
3313static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3314{
3315 struct vxlan_dev *vxlan = netdev_priv(dev);
3316
8b3f9337
RP		 3317 vxlan_flush(vxlan, true);
3318
58ce31cc 3319 gro_cells_destroy(&vxlan->gro_cells);
553675fb 3320 list_del(&vxlan->next);
d342894c 3321 unregister_netdevice_queue(dev, head);
3322}
3323
3324static size_t vxlan_get_size(const struct net_device *dev)
3325{
3326
3327 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3328 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3329 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3330 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3331 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3332 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
e7f70af1 3333 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 3334 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 3335 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3336 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3337 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3338 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3339 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3340 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3341 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3342 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3343 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3344 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3345 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3346 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3347 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3348 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3349 0;
3350}
3351
3352static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3353{
3354 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3355 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3356 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3357 .low = htons(vxlan->cfg.port_min),
3358 .high = htons(vxlan->cfg.port_max),
05f47d69 3359 };
d342894c 3360
54bfd872 3361 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3362 goto nla_put_failure;
3363
e4c7ed41
CW		 3364 if (!vxlan_addr_any(&dst->remote_ip)) {
3365 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3366 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3367 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3368 goto nla_put_failure;
3369#if IS_ENABLED(CONFIG_IPV6)
3370 } else {
930345ea
JB		 3371 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3372 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3373 goto nla_put_failure;
3374#endif
3375 }
3376 }
d342894c 3377
c7995c43 3378 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3379 goto nla_put_failure;
3380
0dfbdf41
TG		 3381 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3382 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3383 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3384 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3385 goto nla_put_failure;
3386#if IS_ENABLED(CONFIG_IPV6)
3387 } else {
930345ea 3388 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3389 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3390 goto nla_put_failure;
3391#endif
3392 }
3393 }
d342894c 3394
0dfbdf41
TG		 3395 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3396 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e7f70af1 3397 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add
DS		 3398 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3399 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3400 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3401 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3402 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3403 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3404 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3405 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3406 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
da8b43c0
AS		 3407 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3408 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3409 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3410 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3411 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3412 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
6ceb31ca 3413 !(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9
TH		 3414 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3415 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3416 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 3417 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3418 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3419 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3420 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3421 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 3422 goto nla_put_failure;
3423
05f47d69 3424 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3425 goto nla_put_failure;
3426
3511494c
TG		 3427 if (vxlan->flags & VXLAN_F_GBP &&
3428 nla_put_flag(skb, IFLA_VXLAN_GBP))
3429 goto nla_put_failure;
3430
e1e5314d
JB		 3431 if (vxlan->flags & VXLAN_F_GPE &&
3432 nla_put_flag(skb, IFLA_VXLAN_GPE))
3433 goto nla_put_failure;
3434
0ace2ca8
TH		 3435 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3436 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3437 goto nla_put_failure;
3438
d342894c 3439 return 0;
3440
3441nla_put_failure:
3442 return -EMSGSIZE;
3443}
3444
1728d4fa
ND		 3445static struct net *vxlan_get_link_net(const struct net_device *dev)
3446{
3447 struct vxlan_dev *vxlan = netdev_priv(dev);
3448
3449 return vxlan->net;
3450}
3451
d342894c 3452static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3453 .kind = "vxlan",
3454 .maxtype = IFLA_VXLAN_MAX,
3455 .policy = vxlan_policy,
3456 .priv_size = sizeof(struct vxlan_dev),
3457 .setup = vxlan_setup,
3458 .validate = vxlan_validate,
3459 .newlink = vxlan_newlink,
8bcdc4f3 3460 .changelink = vxlan_changelink,
d342894c 3461 .dellink = vxlan_dellink,
3462 .get_size = vxlan_get_size,
3463 .fill_info = vxlan_fill_info,
1728d4fa 3464 .get_link_net = vxlan_get_link_net,
d342894c 3465};
3466
cf5da330
ND		 3467struct net_device *vxlan_dev_create(struct net *net, const char *name,
3468 u8 name_assign_type,
3469 struct vxlan_config *conf)
3470{
3471 struct nlattr *tb[IFLA_MAX + 1];
3472 struct net_device *dev;
3473 int err;
3474
3475 memset(&tb, 0, sizeof(tb));
3476
3477 dev = rtnl_create_link(net, name, name_assign_type,
3478 &vxlan_link_ops, tb);
3479 if (IS_ERR(dev))
3480 return dev;
3481
c80498e3 3482 err = __vxlan_dev_create(net, dev, conf);
cf5da330
ND		 3483 if (err < 0) {
3484 free_netdev(dev);
3485 return ERR_PTR(err);
3486 }
3487
3488 err = rtnl_configure_link(dev, NULL);
3489 if (err < 0) {
3490 LIST_HEAD(list_kill);
3491
3492 vxlan_dellink(dev, &list_kill);
3493 unregister_netdevice_many(&list_kill);
3494 return ERR_PTR(err);
3495 }
3496
3497 return dev;
3498}
3499EXPORT_SYMBOL_GPL(vxlan_dev_create);
3500
acaf4e70
DB		 3501static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3502 struct net_device *dev)
3503{
3504 struct vxlan_dev *vxlan, *next;
3505 LIST_HEAD(list_kill);
3506
3507 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3508 struct vxlan_rdst *dst = &vxlan->default_dst;
3509
3510 /* In case we created vxlan device with carrier
3511 * and we loose the carrier due to module unload
3512 * we also need to remove vxlan device. In other
3513 * cases, it's not necessary and remote_ifindex
3514 * is 0 here, so no matches.
3515 */
3516 if (dst->remote_ifindex == dev->ifindex)
3517 vxlan_dellink(vxlan->dev, &list_kill);
3518 }
3519
3520 unregister_netdevice_many(&list_kill);
3521}
3522
b7aade15
HFS		 3523static int vxlan_netdevice_event(struct notifier_block *unused,
3524 unsigned long event, void *ptr)
acaf4e70
DB		 3525{
3526 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3527 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3528
783c1463 3529 if (event == NETDEV_UNREGISTER)
acaf4e70 3530 vxlan_handle_lowerdev_unregister(vn, dev);
7c46a640 3531 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
b7aade15 3532 vxlan_push_rx_ports(dev);
acaf4e70
DB		 3533
3534 return NOTIFY_DONE;
3535}
3536
3537static struct notifier_block vxlan_notifier_block __read_mostly = {
b7aade15 3538 .notifier_call = vxlan_netdevice_event,
acaf4e70
DB		 3539};
3540
d342894c 3541static __net_init int vxlan_init_net(struct net *net)
3542{
3543 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3544 unsigned int h;
d342894c 3545
553675fb 3546 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3547 spin_lock_init(&vn->sock_lock);
d342894c 3548
553675fb 3549 for (h = 0; h < PORT_HASH_SIZE; ++h)
3550 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3551
3552 return 0;
3553}
3554
f01ec1c0
ND		 3555static void __net_exit vxlan_exit_net(struct net *net)
3556{
3557 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3558 struct vxlan_dev *vxlan, *next;
3559 struct net_device *dev, *aux;
3560 LIST_HEAD(list);
3561
3562 rtnl_lock();
3563 for_each_netdev_safe(net, dev, aux)
3564 if (dev->rtnl_link_ops == &vxlan_link_ops)
3565 unregister_netdevice_queue(dev, &list);
3566
3567 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3568 /* If vxlan->dev is in the same netns, it has already been added
3569 * to the list by the previous loop.
3570 */
58ce31cc
TH		 3571 if (!net_eq(dev_net(vxlan->dev), net)) {
3572 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3573 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3574 }
f01ec1c0
ND		 3575 }
3576
3577 unregister_netdevice_many(&list);
3578 rtnl_unlock();
3579}
3580
d342894c 3581static struct pernet_operations vxlan_net_ops = {
3582 .init = vxlan_init_net,
f01ec1c0 3583 .exit = vxlan_exit_net,
d342894c 3584 .id = &vxlan_net_id,
3585 .size = sizeof(struct vxlan_net),
3586};
3587
3588static int __init vxlan_init_module(void)
3589{
3590 int rc;
3591
3592 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3593
783c1463 3594 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3595 if (rc)
3596 goto out1;
3597
acaf4e70 3598 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3599 if (rc)
3600 goto out2;
3601
acaf4e70
DB		 3602 rc = rtnl_link_register(&vxlan_link_ops);
3603 if (rc)
3604 goto out3;
d342894c 3605
acaf4e70
DB		 3606 return 0;
3607out3:
3608 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3609out2:
783c1463 3610 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3611out1:
3612 return rc;
3613}
7332a13b 3614late_initcall(vxlan_init_module);
d342894c 3615
3616static void __exit vxlan_cleanup_module(void)
3617{
b7153984 3618 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3619 unregister_netdevice_notifier(&vxlan_notifier_block);
783c1463
DB		 3620 unregister_pernet_subsys(&vxlan_net_ops);
3621 /* rcu_barrier() is called by netns */
d342894c 3622}
3623module_exit(vxlan_cleanup_module);
3624
3625MODULE_LICENSE("GPL");
3626MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3627MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3628MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3629MODULE_ALIAS_RTNL_LINK("vxlan");
Linux 6.x block layer and io_uring tree(s)