]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
drivers/md/dm-ioctl.c: use kvmalloc rather than opencoded variant
[mirror_ubuntu-artful-kernel.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
d342894c 14#include <linux/module.h>
15#include <linux/errno.h>
16#include <linux/slab.h>
d342894c 17#include <linux/udp.h>
18#include <linux/igmp.h>
d342894c 19#include <linux/if_ether.h>
1b13c97f 20#include <linux/ethtool.h>
e4f67add
DS		 21#include <net/arp.h>
22#include <net/ndisc.h>
d342894c 23#include <net/ip.h>
24#include <net/icmp.h>
d342894c 25#include <net/rtnetlink.h>
d342894c 26#include <net/inet_ecn.h>
27#include <net/net_namespace.h>
28#include <net/netns/generic.h>
012a5729 29#include <net/vxlan.h>
40d29af0 30
e4c7ed41 31#if IS_ENABLED(CONFIG_IPV6)
e4c7ed41 32#include <net/ip6_tunnel.h>
660d98ca 33#include <net/ip6_checksum.h>
e4c7ed41 34#endif
d342894c 35
36#define VXLAN_VERSION "0.1"
37
553675fb 38#define PORT_HASH_BITS 8
39#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 40#define FDB_AGE_DEFAULT 300 /* 5 min */
41#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
42
23c578bf 43/* UDP port for VXLAN traffic.
44 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 45 * for compatibility with early adopters.
23c578bf 46 */
9daaa397
SH		 47static unsigned short vxlan_port __read_mostly = 8472;
48module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 49MODULE_PARM_DESC(udp_port, "Destination UDP port");
50
51static bool log_ecn_error = true;
52module_param(log_ecn_error, bool, 0644);
53MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
54
c7d03a00 55static unsigned int vxlan_net_id;
0dfbdf41 56static struct rtnl_link_ops vxlan_link_ops;
553675fb 57
7256eac1 58static const u8 all_zeros_mac[ETH_ALEN + 2];
afbd8bae 59
205f356d 60static int vxlan_sock_add(struct vxlan_dev *vxlan);
614732ea 61
553675fb 62/* per-network namespace private data for this module */
63struct vxlan_net {
64 struct list_head vxlan_list;
65 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 66 spinlock_t sock_lock;
553675fb 67};
68
d342894c 69/* Forwarding table entry */
70struct vxlan_fdb {
71 struct hlist_node hlist; /* linked list of entries */
72 struct rcu_head rcu;
73 unsigned long updated; /* jiffies */
74 unsigned long used;
3e61aa8f 75 struct list_head remotes;
7177a3b0 76 u8 eth_addr[ETH_ALEN];
d342894c 77 u16 state; /* see ndm_state */
3ad7a4b1 78 __be32 vni;
ae884082 79 u8 flags; /* see ndm_flags */
d342894c 80};
81
d342894c 82/* salt for hash table */
83static u32 vxlan_salt __read_mostly;
84
ee122c79
TG		 85static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
86{
e7030878
TG		 87 return vs->flags & VXLAN_F_COLLECT_METADATA ||
88 ip_tunnel_collect_metadata();
ee122c79
TG		 89}
90
e4c7ed41
CW		 91#if IS_ENABLED(CONFIG_IPV6)
92static inline
93bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
94{
f0ef3126
JB		 95 if (a->sa.sa_family != b->sa.sa_family)
96 return false;
97 if (a->sa.sa_family == AF_INET6)
98 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
99 else
100 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 101}
102
103static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
104{
f0ef3126
JB		 105 if (ipa->sa.sa_family == AF_INET6)
106 return ipv6_addr_any(&ipa->sin6.sin6_addr);
107 else
108 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 109}
110
111static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
112{
f0ef3126
JB		 113 if (ipa->sa.sa_family == AF_INET6)
114 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
115 else
116 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 117}
118
119static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
120{
f0ef3126 121 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 122 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 123 ip->sa.sa_family = AF_INET6;
124 return 0;
125 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 126 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 127 ip->sa.sa_family = AF_INET;
128 return 0;
129 } else {
130 return -EAFNOSUPPORT;
131 }
e4c7ed41
CW		 132}
133
134static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 135 const union vxlan_addr *ip)
e4c7ed41 136{
f0ef3126 137 if (ip->sa.sa_family == AF_INET6)
930345ea 138 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 139 else
930345ea 140 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 141}
142
143#else /* !CONFIG_IPV6 */
144
145static inline
146bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
147{
f0ef3126 148 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 149}
150
151static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
152{
f0ef3126 153 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 154}
155
156static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
157{
f0ef3126 158 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 159}
160
161static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
162{
f0ef3126
JB		 163 if (nla_len(nla) >= sizeof(struct in6_addr)) {
164 return -EAFNOSUPPORT;
165 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 166 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 167 ip->sa.sa_family = AF_INET;
168 return 0;
169 } else {
170 return -EAFNOSUPPORT;
171 }
e4c7ed41
CW		 172}
173
174static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 175 const union vxlan_addr *ip)
e4c7ed41 176{
930345ea 177 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 178}
179#endif
180
553675fb 181/* Virtual Network hash table head */
54bfd872 182static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
553675fb 183{
54bfd872 184 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
553675fb 185}
186
187/* Socket hash table head */
188static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 189{
190 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
191
553675fb 192 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
193}
194
3e61aa8f
SH		 195/* First remote destination for a forwarding entry.
196 * Guaranteed to be non-NULL because remotes are never deleted.
197 */
5ca5461c 198static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 199{
5ca5461c 200 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
201}
202
203static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
204{
205 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 206}
207
ac5132d1
TG		 208/* Find VXLAN socket based on network namespace, address family and UDP port
209 * and enabled unshareable flags.
210 */
211static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
212 __be16 port, u32 flags)
553675fb 213{
214 struct vxlan_sock *vs;
af33c1ad
TH		 215
216 flags &= VXLAN_F_RCV_FLAGS;
553675fb 217
218 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 219 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 220 vxlan_get_sk_family(vs) == family &&
af33c1ad 221 vs->flags == flags)
553675fb 222 return vs;
223 }
224 return NULL;
d342894c 225}
226
54bfd872 227static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni)
d342894c 228{
229 struct vxlan_dev *vxlan;
d342894c 230
b9167b2e
JB		 231 /* For flow based devices, map all packets to VNI 0 */
232 if (vs->flags & VXLAN_F_COLLECT_METADATA)
233 vni = 0;
234
54bfd872
JB		 235 hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) {
236 if (vxlan->default_dst.remote_vni == vni)
d342894c 237 return vxlan;
238 }
239
240 return NULL;
241}
242
5cfccc5a 243/* Look up VNI in a per net namespace table */
54bfd872 244static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni,
ac5132d1
TG		 245 sa_family_t family, __be16 port,
246 u32 flags)
5cfccc5a
PS		 247{
248 struct vxlan_sock *vs;
249
ac5132d1 250 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 251 if (!vs)
252 return NULL;
253
54bfd872 254 return vxlan_vs_find_vni(vs, vni);
5cfccc5a
PS		 255}
256
d342894c 257/* Fill in neighbour message in skbuff. */
258static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 259 const struct vxlan_fdb *fdb,
260 u32 portid, u32 seq, int type, unsigned int flags,
261 const struct vxlan_rdst *rdst)
d342894c 262{
263 unsigned long now = jiffies;
264 struct nda_cacheinfo ci;
265 struct nlmsghdr *nlh;
266 struct ndmsg *ndm;
e4f67add 267 bool send_ip, send_eth;
d342894c 268
269 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
270 if (nlh == NULL)
271 return -EMSGSIZE;
272
273 ndm = nlmsg_data(nlh);
274 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 275
276 send_eth = send_ip = true;
277
278 if (type == RTM_GETNEIGH) {
e4c7ed41 279 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add 280 send_eth = !is_zero_ether_addr(fdb->eth_addr);
8f48ba71 281 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
e4f67add
DS		 282 } else
283 ndm->ndm_family = AF_BRIDGE;
d342894c 284 ndm->ndm_state = fdb->state;
285 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 286 ndm->ndm_flags = fdb->flags;
545469f7 287 ndm->ndm_type = RTN_UNICAST;
d342894c 288
193523bf 289 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 290 nla_put_s32(skb, NDA_LINK_NETNSID,
38f507f1 291 peernet2id(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 292 goto nla_put_failure;
293
e4f67add 294 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 295 goto nla_put_failure;
296
e4c7ed41 297 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 298 goto nla_put_failure;
299
0dfbdf41 300 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 301 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
302 goto nla_put_failure;
c7995c43 303 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
54bfd872 304 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
6681712d 305 goto nla_put_failure;
3ad7a4b1
RP		 306 if ((vxlan->flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
307 nla_put_u32(skb, NDA_SRC_VNI,
308 be32_to_cpu(fdb->vni)))
309 goto nla_put_failure;
6681712d
DS		 310 if (rdst->remote_ifindex &&
311 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 312 goto nla_put_failure;
313
314 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
315 ci.ndm_confirmed = 0;
316 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
317 ci.ndm_refcnt = 0;
318
319 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
320 goto nla_put_failure;
321
053c095a
JB		 322 nlmsg_end(skb, nlh);
323 return 0;
d342894c 324
325nla_put_failure:
326 nlmsg_cancel(skb, nlh);
327 return -EMSGSIZE;
328}
329
330static inline size_t vxlan_nlmsg_size(void)
331{
332 return NLMSG_ALIGN(sizeof(struct ndmsg))
333 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 334 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 335 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 336 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
337 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 338 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 339 + nla_total_size(sizeof(struct nda_cacheinfo));
340}
341
9e4b93f9
ND		 342static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
343 struct vxlan_rdst *rd, int type)
d342894c 344{
345 struct net *net = dev_net(vxlan->dev);
346 struct sk_buff *skb;
347 int err = -ENOBUFS;
348
349 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
350 if (skb == NULL)
351 goto errout;
352
9e4b93f9 353 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 354 if (err < 0) {
355 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
356 WARN_ON(err == -EMSGSIZE);
357 kfree_skb(skb);
358 goto errout;
359 }
360
361 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
362 return;
363errout:
364 if (err < 0)
365 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
366}
367
e4c7ed41 368static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 369{
370 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 371 struct vxlan_fdb f = {
372 .state = NUD_STALE,
373 };
374 struct vxlan_rdst remote = {
e4c7ed41 375 .remote_ip = *ipa, /* goes to NDA_DST */
54bfd872 376 .remote_vni = cpu_to_be32(VXLAN_N_VID),
bb3fd687 377 };
3e61aa8f 378
9e4b93f9 379 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 380}
381
382static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
383{
bb3fd687
SH		 384 struct vxlan_fdb f = {
385 .state = NUD_STALE,
386 };
9e4b93f9 387 struct vxlan_rdst remote = { };
e4f67add 388
e4f67add
DS		 389 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
390
9e4b93f9 391 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 392}
393
d342894c 394/* Hash Ethernet address */
395static u32 eth_hash(const unsigned char *addr)
396{
397 u64 value = get_unaligned((u64 *)addr);
398
399 /* only want 6 bytes */
400#ifdef __BIG_ENDIAN
d342894c 401 value >>= 16;
321fb991 402#else
403 value <<= 16;
d342894c 404#endif
405 return hash_64(value, FDB_HASH_BITS);
406}
407
3ad7a4b1
RP		 408static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
409{
410 /* use 1 byte of OUI and 3 bytes of NIC */
411 u32 key = get_unaligned((u32 *)(addr + 2));
412
413 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
414}
415
d342894c 416/* Hash chain to use given mac address */
417static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
3ad7a4b1 418 const u8 *mac, __be32 vni)
d342894c 419{
3ad7a4b1
RP		 420 if (vxlan->flags & VXLAN_F_COLLECT_METADATA)
421 return &vxlan->fdb_head[eth_vni_hash(mac, vni)];
422 else
423 return &vxlan->fdb_head[eth_hash(mac)];
d342894c 424}
425
426/* Look up Ethernet address in forwarding table */
014be2c8 427static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 428 const u8 *mac, __be32 vni)
d342894c 429{
3ad7a4b1 430 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
d342894c 431 struct vxlan_fdb *f;
d342894c 432
b67bfe0d 433 hlist_for_each_entry_rcu(f, head, hlist) {
3ad7a4b1
RP		 434 if (ether_addr_equal(mac, f->eth_addr)) {
435 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
436 if (vni == f->vni)
437 return f;
438 } else {
439 return f;
440 }
441 }
d342894c 442 }
443
444 return NULL;
445}
446
014be2c8 447static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
3ad7a4b1 448 const u8 *mac, __be32 vni)
014be2c8
SS		 449{
450 struct vxlan_fdb *f;
451
3ad7a4b1 452 f = __vxlan_find_mac(vxlan, mac, vni);
014be2c8
SS		 453 if (f)
454 f->used = jiffies;
455
456 return f;
457}
458
a5e7c10a
MR		 459/* caller should hold vxlan->hash_lock */
460static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 461 union vxlan_addr *ip, __be16 port,
54bfd872 462 __be32 vni, __u32 ifindex)
6681712d 463{
3e61aa8f 464 struct vxlan_rdst *rd;
6681712d 465
3e61aa8f 466 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 467 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 468 rd->remote_port == port &&
469 rd->remote_vni == vni &&
470 rd->remote_ifindex == ifindex)
a5e7c10a 471 return rd;
6681712d 472 }
3e61aa8f 473
a5e7c10a
MR		 474 return NULL;
475}
476
906dc186
TR		 477/* Replace destination of unicast mac */
478static int vxlan_fdb_replace(struct vxlan_fdb *f,
54bfd872
JB		 479 union vxlan_addr *ip, __be16 port, __be32 vni,
480 __u32 ifindex)
906dc186
TR		 481{
482 struct vxlan_rdst *rd;
483
484 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
485 if (rd)
486 return 0;
487
488 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
489 if (!rd)
490 return 0;
0c1d70af
PA		 491
492 dst_cache_reset(&rd->dst_cache);
e4c7ed41 493 rd->remote_ip = *ip;
906dc186
TR		 494 rd->remote_port = port;
495 rd->remote_vni = vni;
496 rd->remote_ifindex = ifindex;
497 return 1;
498}
499
a5e7c10a
MR		 500/* Add/update destinations for multicast */
501static int vxlan_fdb_append(struct vxlan_fdb *f,
54bfd872 502 union vxlan_addr *ip, __be16 port, __be32 vni,
9e4b93f9 503 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 504{
505 struct vxlan_rdst *rd;
506
507 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
508 if (rd)
509 return 0;
510
6681712d
DS		 511 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
512 if (rd == NULL)
513 return -ENOBUFS;
0c1d70af
PA		 514
515 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
516 kfree(rd);
517 return -ENOBUFS;
518 }
519
e4c7ed41 520 rd->remote_ip = *ip;
6681712d
DS		 521 rd->remote_port = port;
522 rd->remote_vni = vni;
523 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 524
525 list_add_tail_rcu(&rd->list, &f->remotes);
526
9e4b93f9 527 *rdp = rd;
6681712d
DS		 528 return 1;
529}
530
dfd8645e
TH		 531static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
532 unsigned int off,
533 struct vxlanhdr *vh, size_t hdrlen,
54bfd872
JB		 534 __be32 vni_field,
535 struct gro_remcsum *grc,
0ace2ca8 536 bool nopartial)
dfd8645e 537{
b7fe10e5 538 size_t start, offset;
dfd8645e
TH		 539
540 if (skb->remcsum_offload)
b7fe10e5 541 return vh;
dfd8645e
TH		 542
543 if (!NAPI_GRO_CB(skb)->csum_valid)
544 return NULL;
545
54bfd872
JB		 546 start = vxlan_rco_start(vni_field);
547 offset = start + vxlan_rco_offset(vni_field);
dfd8645e 548
b7fe10e5
TH		 549 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
550 start, offset, grc, nopartial);
dfd8645e
TH		 551
552 skb->remcsum_offload = 1;
553
554 return vh;
555}
556
5602c48c
TH		 557static struct sk_buff **vxlan_gro_receive(struct sock *sk,
558 struct sk_buff **head,
559 struct sk_buff *skb)
dc01e7d3
OG		 560{
561 struct sk_buff *p, **pp = NULL;
562 struct vxlanhdr *vh, *vh2;
9b174d88 563 unsigned int hlen, off_vx;
dc01e7d3 564 int flush = 1;
5602c48c 565 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
54bfd872 566 __be32 flags;
26c4f7da
TH		 567 struct gro_remcsum grc;
568
569 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 570
571 off_vx = skb_gro_offset(skb);
572 hlen = off_vx + sizeof(*vh);
573 vh = skb_gro_header_fast(skb, off_vx);
574 if (skb_gro_header_hard(skb, hlen)) {
575 vh = skb_gro_header_slow(skb, hlen, off_vx);
576 if (unlikely(!vh))
577 goto out;
578 }
dc01e7d3 579
dfd8645e
TH		 580 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
581
54bfd872 582 flags = vh->vx_flags;
dfd8645e
TH		 583
584 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
585 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
54bfd872 586 vh->vx_vni, &grc,
0ace2ca8
TH		 587 !!(vs->flags &
588 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 589
590 if (!vh)
591 goto out;
592 }
593
b7fe10e5
TH		 594 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
595
dc01e7d3
OG		 596 for (p = *head; p; p = p->next) {
597 if (!NAPI_GRO_CB(p)->same_flow)
598 continue;
599
600 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 601 if (vh->vx_flags != vh2->vx_flags ||
602 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 603 NAPI_GRO_CB(p)->same_flow = 0;
604 continue;
605 }
dc01e7d3
OG		 606 }
607
fcd91dd4 608 pp = call_gro_receive(eth_gro_receive, head, skb);
c194cf93 609 flush = 0;
dc01e7d3 610
dc01e7d3 611out:
26c4f7da 612 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 613 NAPI_GRO_CB(skb)->flush |= flush;
614
615 return pp;
616}
617
5602c48c 618static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
dc01e7d3 619{
229740c6
JR		 620 /* Sets 'skb->inner_mac_header' since we are always called with
621 * 'skb->encapsulation' set.
622 */
9b174d88 623 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 624}
625
d342894c 626/* Add new entry to forwarding table -- assumes lock held */
627static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 628 const u8 *mac, union vxlan_addr *ip,
6681712d 629 __u16 state, __u16 flags,
3ad7a4b1
RP		 630 __be16 port, __be32 src_vni, __be32 vni,
631 __u32 ifindex, __u8 ndm_flags)
d342894c 632{
9e4b93f9 633 struct vxlan_rdst *rd = NULL;
d342894c 634 struct vxlan_fdb *f;
635 int notify = 0;
17b46365 636 int rc;
d342894c 637
3ad7a4b1 638 f = __vxlan_find_mac(vxlan, mac, src_vni);
d342894c 639 if (f) {
640 if (flags & NLM_F_EXCL) {
641 netdev_dbg(vxlan->dev,
642 "lost race to create %pM\n", mac);
643 return -EEXIST;
644 }
645 if (f->state != state) {
646 f->state = state;
647 f->updated = jiffies;
648 notify = 1;
649 }
ae884082
DS		 650 if (f->flags != ndm_flags) {
651 f->flags = ndm_flags;
652 f->updated = jiffies;
653 notify = 1;
654 }
906dc186
TR		 655 if ((flags & NLM_F_REPLACE)) {
656 /* Only change unicasts */
657 if (!(is_multicast_ether_addr(f->eth_addr) ||
658 is_zero_ether_addr(f->eth_addr))) {
60840429 659 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 660 ifindex);
906dc186
TR		 661 } else
662 return -EOPNOTSUPP;
663 }
6681712d 664 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 665 (is_multicast_ether_addr(f->eth_addr) ||
666 is_zero_ether_addr(f->eth_addr))) {
17b46365 667 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
6681712d
DS		 668
669 if (rc < 0)
670 return rc;
671 notify |= rc;
672 }
d342894c 673 } else {
674 if (!(flags & NLM_F_CREATE))
675 return -ENOENT;
676
0dfbdf41
TG		 677 if (vxlan->cfg.addrmax &&
678 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 679 return -ENOSPC;
680
906dc186
TR		 681 /* Disallow replace to add a multicast entry */
682 if ((flags & NLM_F_REPLACE) &&
683 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
684 return -EOPNOTSUPP;
685
e4c7ed41 686 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 687 f = kmalloc(sizeof(*f), GFP_ATOMIC);
688 if (!f)
689 return -ENOMEM;
690
691 notify = 1;
d342894c 692 f->state = state;
ae884082 693 f->flags = ndm_flags;
d342894c 694 f->updated = f->used = jiffies;
3ad7a4b1 695 f->vni = src_vni;
3e61aa8f 696 INIT_LIST_HEAD(&f->remotes);
d342894c 697 memcpy(f->eth_addr, mac, ETH_ALEN);
698
17b46365
HY		 699 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
700 if (rc < 0) {
701 kfree(f);
702 return rc;
703 }
3e61aa8f 704
d342894c 705 ++vxlan->addrcnt;
706 hlist_add_head_rcu(&f->hlist,
3ad7a4b1 707 vxlan_fdb_head(vxlan, mac, src_vni));
d342894c 708 }
709
9e4b93f9
ND		 710 if (notify) {
711 if (rd == NULL)
712 rd = first_remote_rtnl(f);
713 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
714 }
d342894c 715
716 return 0;
717}
718
6706c82e 719static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 720{
721 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 722 struct vxlan_rdst *rd, *nd;
6681712d 723
0c1d70af
PA		 724 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
725 dst_cache_destroy(&rd->dst_cache);
6681712d 726 kfree(rd);
0c1d70af 727 }
6681712d
DS		 728 kfree(f);
729}
730
d342894c 731static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
732{
733 netdev_dbg(vxlan->dev,
734 "delete %pM\n", f->eth_addr);
735
736 --vxlan->addrcnt;
9e4b93f9 737 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 738
739 hlist_del_rcu(&f->hlist);
6681712d 740 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 741}
742
f0b074be 743static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
3ad7a4b1
RP		 744 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
745 __be32 *vni, u32 *ifindex)
d342894c 746{
6681712d 747 struct net *net = dev_net(vxlan->dev);
e4c7ed41 748 int err;
d342894c 749
f0b074be 750 if (tb[NDA_DST]) {
e4c7ed41
CW		 751 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
752 if (err)
753 return err;
f0b074be 754 } else {
e4c7ed41
CW		 755 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
756 if (remote->sa.sa_family == AF_INET) {
757 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
758 ip->sa.sa_family = AF_INET;
759#if IS_ENABLED(CONFIG_IPV6)
760 } else {
761 ip->sin6.sin6_addr = in6addr_any;
762 ip->sa.sa_family = AF_INET6;
763#endif
764 }
f0b074be 765 }
d342894c 766
6681712d 767 if (tb[NDA_PORT]) {
73cf3317 768 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 769 return -EINVAL;
f0b074be
MR		 770 *port = nla_get_be16(tb[NDA_PORT]);
771 } else {
0dfbdf41 772 *port = vxlan->cfg.dst_port;
f0b074be 773 }
6681712d
DS		 774
775 if (tb[NDA_VNI]) {
776 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
777 return -EINVAL;
54bfd872 778 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
f0b074be
MR		 779 } else {
780 *vni = vxlan->default_dst.remote_vni;
781 }
6681712d 782
3ad7a4b1
RP		 783 if (tb[NDA_SRC_VNI]) {
784 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
785 return -EINVAL;
786 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
787 } else {
788 *src_vni = vxlan->default_dst.remote_vni;
789 }
790
6681712d 791 if (tb[NDA_IFINDEX]) {
5abb0029 792 struct net_device *tdev;
6681712d
DS		 793
794 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
795 return -EINVAL;
f0b074be 796 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 797 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 798 if (!tdev)
6681712d 799 return -EADDRNOTAVAIL;
f0b074be
MR		 800 } else {
801 *ifindex = 0;
802 }
803
804 return 0;
805}
806
807/* Add static entry (via netlink) */
808static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
809 struct net_device *dev,
f6f6424b 810 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 811{
812 struct vxlan_dev *vxlan = netdev_priv(dev);
813 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 814 union vxlan_addr ip;
f0b074be 815 __be16 port;
3ad7a4b1 816 __be32 src_vni, vni;
54bfd872 817 u32 ifindex;
f0b074be
MR		 818 int err;
819
820 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
821 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
822 ndm->ndm_state);
823 return -EINVAL;
824 }
825
826 if (tb[NDA_DST] == NULL)
827 return -EINVAL;
828
3ad7a4b1 829 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
f0b074be
MR		 830 if (err)
831 return err;
6681712d 832
5933a7bb
MR		 833 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
834 return -EAFNOSUPPORT;
835
d342894c 836 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 837 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
3ad7a4b1 838 port, src_vni, vni, ifindex, ndm->ndm_flags);
d342894c 839 spin_unlock_bh(&vxlan->hash_lock);
840
841 return err;
842}
843
3ad7a4b1
RP		 844static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
845 const unsigned char *addr, union vxlan_addr ip,
846 __be16 port, __be32 src_vni, u32 vni, u32 ifindex,
847 u16 vid)
d342894c 848{
d342894c 849 struct vxlan_fdb *f;
bc7892ba 850 struct vxlan_rdst *rd = NULL;
3ad7a4b1 851 int err = -ENOENT;
bc7892ba 852
3ad7a4b1 853 f = vxlan_find_mac(vxlan, addr, src_vni);
bc7892ba 854 if (!f)
3ad7a4b1 855 return err;
bc7892ba 856
e4c7ed41
CW		 857 if (!vxlan_addr_any(&ip)) {
858 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 859 if (!rd)
860 goto out;
861 }
862
bc7892ba
MR		 863 /* remove a destination if it's not the only one on the list,
864 * otherwise destroy the fdb entry
865 */
866 if (rd && !list_is_singular(&f->remotes)) {
867 list_del_rcu(&rd->list);
9e4b93f9 868 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 869 kfree_rcu(rd, rcu);
bc7892ba 870 goto out;
d342894c 871 }
bc7892ba
MR		 872
873 vxlan_fdb_destroy(vxlan, f);
874
875out:
3ad7a4b1
RP		 876 return 0;
877}
878
879/* Delete entry (via netlink) */
880static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
881 struct net_device *dev,
882 const unsigned char *addr, u16 vid)
883{
884 struct vxlan_dev *vxlan = netdev_priv(dev);
885 union vxlan_addr ip;
886 __be32 src_vni, vni;
887 __be16 port;
888 u32 ifindex;
889 int err;
890
891 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
892 if (err)
893 return err;
894
895 spin_lock_bh(&vxlan->hash_lock);
896 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
897 vid);
d342894c 898 spin_unlock_bh(&vxlan->hash_lock);
899
900 return err;
901}
902
903/* Dump forwarding table */
904static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3 905 struct net_device *dev,
d297653d 906 struct net_device *filter_dev, int *idx)
d342894c 907{
908 struct vxlan_dev *vxlan = netdev_priv(dev);
909 unsigned int h;
d297653d 910 int err = 0;
d342894c 911
912 for (h = 0; h < FDB_HASH_SIZE; ++h) {
913 struct vxlan_fdb *f;
d342894c 914
b67bfe0d 915 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 916 struct vxlan_rdst *rd;
6681712d 917
3e61aa8f 918 list_for_each_entry_rcu(rd, &f->remotes, list) {
d297653d 919 if (*idx < cb->args[2])
07a51cd3
AW		 920 goto skip;
921
6681712d
DS		 922 err = vxlan_fdb_info(skb, vxlan, f,
923 NETLINK_CB(cb->skb).portid,
924 cb->nlh->nlmsg_seq,
925 RTM_NEWNEIGH,
926 NLM_F_MULTI, rd);
d297653d 927 if (err < 0)
3e61aa8f 928 goto out;
3e61aa8f 929skip:
d297653d 930 *idx += 1;
07a51cd3 931 }
d342894c 932 }
933 }
3e61aa8f 934out:
d297653d 935 return err;
d342894c 936}
937
938/* Watch incoming packets to learn mapping between Ethernet address
939 * and Tunnel endpoint.
c4b49512 940 * Return true if packet is bogus and should be dropped.
d342894c 941 */
26a41ae6 942static bool vxlan_snoop(struct net_device *dev,
3ad7a4b1
RP		 943 union vxlan_addr *src_ip, const u8 *src_mac,
944 __be32 vni)
d342894c 945{
946 struct vxlan_dev *vxlan = netdev_priv(dev);
947 struct vxlan_fdb *f;
d342894c 948
3ad7a4b1 949 f = vxlan_find_mac(vxlan, src_mac, vni);
d342894c 950 if (likely(f)) {
5ca5461c 951 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 952
e4c7ed41 953 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 954 return false;
955
956 /* Don't migrate static entries, drop packets */
eb064c3b 957 if (f->state & NUD_NOARP)
26a41ae6 958 return true;
d342894c 959
960 if (net_ratelimit())
961 netdev_info(dev,
e4c7ed41 962 "%pM migrated from %pIS to %pIS\n",
a4870f79 963 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 964
e4c7ed41 965 rdst->remote_ip = *src_ip;
d342894c 966 f->updated = jiffies;
9e4b93f9 967 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 968 } else {
969 /* learned new entry */
970 spin_lock(&vxlan->hash_lock);
3bf74b1a 971
972 /* close off race between vxlan_flush and incoming packets */
973 if (netif_running(dev))
974 vxlan_fdb_create(vxlan, src_mac, src_ip,
975 NUD_REACHABLE,
976 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 977 vxlan->cfg.dst_port,
3ad7a4b1 978 vni,
3bf74b1a 979 vxlan->default_dst.remote_vni,
980 0, NTF_SELF);
d342894c 981 spin_unlock(&vxlan->hash_lock);
982 }
26a41ae6 983
984 return false;
d342894c 985}
986
d342894c 987/* See if multicast group is already in use by other ID */
95ab0991 988static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 989{
553675fb 990 struct vxlan_dev *vxlan;
c6fcc4fc 991 struct vxlan_sock *sock4;
4053ab1b
AB		 992#if IS_ENABLED(CONFIG_IPV6)
993 struct vxlan_sock *sock6;
994#endif
b1be00a6 995 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 996
c6fcc4fc 997 sock4 = rtnl_dereference(dev->vn4_sock);
998
95ab0991
G		 999 /* The vxlan_sock is only used by dev, leaving group has
1000 * no effect on other vxlan devices.
1001 */
c6fcc4fc 1002 if (family == AF_INET && sock4 && atomic_read(&sock4->refcnt) == 1)
95ab0991 1003 return false;
b1be00a6 1004#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1005 sock6 = rtnl_dereference(dev->vn6_sock);
1006 if (family == AF_INET6 && sock6 && atomic_read(&sock6->refcnt) == 1)
b1be00a6
JB		 1007 return false;
1008#endif
95ab0991 1009
553675fb 1010 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1011 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1012 continue;
d342894c 1013
c6fcc4fc 1014 if (family == AF_INET &&
1015 rtnl_dereference(vxlan->vn4_sock) != sock4)
95ab0991 1016 continue;
b1be00a6 1017#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1018 if (family == AF_INET6 &&
1019 rtnl_dereference(vxlan->vn6_sock) != sock6)
b1be00a6
JB		 1020 continue;
1021#endif
95ab0991
G		 1022
1023 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1024 &dev->default_dst.remote_ip))
1025 continue;
1026
1027 if (vxlan->default_dst.remote_ifindex !=
1028 dev->default_dst.remote_ifindex)
1029 continue;
1030
1031 return true;
553675fb 1032 }
d342894c 1033
1034 return false;
1035}
1036
544a773a 1037static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
7c47cedf 1038{
b1be00a6 1039 struct vxlan_net *vn;
012a5729 1040
b1be00a6 1041 if (!vs)
544a773a 1042 return false;
7c47cedf 1043 if (!atomic_dec_and_test(&vs->refcnt))
544a773a 1044 return false;
d342894c 1045
b1be00a6 1046 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1047 spin_lock(&vn->sock_lock);
7c47cedf 1048 hlist_del_rcu(&vs->hlist);
e7b3db5e 1049 udp_tunnel_notify_del_rx_port(vs->sock,
b9adcd69
AD		 1050 (vs->flags & VXLAN_F_GPE) ?
1051 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 1052 UDP_TUNNEL_TYPE_VXLAN);
1c51a915
SH		 1053 spin_unlock(&vn->sock_lock);
1054
544a773a 1055 return true;
d342894c 1056}
1057
b1be00a6
JB		 1058static void vxlan_sock_release(struct vxlan_dev *vxlan)
1059{
c6fcc4fc 1060 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
b1be00a6 1061#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1062 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1063
1064 rcu_assign_pointer(vxlan->vn6_sock, NULL);
544a773a
HFS		 1065#endif
1066
c6fcc4fc 1067 rcu_assign_pointer(vxlan->vn4_sock, NULL);
544a773a
HFS		 1068 synchronize_net();
1069
c6fcc4fc 1070 if (__vxlan_sock_release_prep(sock4)) {
1071 udp_tunnel_sock_release(sock4->sock);
1072 kfree(sock4);
544a773a
HFS		 1073 }
1074
1075#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 1076 if (__vxlan_sock_release_prep(sock6)) {
1077 udp_tunnel_sock_release(sock6->sock);
1078 kfree(sock6);
544a773a 1079 }
b1be00a6
JB		 1080#endif
1081}
1082
56ef9c90 1083/* Update multicast group membership when first VNI on
c4b49512 1084 * multicast address is brought up
7c47cedf 1085 */
56ef9c90 1086static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1087{
b1be00a6 1088 struct sock *sk;
e4c7ed41
CW		 1089 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1090 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1091 int ret = -EINVAL;
d342894c 1092
e4c7ed41 1093 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1094 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1095 struct ip_mreqn mreq = {
1096 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1097 .imr_ifindex = ifindex,
1098 };
1099
c6fcc4fc 1100 sk = sock4->sock->sk;
b1be00a6 1101 lock_sock(sk);
56ef9c90 1102 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1103 release_sock(sk);
e4c7ed41
CW		 1104#if IS_ENABLED(CONFIG_IPV6)
1105 } else {
c6fcc4fc 1106 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1107
1108 sk = sock6->sock->sk;
b1be00a6 1109 lock_sock(sk);
56ef9c90
MRL		 1110 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1111 &ip->sin6.sin6_addr);
b1be00a6 1112 release_sock(sk);
e4c7ed41
CW		 1113#endif
1114 }
3fc2de2f 1115
56ef9c90 1116 return ret;
3fc2de2f 1117}
1118
1119/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1120static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1121{
b1be00a6 1122 struct sock *sk;
e4c7ed41
CW		 1123 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1124 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1125 int ret = -EINVAL;
3fc2de2f 1126
e4c7ed41 1127 if (ip->sa.sa_family == AF_INET) {
c6fcc4fc 1128 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
e4c7ed41
CW		 1129 struct ip_mreqn mreq = {
1130 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1131 .imr_ifindex = ifindex,
1132 };
1133
c6fcc4fc 1134 sk = sock4->sock->sk;
b1be00a6 1135 lock_sock(sk);
56ef9c90 1136 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1137 release_sock(sk);
e4c7ed41
CW		 1138#if IS_ENABLED(CONFIG_IPV6)
1139 } else {
c6fcc4fc 1140 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1141
1142 sk = sock6->sock->sk;
b1be00a6 1143 lock_sock(sk);
56ef9c90
MRL		 1144 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1145 &ip->sin6.sin6_addr);
b1be00a6 1146 release_sock(sk);
e4c7ed41
CW		 1147#endif
1148 }
d342894c 1149
56ef9c90 1150 return ret;
d342894c 1151}
1152
f14ecebb
JB		 1153static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1154 struct sk_buff *skb, u32 vxflags)
dfd8645e 1155{
7d34fa75 1156 size_t start, offset;
dfd8645e 1157
f14ecebb
JB		 1158 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1159 goto out;
b7fe10e5 1160
f14ecebb
JB		 1161 start = vxlan_rco_start(unparsed->vx_vni);
1162 offset = start + vxlan_rco_offset(unparsed->vx_vni);
dfd8645e 1163
7d34fa75 1164 if (!pskb_may_pull(skb, offset + sizeof(u16)))
be5cfeab 1165 return false;
dfd8645e 1166
be5cfeab
JB		 1167 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1168 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
f14ecebb
JB		 1169out:
1170 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1171 unparsed->vx_vni &= VXLAN_VNI_MASK;
be5cfeab 1172 return true;
dfd8645e
TH		 1173}
1174
f14ecebb 1175static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
64f87d36 1176 struct sk_buff *skb, u32 vxflags,
10a5af23 1177 struct vxlan_metadata *md)
3288af08 1178{
f14ecebb 1179 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
10a5af23 1180 struct metadata_dst *tun_dst;
f14ecebb
JB		 1181
1182 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1183 goto out;
3288af08 1184
3288af08
JB		 1185 md->gbp = ntohs(gbp->policy_id);
1186
10a5af23 1187 tun_dst = (struct metadata_dst *)skb_dst(skb);
810813c4 1188 if (tun_dst) {
3288af08 1189 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
810813c4
DM		 1190 tun_dst->u.tun_info.options_len = sizeof(*md);
1191 }
3288af08
JB		 1192 if (gbp->dont_learn)
1193 md->gbp |= VXLAN_GBP_DONT_LEARN;
1194
1195 if (gbp->policy_applied)
1196 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
f14ecebb 1197
64f87d36
JB		 1198 /* In flow-based mode, GBP is carried in dst_metadata */
1199 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1200 skb->mark = md->gbp;
f14ecebb
JB		 1201out:
1202 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
3288af08
JB		 1203}
1204
e1e5314d 1205static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
61618eea 1206 __be16 *protocol,
e1e5314d
JB		 1207 struct sk_buff *skb, u32 vxflags)
1208{
1209 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1210
1211 /* Need to have Next Protocol set for interfaces in GPE mode. */
1212 if (!gpe->np_applied)
1213 return false;
1214 /* "The initial version is 0. If a receiver does not support the
1215 * version indicated it MUST drop the packet.
1216 */
1217 if (gpe->version != 0)
1218 return false;
1219 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1220 * processing MUST occur." However, we don't implement OAM
1221 * processing, thus drop the packet.
1222 */
1223 if (gpe->oam_flag)
1224 return false;
1225
1226 switch (gpe->next_protocol) {
1227 case VXLAN_GPE_NP_IPV4:
1228 *protocol = htons(ETH_P_IP);
1229 break;
1230 case VXLAN_GPE_NP_IPV6:
1231 *protocol = htons(ETH_P_IPV6);
1232 break;
1233 case VXLAN_GPE_NP_ETHERNET:
1234 *protocol = htons(ETH_P_TEB);
1235 break;
1236 default:
1237 return false;
1238 }
1239
1240 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1241 return true;
1242}
1243
1ab016e2
JB		 1244static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1245 struct vxlan_sock *vs,
3ad7a4b1 1246 struct sk_buff *skb, __be32 vni)
614732ea 1247{
614732ea 1248 union vxlan_addr saddr;
614732ea 1249
614732ea 1250 skb_reset_mac_header(skb);
614732ea
TG		 1251 skb->protocol = eth_type_trans(skb, vxlan->dev);
1252 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1253
1254 /* Ignore packet loops (and multicast echo) */
1255 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1ab016e2 1256 return false;
614732ea 1257
760c6805 1258 /* Get address from the outer IP header */
ce212d0f 1259 if (vxlan_get_sk_family(vs) == AF_INET) {
1ab016e2 1260 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
614732ea
TG		 1261 saddr.sa.sa_family = AF_INET;
1262#if IS_ENABLED(CONFIG_IPV6)
1263 } else {
1ab016e2 1264 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
614732ea
TG		 1265 saddr.sa.sa_family = AF_INET6;
1266#endif
1267 }
1268
1ab016e2 1269 if ((vxlan->flags & VXLAN_F_LEARN) &&
3ad7a4b1 1270 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, vni))
1ab016e2
JB		 1271 return false;
1272
1273 return true;
1274}
1275
760c6805
JB		 1276static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1277 struct sk_buff *skb)
1278{
1279 int err = 0;
1280
1281 if (vxlan_get_sk_family(vs) == AF_INET)
1282 err = IP_ECN_decapsulate(oiph, skb);
1283#if IS_ENABLED(CONFIG_IPV6)
1284 else
1285 err = IP6_ECN_decapsulate(oiph, skb);
1286#endif
1287
1288 if (unlikely(err) && log_ecn_error) {
1289 if (vxlan_get_sk_family(vs) == AF_INET)
1290 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1291 &((struct iphdr *)oiph)->saddr,
1292 ((struct iphdr *)oiph)->tos);
1293 else
1294 net_info_ratelimited("non-ECT from %pI6\n",
1295 &((struct ipv6hdr *)oiph)->saddr);
1296 }
1297 return err <= 1;
1298}
1299
d342894c 1300/* Callback from net/ipv4/udp.c to receive packets */
f2d1968e 1301static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
d342894c 1302{
f2d1968e 1303 struct pcpu_sw_netstats *stats;
c9e78efb 1304 struct vxlan_dev *vxlan;
5cfccc5a 1305 struct vxlan_sock *vs;
f14ecebb 1306 struct vxlanhdr unparsed;
ee122c79
TG		 1307 struct vxlan_metadata _md;
1308 struct vxlan_metadata *md = &_md;
61618eea 1309 __be16 protocol = htons(ETH_P_TEB);
e1e5314d 1310 bool raw_proto = false;
f2d1968e 1311 void *oiph;
3ad7a4b1 1312 __be32 vni = 0;
d342894c 1313
e1e5314d 1314 /* Need UDP and VXLAN header to be present */
7ce04758 1315 if (!pskb_may_pull(skb, VXLAN_HLEN))
e5aed006 1316 goto drop;
d342894c 1317
f14ecebb 1318 unparsed = *vxlan_hdr(skb);
288b01c8
JB		 1319 /* VNI flag always required to be set */
1320 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1321 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1322 ntohl(vxlan_hdr(skb)->vx_flags),
1323 ntohl(vxlan_hdr(skb)->vx_vni));
1324 /* Return non vxlan pkt */
e5aed006 1325 goto drop;
d342894c 1326 }
288b01c8
JB		 1327 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1328 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
d342894c 1329
559835ea 1330 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1331 if (!vs)
d342894c 1332 goto drop;
d342894c 1333
3ad7a4b1
RP		 1334 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1335
3ad7a4b1 1336 vxlan = vxlan_vs_find_vni(vs, vni);
c9e78efb
JB		 1337 if (!vxlan)
1338 goto drop;
1339
e1e5314d
JB		 1340 /* For backwards compatibility, only allow reserved fields to be
1341 * used by VXLAN extensions if explicitly requested.
1342 */
1343 if (vs->flags & VXLAN_F_GPE) {
1344 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1345 goto drop;
1346 raw_proto = true;
1347 }
1348
1349 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1350 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1351 goto drop;
c9e78efb 1352
ee122c79 1353 if (vxlan_collect_metadata(vs)) {
10a5af23 1354 struct metadata_dst *tun_dst;
07dabf20 1355
c29a70d2 1356 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
d817f432 1357 key32_to_tunnel_id(vni), sizeof(*md));
c29a70d2 1358
ee122c79
TG		 1359 if (!tun_dst)
1360 goto drop;
1361
0f1b7354 1362 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
10a5af23
JB		 1363
1364 skb_dst_set(skb, (struct dst_entry *)tun_dst);
ee122c79
TG		 1365 } else {
1366 memset(md, 0, sizeof(*md));
1367 }
1368
f14ecebb
JB		 1369 if (vs->flags & VXLAN_F_REMCSUM_RX)
1370 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1371 goto drop;
1372 if (vs->flags & VXLAN_F_GBP)
10a5af23 1373 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
e1e5314d
JB		 1374 /* Note that GBP and GPE can never be active together. This is
1375 * ensured in vxlan_dev_configure.
1376 */
3511494c 1377
f14ecebb 1378 if (unparsed.vx_flags || unparsed.vx_vni) {
3bf39475
TH		 1379 /* If there are any unprocessed flags remaining treat
1380 * this as a malformed packet. This behavior diverges from
1381 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1382 * in reserved fields are to be ignored. The approach here
c4b49512 1383 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1384 * is more robust and provides a little more security in
1385 * adding extensions to VXLAN.
1386 */
288b01c8 1387 goto drop;
3bf39475
TH		 1388 }
1389
e1e5314d 1390 if (!raw_proto) {
3ad7a4b1 1391 if (!vxlan_set_mac(vxlan, vs, skb, vni))
e1e5314d
JB		 1392 goto drop;
1393 } else {
8be0cfa4 1394 skb_reset_mac_header(skb);
e1e5314d
JB		 1395 skb->dev = vxlan->dev;
1396 skb->pkt_type = PACKET_HOST;
1397 }
f2d1968e 1398
f2d1968e
JB		 1399 oiph = skb_network_header(skb);
1400 skb_reset_network_header(skb);
1401
1402 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1403 ++vxlan->dev->stats.rx_frame_errors;
1404 ++vxlan->dev->stats.rx_errors;
1405 goto drop;
1406 }
1407
1408 stats = this_cpu_ptr(vxlan->dev->tstats);
1409 u64_stats_update_begin(&stats->syncp);
1410 stats->rx_packets++;
1411 stats->rx_bytes += skb->len;
1412 u64_stats_update_end(&stats->syncp);
1413
1414 gro_cells_receive(&vxlan->gro_cells, skb);
5cfccc5a
PS		 1415 return 0;
1416
1417drop:
288b01c8
JB		 1418 /* Consume bad packet */
1419 kfree_skb(skb);
1420 return 0;
5cfccc5a
PS		 1421}
1422
3ad7a4b1 1423static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
e4f67add
DS		 1424{
1425 struct vxlan_dev *vxlan = netdev_priv(dev);
1426 struct arphdr *parp;
1427 u8 *arpptr, *sha;
1428 __be32 sip, tip;
1429 struct neighbour *n;
1430
1431 if (dev->flags & IFF_NOARP)
1432 goto out;
1433
1434 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1435 dev->stats.tx_dropped++;
1436 goto out;
1437 }
1438 parp = arp_hdr(skb);
1439
1440 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1441 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1442 parp->ar_pro != htons(ETH_P_IP) ||
1443 parp->ar_op != htons(ARPOP_REQUEST) ||
1444 parp->ar_hln != dev->addr_len ||
1445 parp->ar_pln != 4)
1446 goto out;
1447 arpptr = (u8 *)parp + sizeof(struct arphdr);
1448 sha = arpptr;
1449 arpptr += dev->addr_len; /* sha */
1450 memcpy(&sip, arpptr, sizeof(sip));
1451 arpptr += sizeof(sip);
1452 arpptr += dev->addr_len; /* tha */
1453 memcpy(&tip, arpptr, sizeof(tip));
1454
1455 if (ipv4_is_loopback(tip) ||
1456 ipv4_is_multicast(tip))
1457 goto out;
1458
1459 n = neigh_lookup(&arp_tbl, &tip, dev);
1460
1461 if (n) {
e4f67add
DS		 1462 struct vxlan_fdb *f;
1463 struct sk_buff *reply;
1464
1465 if (!(n->nud_state & NUD_CONNECTED)) {
1466 neigh_release(n);
1467 goto out;
1468 }
1469
3ad7a4b1 1470 f = vxlan_find_mac(vxlan, n->ha, vni);
e4c7ed41 1471 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1472 /* bridge-local neighbor */
1473 neigh_release(n);
1474 goto out;
1475 }
1476
1477 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1478 n->ha, sha);
1479
1480 neigh_release(n);
1481
7346135d
DS		 1482 if (reply == NULL)
1483 goto out;
1484
e4f67add
DS		 1485 skb_reset_mac_header(reply);
1486 __skb_pull(reply, skb_network_offset(reply));
1487 reply->ip_summed = CHECKSUM_UNNECESSARY;
1488 reply->pkt_type = PACKET_HOST;
1489
1490 if (netif_rx_ni(reply) == NET_RX_DROP)
1491 dev->stats.rx_dropped++;
e4c7ed41
CW		 1492 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1493 union vxlan_addr ipa = {
1494 .sin.sin_addr.s_addr = tip,
a45e92a5 1495 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1496 };
1497
1498 vxlan_ip_miss(dev, &ipa);
1499 }
e4f67add
DS		 1500out:
1501 consume_skb(skb);
1502 return NETDEV_TX_OK;
1503}
1504
f564f45c 1505#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1506static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1507 struct neighbour *n, bool isrouter)
1508{
1509 struct net_device *dev = request->dev;
1510 struct sk_buff *reply;
1511 struct nd_msg *ns, *na;
1512 struct ipv6hdr *pip6;
1513 u8 *daddr;
1514 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1515 int ns_olen;
1516 int i, len;
1517
f1fb08f6 1518 if (dev == NULL || !pskb_may_pull(request, request->len))
4b29dba9
DS		 1519 return NULL;
1520
1521 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1522 sizeof(*na) + na_olen + dev->needed_tailroom;
1523 reply = alloc_skb(len, GFP_ATOMIC);
1524 if (reply == NULL)
1525 return NULL;
1526
1527 reply->protocol = htons(ETH_P_IPV6);
1528 reply->dev = dev;
1529 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1530 skb_push(reply, sizeof(struct ethhdr));
6297b91c 1531 skb_reset_mac_header(reply);
4b29dba9 1532
f1fb08f6 1533 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
4b29dba9
DS		 1534
1535 daddr = eth_hdr(request)->h_source;
f1fb08f6
VB		 1536 ns_olen = request->len - skb_network_offset(request) -
1537 sizeof(struct ipv6hdr) - sizeof(*ns);
4b29dba9
DS		 1538 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1539 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1540 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1541 break;
1542 }
1543 }
1544
1545 /* Ethernet header */
1546 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1547 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1548 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1549 reply->protocol = htons(ETH_P_IPV6);
1550
1551 skb_pull(reply, sizeof(struct ethhdr));
6297b91c 1552 skb_reset_network_header(reply);
4b29dba9
DS		 1553 skb_put(reply, sizeof(struct ipv6hdr));
1554
1555 /* IPv6 header */
1556
1557 pip6 = ipv6_hdr(reply);
1558 memset(pip6, 0, sizeof(struct ipv6hdr));
1559 pip6->version = 6;
1560 pip6->priority = ipv6_hdr(request)->priority;
1561 pip6->nexthdr = IPPROTO_ICMPV6;
1562 pip6->hop_limit = 255;
1563 pip6->daddr = ipv6_hdr(request)->saddr;
1564 pip6->saddr = *(struct in6_addr *)n->primary_key;
1565
1566 skb_pull(reply, sizeof(struct ipv6hdr));
6297b91c 1567 skb_reset_transport_header(reply);
4b29dba9
DS		 1568
1569 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1570
1571 /* Neighbor Advertisement */
1572 memset(na, 0, sizeof(*na)+na_olen);
1573 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1574 na->icmph.icmp6_router = isrouter;
1575 na->icmph.icmp6_override = 1;
1576 na->icmph.icmp6_solicited = 1;
1577 na->target = ns->target;
1578 ether_addr_copy(&na->opt[2], n->ha);
1579 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1580 na->opt[1] = na_olen >> 3;
1581
1582 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1583 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1584 csum_partial(na, sizeof(*na)+na_olen, 0));
1585
1586 pip6->payload_len = htons(sizeof(*na)+na_olen);
1587
1588 skb_push(reply, sizeof(struct ipv6hdr));
1589
1590 reply->ip_summed = CHECKSUM_UNNECESSARY;
1591
1592 return reply;
1593}
1594
3ad7a4b1 1595static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
f564f45c
CW		 1596{
1597 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1598 struct nd_msg *msg;
f564f45c 1599 const struct ipv6hdr *iphdr;
8dcd81a9 1600 const struct in6_addr *daddr;
4b29dba9
DS		 1601 struct neighbour *n;
1602 struct inet6_dev *in6_dev;
f564f45c
CW		 1603
1604 in6_dev = __in6_dev_get(dev);
1605 if (!in6_dev)
1606 goto out;
1607
f1fb08f6
VB		 1608 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + sizeof(struct nd_msg)))
1609 goto out;
1610
f564f45c 1611 iphdr = ipv6_hdr(skb);
f564f45c
CW		 1612 daddr = &iphdr->daddr;
1613
f1fb08f6 1614 msg = (struct nd_msg *)(iphdr + 1);
f564f45c
CW		 1615 if (msg->icmph.icmp6_code != 0 ||
1616 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1617 goto out;
1618
4b29dba9
DS		 1619 if (ipv6_addr_loopback(daddr) ||
1620 ipv6_addr_is_multicast(&msg->target))
1621 goto out;
1622
1623 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1624
1625 if (n) {
1626 struct vxlan_fdb *f;
4b29dba9 1627 struct sk_buff *reply;
f564f45c
CW		 1628
1629 if (!(n->nud_state & NUD_CONNECTED)) {
1630 neigh_release(n);
1631 goto out;
1632 }
1633
3ad7a4b1 1634 f = vxlan_find_mac(vxlan, n->ha, vni);
f564f45c
CW		 1635 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1636 /* bridge-local neighbor */
1637 neigh_release(n);
1638 goto out;
1639 }
1640
4b29dba9
DS		 1641 reply = vxlan_na_create(skb, n,
1642 !!(f ? f->flags & NTF_ROUTER : 0));
1643
f564f45c 1644 neigh_release(n);
4b29dba9
DS		 1645
1646 if (reply == NULL)
1647 goto out;
1648
1649 if (netif_rx_ni(reply) == NET_RX_DROP)
1650 dev->stats.rx_dropped++;
1651
f564f45c 1652 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1653 union vxlan_addr ipa = {
1654 .sin6.sin6_addr = msg->target,
a45e92a5 1655 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1656 };
1657
f564f45c
CW		 1658 vxlan_ip_miss(dev, &ipa);
1659 }
1660
1661out:
1662 consume_skb(skb);
1663 return NETDEV_TX_OK;
1664}
1665#endif
1666
e4f67add
DS		 1667static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1668{
1669 struct vxlan_dev *vxlan = netdev_priv(dev);
1670 struct neighbour *n;
e4f67add
DS		 1671
1672 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1673 return false;
1674
1675 n = NULL;
1676 switch (ntohs(eth_hdr(skb)->h_proto)) {
1677 case ETH_P_IP:
e15a00aa
CW		 1678 {
1679 struct iphdr *pip;
1680
e4f67add
DS		 1681 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1682 return false;
1683 pip = ip_hdr(skb);
1684 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1685 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1686 union vxlan_addr ipa = {
1687 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1688 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1689 };
1690
1691 vxlan_ip_miss(dev, &ipa);
1692 return false;
1693 }
1694
e4f67add 1695 break;
e15a00aa
CW		 1696 }
1697#if IS_ENABLED(CONFIG_IPV6)
1698 case ETH_P_IPV6:
1699 {
1700 struct ipv6hdr *pip6;
1701
1702 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1703 return false;
1704 pip6 = ipv6_hdr(skb);
1705 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1706 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1707 union vxlan_addr ipa = {
1708 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1709 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1710 };
1711
1712 vxlan_ip_miss(dev, &ipa);
1713 return false;
1714 }
1715
1716 break;
1717 }
1718#endif
e4f67add
DS		 1719 default:
1720 return false;
1721 }
1722
1723 if (n) {
1724 bool diff;
1725
7367d0b5 1726 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1727 if (diff) {
1728 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1729 dev->addr_len);
1730 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1731 }
1732 neigh_release(n);
1733 return diff;
e4c7ed41
CW		 1734 }
1735
e4f67add
DS		 1736 return false;
1737}
1738
af33c1ad 1739static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1740 struct vxlan_metadata *md)
1741{
1742 struct vxlanhdr_gbp *gbp;
1743
db79a621
TG		 1744 if (!md->gbp)
1745 return;
1746
3511494c 1747 gbp = (struct vxlanhdr_gbp *)vxh;
54bfd872 1748 vxh->vx_flags |= VXLAN_HF_GBP;
3511494c
TG		 1749
1750 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1751 gbp->dont_learn = 1;
1752
1753 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1754 gbp->policy_applied = 1;
1755
1756 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1757}
1758
e1e5314d
JB		 1759static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1760 __be16 protocol)
1761{
1762 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1763
1764 gpe->np_applied = 1;
1765
1766 switch (protocol) {
1767 case htons(ETH_P_IP):
1768 gpe->next_protocol = VXLAN_GPE_NP_IPV4;
1769 return 0;
1770 case htons(ETH_P_IPV6):
1771 gpe->next_protocol = VXLAN_GPE_NP_IPV6;
1772 return 0;
1773 case htons(ETH_P_TEB):
1774 gpe->next_protocol = VXLAN_GPE_NP_ETHERNET;
1775 return 0;
1776 }
1777 return -EPFNOSUPPORT;
1778}
1779
f491e56d
JB		 1780static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1781 int iphdr_len, __be32 vni,
1782 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1783 bool udp_sum)
e4c7ed41 1784{
e4c7ed41 1785 struct vxlanhdr *vxh;
e4c7ed41
CW		 1786 int min_headroom;
1787 int err;
dfd8645e 1788 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
e1e5314d 1789 __be16 inner_protocol = htons(ETH_P_TEB);
dfd8645e 1790
af33c1ad 1791 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1792 skb->ip_summed == CHECKSUM_PARTIAL) {
1793 int csum_start = skb_checksum_start_offset(skb);
1794
1795 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1796 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1797 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1798 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1799 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1800 }
e4c7ed41 1801
e4c7ed41 1802 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
4a4f86cc 1803 + VXLAN_HLEN + iphdr_len;
649c5b8b
PS		 1804
1805 /* Need space for new headers (invalidates iph ptr) */
1806 err = skb_cow_head(skb, min_headroom);
e1e5314d 1807 if (unlikely(err))
c46b7897 1808 return err;
649c5b8b 1809
aed069df
AD		 1810 err = iptunnel_handle_offloads(skb, type);
1811 if (err)
c46b7897 1812 return err;
b736a623 1813
49560532 1814 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
54bfd872
JB		 1815 vxh->vx_flags = VXLAN_HF_VNI;
1816 vxh->vx_vni = vxlan_vni_field(vni);
49560532 1817
dfd8645e 1818 if (type & SKB_GSO_TUNNEL_REMCSUM) {
54bfd872 1819 unsigned int start;
dfd8645e 1820
54bfd872
JB		 1821 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1822 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1823 vxh->vx_flags |= VXLAN_HF_RCO;
dfd8645e
TH		 1824
1825 if (!skb_is_gso(skb)) {
1826 skb->ip_summed = CHECKSUM_NONE;
1827 skb->encapsulation = 0;
1828 }
1829 }
1830
af33c1ad
TH		 1831 if (vxflags & VXLAN_F_GBP)
1832 vxlan_build_gbp_hdr(vxh, vxflags, md);
e1e5314d
JB		 1833 if (vxflags & VXLAN_F_GPE) {
1834 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1835 if (err < 0)
c46b7897 1836 return err;
e1e5314d
JB		 1837 inner_protocol = skb->protocol;
1838 }
3511494c 1839
e1e5314d 1840 skb_set_inner_protocol(skb, inner_protocol);
039f5062 1841 return 0;
49560532 1842}
49560532 1843
655c3de1 1844static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
1845 struct vxlan_sock *sock4,
1a8496ba 1846 struct sk_buff *skb, int oif, u8 tos,
4ecb1d83 1847 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
0c1d70af 1848 struct dst_cache *dst_cache,
db3c6139 1849 const struct ip_tunnel_info *info)
1a8496ba 1850{
db3c6139 1851 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1a8496ba
JB		 1852 struct rtable *rt = NULL;
1853 struct flowi4 fl4;
1854
655c3de1 1855 if (!sock4)
1856 return ERR_PTR(-EIO);
1857
db3c6139
DB		 1858 if (tos && !info)
1859 use_cache = false;
1860 if (use_cache) {
0c1d70af
PA		 1861 rt = dst_cache_get_ip4(dst_cache, saddr);
1862 if (rt)
1863 return rt;
1864 }
1865
1a8496ba
JB		 1866 memset(&fl4, 0, sizeof(fl4));
1867 fl4.flowi4_oif = oif;
1868 fl4.flowi4_tos = RT_TOS(tos);
1869 fl4.flowi4_mark = skb->mark;
1870 fl4.flowi4_proto = IPPROTO_UDP;
1871 fl4.daddr = daddr;
272d96a5 1872 fl4.saddr = *saddr;
4ecb1d83
MP		 1873 fl4.fl4_dport = dport;
1874 fl4.fl4_sport = sport;
1a8496ba
JB		 1875
1876 rt = ip_route_output_key(vxlan->net, &fl4);
655c3de1 1877 if (likely(!IS_ERR(rt))) {
1878 if (rt->dst.dev == dev) {
1879 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
1880 ip_rt_put(rt);
1881 return ERR_PTR(-ELOOP);
1882 }
1883
1a8496ba 1884 *saddr = fl4.saddr;
0c1d70af
PA		 1885 if (use_cache)
1886 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
655c3de1 1887 } else {
1888 netdev_dbg(dev, "no route to %pI4\n", &daddr);
1889 return ERR_PTR(-ENETUNREACH);
0c1d70af 1890 }
1a8496ba
JB		 1891 return rt;
1892}
1893
e5d4b29f
JB		 1894#if IS_ENABLED(CONFIG_IPV6)
1895static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
655c3de1 1896 struct net_device *dev,
03dc52a8 1897 struct vxlan_sock *sock6,
1400615d 1898 struct sk_buff *skb, int oif, u8 tos,
e7f70af1 1899 __be32 label,
e5d4b29f 1900 const struct in6_addr *daddr,
0c1d70af 1901 struct in6_addr *saddr,
4ecb1d83 1902 __be16 dport, __be16 sport,
db3c6139
DB		 1903 struct dst_cache *dst_cache,
1904 const struct ip_tunnel_info *info)
e5d4b29f 1905{
db3c6139 1906 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
e5d4b29f
JB		 1907 struct dst_entry *ndst;
1908 struct flowi6 fl6;
1909 int err;
1910
c6fcc4fc 1911 if (!sock6)
1912 return ERR_PTR(-EIO);
1913
1400615d
DB		 1914 if (tos && !info)
1915 use_cache = false;
db3c6139 1916 if (use_cache) {
0c1d70af
PA		 1917 ndst = dst_cache_get_ip6(dst_cache, saddr);
1918 if (ndst)
1919 return ndst;
1920 }
1921
e5d4b29f
JB		 1922 memset(&fl6, 0, sizeof(fl6));
1923 fl6.flowi6_oif = oif;
1924 fl6.daddr = *daddr;
272d96a5 1925 fl6.saddr = *saddr;
eaa93bf4 1926 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
e5d4b29f
JB		 1927 fl6.flowi6_mark = skb->mark;
1928 fl6.flowi6_proto = IPPROTO_UDP;
4ecb1d83
MP		 1929 fl6.fl6_dport = dport;
1930 fl6.fl6_sport = sport;
e5d4b29f
JB		 1931
1932 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
c6fcc4fc 1933 sock6->sock->sk,
e5d4b29f 1934 &ndst, &fl6);
655c3de1 1935 if (unlikely(err < 0)) {
1936 netdev_dbg(dev, "no route to %pI6\n", daddr);
1937 return ERR_PTR(-ENETUNREACH);
1938 }
1939
1940 if (unlikely(ndst->dev == dev)) {
1941 netdev_dbg(dev, "circular route to %pI6\n", daddr);
1942 dst_release(ndst);
1943 return ERR_PTR(-ELOOP);
1944 }
e5d4b29f
JB		 1945
1946 *saddr = fl6.saddr;
db3c6139 1947 if (use_cache)
0c1d70af 1948 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 1949 return ndst;
1950}
1951#endif
1952
9dcc71e1
SS		 1953/* Bypass encapsulation if the destination is local */
1954static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
3ad7a4b1 1955 struct vxlan_dev *dst_vxlan, __be32 vni)
9dcc71e1 1956{
8f84985f 1957 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1958 union vxlan_addr loopback;
1959 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1960 struct net_device *dev = skb->dev;
1961 int len = skb->len;
9dcc71e1 1962
8f84985f
LR		 1963 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1964 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1965 skb->pkt_type = PACKET_HOST;
1966 skb->encapsulation = 0;
1967 skb->dev = dst_vxlan->dev;
1968 __skb_pull(skb, skb_network_offset(skb));
1969
e4c7ed41
CW		 1970 if (remote_ip->sa.sa_family == AF_INET) {
1971 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1972 loopback.sa.sa_family = AF_INET;
1973#if IS_ENABLED(CONFIG_IPV6)
1974 } else {
1975 loopback.sin6.sin6_addr = in6addr_loopback;
1976 loopback.sa.sa_family = AF_INET6;
1977#endif
1978 }
1979
9dcc71e1 1980 if (dst_vxlan->flags & VXLAN_F_LEARN)
3ad7a4b1 1981 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source, vni);
9dcc71e1
SS		 1982
1983 u64_stats_update_begin(&tx_stats->syncp);
1984 tx_stats->tx_packets++;
ce6502a8 1985 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1986 u64_stats_update_end(&tx_stats->syncp);
1987
1988 if (netif_rx(skb) == NET_RX_SUCCESS) {
1989 u64_stats_update_begin(&rx_stats->syncp);
1990 rx_stats->rx_packets++;
ce6502a8 1991 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1992 u64_stats_update_end(&rx_stats->syncp);
1993 } else {
ce6502a8 1994 dev->stats.rx_dropped++;
9dcc71e1
SS		 1995 }
1996}
1997
fee1fad7 1998static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
1999 struct vxlan_dev *vxlan, union vxlan_addr *daddr,
1f6cc07e 2000 __be16 dst_port, __be32 vni, struct dst_entry *dst,
fee1fad7 2001 u32 rt_flags)
2002{
2003#if IS_ENABLED(CONFIG_IPV6)
2004 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2005 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2006 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2007 */
2008 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2009#endif
2010 /* Bypass encapsulation if the destination is local */
2011 if (rt_flags & RTCF_LOCAL &&
2012 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2013 struct vxlan_dev *dst_vxlan;
2014
2015 dst_release(dst);
2016 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
2017 daddr->sa.sa_family, dst_port,
2018 vxlan->flags);
2019 if (!dst_vxlan) {
2020 dev->stats.tx_errors++;
2021 kfree_skb(skb);
2022
2023 return -ENOENT;
2024 }
3ad7a4b1 2025 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
fee1fad7 2026 return 1;
2027 }
2028
2029 return 0;
2030}
2031
4ad16930 2032static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
3ad7a4b1
RP		 2033 __be32 default_vni, struct vxlan_rdst *rdst,
2034 bool did_rsc)
d342894c 2035{
d71785ff 2036 struct dst_cache *dst_cache;
3093fbe7 2037 struct ip_tunnel_info *info;
d342894c 2038 struct vxlan_dev *vxlan = netdev_priv(dev);
0770b53b 2039 const struct iphdr *old_iph = ip_hdr(skb);
e4c7ed41 2040 union vxlan_addr *dst;
272d96a5 2041 union vxlan_addr remote_ip, local_ip;
ee122c79
TG		 2042 struct vxlan_metadata _md;
2043 struct vxlan_metadata *md = &_md;
e4c7ed41 2044 __be16 src_port = 0, dst_port;
655c3de1 2045 struct dst_entry *ndst = NULL;
e7f70af1 2046 __be32 vni, label;
d342894c 2047 __u8 tos, ttl;
0e6fbc5b 2048 int err;
ee122c79 2049 u32 flags = vxlan->flags;
b4ed5cad 2050 bool udp_sum = false;
f491e56d 2051 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 2052
61adedf3 2053 info = skb_tunnel_info(skb);
3093fbe7 2054
ee122c79 2055 if (rdst) {
0770b53b 2056 dst = &rdst->remote_ip;
2057 if (vxlan_addr_any(dst)) {
2058 if (did_rsc) {
2059 /* short-circuited back to local bridge */
3ad7a4b1 2060 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
0770b53b 2061 return;
2062 }
2063 goto drop;
2064 }
2065
0dfbdf41 2066 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
3ad7a4b1 2067 vni = (rdst->remote_vni) ? : default_vni;
1158632b 2068 local_ip = vxlan->cfg.saddr;
d71785ff 2069 dst_cache = &rdst->dst_cache;
0770b53b 2070 md->gbp = skb->mark;
2071 ttl = vxlan->cfg.ttl;
2072 if (!ttl && vxlan_addr_multicast(dst))
2073 ttl = 1;
2074
2075 tos = vxlan->cfg.tos;
2076 if (tos == 1)
2077 tos = ip_tunnel_get_dsfield(old_iph, skb);
2078
2079 if (dst->sa.sa_family == AF_INET)
2080 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2081 else
2082 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2083 label = vxlan->cfg.label;
ee122c79
TG		 2084 } else {
2085 if (!info) {
2086 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2087 dev->name);
2088 goto drop;
2089 }
b1be00a6 2090 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
272d96a5 2091 if (remote_ip.sa.sa_family == AF_INET) {
a725e514 2092 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
272d96a5 2093 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2094 } else {
a725e514 2095 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
272d96a5 2096 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2097 }
ee122c79 2098 dst = &remote_ip;
0770b53b 2099 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2100 vni = tunnel_id_to_key32(info->key.tun_id);
d71785ff 2101 dst_cache = &info->dst_cache;
0770b53b 2102 if (info->options_len)
2103 md = ip_tunnel_info_opts(info);
a725e514
JB		 2104 ttl = info->key.ttl;
2105 tos = info->key.tos;
e7f70af1 2106 label = info->key.label;
b4ed5cad 2107 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514 2108 }
0770b53b 2109 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2110 vxlan->cfg.port_max, true);
a725e514 2111
56de859e 2112 rcu_read_lock();
e4c7ed41 2113 if (dst->sa.sa_family == AF_INET) {
c6fcc4fc 2114 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
c46b7897 2115 struct rtable *rt;
0770b53b 2116 __be16 df = 0;
c6fcc4fc 2117
655c3de1 2118 rt = vxlan_get_route(vxlan, dev, sock4, skb,
1a8496ba 2119 rdst ? rdst->remote_ifindex : 0, tos,
272d96a5 2120 dst->sin.sin_addr.s_addr,
1158632b 2121 &local_ip.sin.sin_addr.s_addr,
4ecb1d83 2122 dst_port, src_port,
d71785ff 2123 dst_cache, info);
8ebd115b
DM		 2124 if (IS_ERR(rt)) {
2125 err = PTR_ERR(rt);
c46b7897 2126 goto tx_error;
8ebd115b 2127 }
e4c7ed41
CW		 2128
2129 /* Bypass encapsulation if the destination is local */
fee1fad7 2130 if (!info) {
2131 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2132 dst_port, vni, &rt->dst,
2133 rt->rt_flags);
2134 if (err)
56de859e 2135 goto out_unlock;
fee1fad7 2136 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
6ceb31ca 2137 df = htons(IP_DF);
fee1fad7 2138 }
6ceb31ca 2139
c46b7897 2140 ndst = &rt->dst;
e4c7ed41
CW		 2141 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2142 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
c46b7897 2143 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
54bfd872 2144 vni, md, flags, udp_sum);
f491e56d 2145 if (err < 0)
c46b7897 2146 goto tx_error;
f491e56d 2147
1158632b 2148 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
f491e56d
JB		 2149 dst->sin.sin_addr.s_addr, tos, ttl, df,
2150 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2151#if IS_ENABLED(CONFIG_IPV6)
2152 } else {
c6fcc4fc 2153 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
e4c7ed41 2154
655c3de1 2155 ndst = vxlan6_get_route(vxlan, dev, sock6, skb,
1400615d 2156 rdst ? rdst->remote_ifindex : 0, tos,
272d96a5 2157 label, &dst->sin6.sin6_addr,
1158632b 2158 &local_ip.sin6.sin6_addr,
4ecb1d83 2159 dst_port, src_port,
db3c6139 2160 dst_cache, info);
e5d4b29f 2161 if (IS_ERR(ndst)) {
8ebd115b 2162 err = PTR_ERR(ndst);
c46b7897 2163 ndst = NULL;
9dcc71e1 2164 goto tx_error;
e4c7ed41 2165 }
655c3de1 2166
fee1fad7 2167 if (!info) {
2168 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
49560532 2169
fee1fad7 2170 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2171 dst_port, vni, ndst,
2172 rt6i_flags);
2173 if (err)
56de859e 2174 goto out_unlock;
fee1fad7 2175 }
35e2d115 2176
1400615d 2177 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e4c7ed41 2178 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2179 skb_scrub_packet(skb, xnet);
2180 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
54bfd872 2181 vni, md, flags, udp_sum);
c46b7897 2182 if (err < 0)
2183 goto tx_error;
2184
0770b53b 2185 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
1158632b 2186 &local_ip.sin6.sin6_addr,
272d96a5 2187 &dst->sin6.sin6_addr, tos, ttl,
e7f70af1 2188 label, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2189#endif
2190 }
56de859e
JK		 2191out_unlock:
2192 rcu_read_unlock();
4ad16930 2193 return;
d342894c 2194
2195drop:
2196 dev->stats.tx_dropped++;
c46b7897 2197 dev_kfree_skb(skb);
2198 return;
d342894c 2199
2200tx_error:
56de859e 2201 rcu_read_unlock();
655c3de1 2202 if (err == -ELOOP)
2203 dev->stats.collisions++;
2204 else if (err == -ENETUNREACH)
2205 dev->stats.tx_carrier_errors++;
c46b7897 2206 dst_release(ndst);
d342894c 2207 dev->stats.tx_errors++;
c46b7897 2208 kfree_skb(skb);
d342894c 2209}
2210
6681712d
DS		 2211/* Transmit local packets over Vxlan
2212 *
2213 * Outer IP header inherits ECN and DF from inner header.
2214 * Outer UDP destination is the VXLAN assigned port.
2215 * source port is based on hash of flow
2216 */
2217static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2218{
2219 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2220 const struct ip_tunnel_info *info;
6681712d
DS		 2221 struct ethhdr *eth;
2222 bool did_rsc = false;
8f646c92 2223 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2224 struct vxlan_fdb *f;
3ad7a4b1 2225 __be32 vni = 0;
6681712d 2226
61adedf3 2227 info = skb_tunnel_info(skb);
3093fbe7 2228
6681712d 2229 skb_reset_mac_header(skb);
6681712d 2230
47e5d1b0 2231 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
3ad7a4b1
RP		 2232 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2233 info->mode & IP_TUNNEL_INFO_TX) {
2234 vni = tunnel_id_to_key32(info->key.tun_id);
2235 } else {
2236 if (info && info->mode & IP_TUNNEL_INFO_TX)
2237 vxlan_xmit_one(skb, dev, vni, NULL, false);
2238 else
2239 kfree_skb(skb);
2240 return NETDEV_TX_OK;
2241 }
47e5d1b0
JB		 2242 }
2243
2244 if (vxlan->flags & VXLAN_F_PROXY) {
2245 eth = eth_hdr(skb);
f564f45c 2246 if (ntohs(eth->h_proto) == ETH_P_ARP)
3ad7a4b1 2247 return arp_reduce(dev, skb, vni);
f564f45c 2248#if IS_ENABLED(CONFIG_IPV6)
f1fb08f6
VB		 2249 else if (ntohs(eth->h_proto) == ETH_P_IPV6) {
2250 struct ipv6hdr *hdr, _hdr;
2251 if ((hdr = skb_header_pointer(skb,
2252 skb_network_offset(skb),
2253 sizeof(_hdr), &_hdr)) &&
2254 hdr->nexthdr == IPPROTO_ICMPV6)
2255 return neigh_reduce(dev, skb, vni);
f564f45c
CW		 2256 }
2257#endif
2258 }
6681712d 2259
47e5d1b0 2260 eth = eth_hdr(skb);
3ad7a4b1 2261 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2262 did_rsc = false;
2263
2264 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2265 (ntohs(eth->h_proto) == ETH_P_IP ||
2266 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2267 did_rsc = route_shortcircuit(dev, skb);
2268 if (did_rsc)
3ad7a4b1 2269 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
ae884082
DS		 2270 }
2271
6681712d 2272 if (f == NULL) {
3ad7a4b1 2273 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae
MR		 2274 if (f == NULL) {
2275 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2276 !is_multicast_ether_addr(eth->h_dest))
2277 vxlan_fdb_miss(vxlan, eth->h_dest);
2278
2279 dev->stats.tx_dropped++;
8f646c92 2280 kfree_skb(skb);
afbd8bae
MR		 2281 return NETDEV_TX_OK;
2282 }
2283 }
6681712d 2284
afbd8bae
MR		 2285 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2286 struct sk_buff *skb1;
6681712d 2287
8f646c92
ED		 2288 if (!fdst) {
2289 fdst = rdst;
2290 continue;
2291 }
afbd8bae
MR		 2292 skb1 = skb_clone(skb, GFP_ATOMIC);
2293 if (skb1)
3ad7a4b1 2294 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
6681712d
DS		 2295 }
2296
8f646c92 2297 if (fdst)
3ad7a4b1 2298 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
8f646c92
ED		 2299 else
2300 kfree_skb(skb);
4ad16930 2301 return NETDEV_TX_OK;
6681712d
DS		 2302}
2303
d342894c 2304/* Walk the forwarding table and purge stale entries */
2305static void vxlan_cleanup(unsigned long arg)
2306{
2307 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2308 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2309 unsigned int h;
2310
2311 if (!netif_running(vxlan->dev))
2312 return;
2313
d342894c 2314 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2315 struct hlist_node *p, *n;
14e1d0fa
SD		 2316
2317 spin_lock_bh(&vxlan->hash_lock);
d342894c 2318 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2319 struct vxlan_fdb *f
2320 = container_of(p, struct vxlan_fdb, hlist);
2321 unsigned long timeout;
2322
efb5f68f 2323 if (f->state & (NUD_PERMANENT | NUD_NOARP))
d342894c 2324 continue;
2325
def499c9
RP		 2326 if (f->flags & NTF_EXT_LEARNED)
2327 continue;
2328
0dfbdf41 2329 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2330 if (time_before_eq(timeout, jiffies)) {
2331 netdev_dbg(vxlan->dev,
2332 "garbage collect %pM\n",
2333 f->eth_addr);
2334 f->state = NUD_STALE;
2335 vxlan_fdb_destroy(vxlan, f);
2336 } else if (time_before(timeout, next_timer))
2337 next_timer = timeout;
2338 }
14e1d0fa 2339 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2340 }
d342894c 2341
2342 mod_timer(&vxlan->age_timer, next_timer);
2343}
2344
9c2e24e1
PS		 2345static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2346{
56ef9c90 2347 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
54bfd872 2348 __be32 vni = vxlan->default_dst.remote_vni;
9c2e24e1 2349
56ef9c90 2350 spin_lock(&vn->sock_lock);
9c2e24e1 2351 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2352 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2353}
2354
d342894c 2355/* Setup stats when device is created */
2356static int vxlan_init(struct net_device *dev)
2357{
1c213bd2 2358 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2359 if (!dev->tstats)
d342894c 2360 return -ENOMEM;
2361
2362 return 0;
2363}
2364
3ad7a4b1 2365static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
afbd8bae
MR		 2366{
2367 struct vxlan_fdb *f;
2368
2369 spin_lock_bh(&vxlan->hash_lock);
3ad7a4b1 2370 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
afbd8bae
MR		 2371 if (f)
2372 vxlan_fdb_destroy(vxlan, f);
2373 spin_unlock_bh(&vxlan->hash_lock);
2374}
2375
ebf4063e
SH		 2376static void vxlan_uninit(struct net_device *dev)
2377{
2378 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2379
3ad7a4b1 2380 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
afbd8bae 2381
ebf4063e
SH		 2382 free_percpu(dev->tstats);
2383}
2384
d342894c 2385/* Start ageing timer and join group when device is brought up */
2386static int vxlan_open(struct net_device *dev)
2387{
2388 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2389 int ret;
56ef9c90 2390
205f356d
JB		 2391 ret = vxlan_sock_add(vxlan);
2392 if (ret < 0)
2393 return ret;
d342894c 2394
79d4a94f 2395 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2396 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2397 if (ret == -EADDRINUSE)
2398 ret = 0;
56ef9c90 2399 if (ret) {
205f356d 2400 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2401 return ret;
2402 }
d342894c 2403 }
2404
0dfbdf41 2405 if (vxlan->cfg.age_interval)
d342894c 2406 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2407
56ef9c90 2408 return ret;
d342894c 2409}
2410
2411/* Purge the forwarding table */
8b3f9337 2412static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
d342894c 2413{
31fec5aa 2414 unsigned int h;
d342894c 2415
2416 spin_lock_bh(&vxlan->hash_lock);
2417 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2418 struct hlist_node *p, *n;
2419 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2420 struct vxlan_fdb *f
2421 = container_of(p, struct vxlan_fdb, hlist);
8b3f9337
RP		 2422 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2423 continue;
afbd8bae
MR		 2424 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2425 if (!is_zero_ether_addr(f->eth_addr))
2426 vxlan_fdb_destroy(vxlan, f);
d342894c 2427 }
2428 }
2429 spin_unlock_bh(&vxlan->hash_lock);
2430}
2431
2432/* Cleanup timer and forwarding table on shutdown */
2433static int vxlan_stop(struct net_device *dev)
2434{
2435 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2436 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2437 int ret = 0;
d342894c 2438
24c0e683 2439 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2440 !vxlan_group_used(vn, vxlan))
56ef9c90 2441 ret = vxlan_igmp_leave(vxlan);
d342894c 2442
2443 del_timer_sync(&vxlan->age_timer);
2444
8b3f9337 2445 vxlan_flush(vxlan, false);
205f356d 2446 vxlan_sock_release(vxlan);
d342894c 2447
56ef9c90 2448 return ret;
d342894c 2449}
2450
d342894c 2451/* Stub, nothing needs to be done. */
2452static void vxlan_set_multicast_list(struct net_device *dev)
2453{
2454}
2455
91572088 2456static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
345010b5 2457{
91572088
JW		 2458 struct vxlan_dev *vxlan = netdev_priv(dev);
2459 struct vxlan_rdst *dst = &vxlan->default_dst;
2460 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2461 dst->remote_ifindex);
2462 bool use_ipv6 = false;
345010b5
DB		 2463
2464 if (dst->remote_ip.sa.sa_family == AF_INET6)
91572088 2465 use_ipv6 = true;
345010b5 2466
91572088
JW		 2467 /* This check is different than dev->max_mtu, because it looks at
2468 * the lowerdev->mtu, rather than the static dev->max_mtu
2469 */
2470 if (lowerdev) {
2471 int max_mtu = lowerdev->mtu -
2472 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2473 if (new_mtu > max_mtu)
72564b59 2474 return -EINVAL;
72564b59
DW		 2475 }
2476
345010b5
DB		 2477 dev->mtu = new_mtu;
2478 return 0;
2479}
2480
fc4099f1
PS		 2481static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2482{
2483 struct vxlan_dev *vxlan = netdev_priv(dev);
2484 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2485 __be16 sport, dport;
2486
2487 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2488 vxlan->cfg.port_max, true);
2489 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2490
239e944f 2491 if (ip_tunnel_info_af(info) == AF_INET) {
c6fcc4fc 2492 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
1a8496ba
JB		 2493 struct rtable *rt;
2494
655c3de1 2495 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
1a8496ba 2496 info->key.u.ipv4.dst,
22f0708a
PA		 2497 &info->key.u.ipv4.src, dport, sport,
2498 &info->dst_cache, info);
1a8496ba
JB		 2499 if (IS_ERR(rt))
2500 return PTR_ERR(rt);
2501 ip_rt_put(rt);
239e944f
JB		 2502 } else {
2503#if IS_ENABLED(CONFIG_IPV6)
03dc52a8 2504 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
239e944f
JB		 2505 struct dst_entry *ndst;
2506
655c3de1 2507 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
e7f70af1 2508 info->key.label, &info->key.u.ipv6.dst,
22f0708a
PA		 2509 &info->key.u.ipv6.src, dport, sport,
2510 &info->dst_cache, info);
239e944f
JB		 2511 if (IS_ERR(ndst))
2512 return PTR_ERR(ndst);
2513 dst_release(ndst);
239e944f
JB		 2514#else /* !CONFIG_IPV6 */
2515 return -EPFNOSUPPORT;
2516#endif
2517 }
1a8496ba
JB		 2518 info->key.tp_src = sport;
2519 info->key.tp_dst = dport;
239e944f 2520 return 0;
fc4099f1
PS		 2521}
2522
0c867c9b 2523static const struct net_device_ops vxlan_netdev_ether_ops = {
d342894c 2524 .ndo_init = vxlan_init,
ebf4063e 2525 .ndo_uninit = vxlan_uninit,
d342894c 2526 .ndo_open = vxlan_open,
2527 .ndo_stop = vxlan_stop,
2528 .ndo_start_xmit = vxlan_xmit,
e8171045 2529 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2530 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2531 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2532 .ndo_validate_addr = eth_validate_addr,
2533 .ndo_set_mac_address = eth_mac_addr,
2534 .ndo_fdb_add = vxlan_fdb_add,
2535 .ndo_fdb_del = vxlan_fdb_delete,
2536 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2537 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2538};
2539
e1e5314d
JB		 2540static const struct net_device_ops vxlan_netdev_raw_ops = {
2541 .ndo_init = vxlan_init,
2542 .ndo_uninit = vxlan_uninit,
2543 .ndo_open = vxlan_open,
2544 .ndo_stop = vxlan_stop,
2545 .ndo_start_xmit = vxlan_xmit,
2546 .ndo_get_stats64 = ip_tunnel_get_stats64,
2547 .ndo_change_mtu = vxlan_change_mtu,
2548 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2549};
2550
d342894c 2551/* Info for udev, that this is a virtual tunnel endpoint */
2552static struct device_type vxlan_type = {
2553 .name = "vxlan",
2554};
2555
e5de25dc 2556/* Calls the ndo_udp_tunnel_add of the caller in order to
35e42379 2557 * supply the listening VXLAN udp ports. Callers are expected
e5de25dc 2558 * to implement the ndo_udp_tunnel_add.
53cf5275 2559 */
b7aade15 2560static void vxlan_push_rx_ports(struct net_device *dev)
53cf5275
JG		 2561{
2562 struct vxlan_sock *vs;
2563 struct net *net = dev_net(dev);
2564 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
35e42379 2565 unsigned int i;
53cf5275
JG		 2566
2567 spin_lock(&vn->sock_lock);
2568 for (i = 0; i < PORT_HASH_SIZE; ++i) {
e7b3db5e
AD		 2569 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist)
2570 udp_tunnel_push_rx_port(dev, vs->sock,
b9adcd69
AD		 2571 (vs->flags & VXLAN_F_GPE) ?
2572 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 2573 UDP_TUNNEL_TYPE_VXLAN);
53cf5275
JG		 2574 }
2575 spin_unlock(&vn->sock_lock);
2576}
53cf5275 2577
d342894c 2578/* Initialize the device structure. */
2579static void vxlan_setup(struct net_device *dev)
2580{
2581 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2582 unsigned int h;
d342894c 2583
65226ef8
JB		 2584 eth_hw_addr_random(dev);
2585 ether_setup(dev);
2586
ebf4063e 2587 dev->destructor = free_netdev;
d342894c 2588 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2589
d342894c 2590 dev->features |= NETIF_F_LLTX;
d6727fe3 2591 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2592 dev->features |= NETIF_F_RXCSUM;
05c0db08 2593 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2594
1eaa8178 2595 dev->vlan_features = dev->features;
0afb1666 2596 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2597 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
02875878 2598 netif_keep_dst(dev);
0c867c9b 2599 dev->priv_flags |= IFF_NO_QUEUE;
d342894c 2600
553675fb 2601 INIT_LIST_HEAD(&vxlan->next);
d342894c 2602 spin_lock_init(&vxlan->hash_lock);
2603
2604 init_timer_deferrable(&vxlan->age_timer);
2605 vxlan->age_timer.function = vxlan_cleanup;
2606 vxlan->age_timer.data = (unsigned long) vxlan;
2607
0dfbdf41 2608 vxlan->cfg.dst_port = htons(vxlan_port);
05f47d69 2609
d342894c 2610 vxlan->dev = dev;
2611
58ce31cc
TH		 2612 gro_cells_init(&vxlan->gro_cells, dev);
2613
d342894c 2614 for (h = 0; h < FDB_HASH_SIZE; ++h)
2615 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2616}
2617
0c867c9b
JB		 2618static void vxlan_ether_setup(struct net_device *dev)
2619{
0c867c9b
JB		 2620 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2621 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2622 dev->netdev_ops = &vxlan_netdev_ether_ops;
2623}
2624
e1e5314d
JB		 2625static void vxlan_raw_setup(struct net_device *dev)
2626{
65226ef8 2627 dev->header_ops = NULL;
e1e5314d
JB		 2628 dev->type = ARPHRD_NONE;
2629 dev->hard_header_len = 0;
2630 dev->addr_len = 0;
e1e5314d
JB		 2631 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2632 dev->netdev_ops = &vxlan_netdev_raw_ops;
2633}
2634
d342894c 2635static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2636 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2637 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2638 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2639 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2640 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2641 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2642 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2643 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
e7f70af1 2644 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
d342894c 2645 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2646 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2647 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2648 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2649 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2650 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2651 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2652 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2653 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2654 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2655 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2656 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2657 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2658 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2659 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2660 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
e1e5314d 2661 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
0ace2ca8 2662 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2663};
2664
2665static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2666{
2667 if (tb[IFLA_ADDRESS]) {
2668 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2669 pr_debug("invalid link address (not ethernet)\n");
2670 return -EINVAL;
2671 }
2672
2673 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2674 pr_debug("invalid all zero ethernet address\n");
2675 return -EADDRNOTAVAIL;
2676 }
2677 }
2678
2679 if (!data)
2680 return -EINVAL;
2681
2682 if (data[IFLA_VXLAN_ID]) {
2683 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
4e37d691 2684 if (id >= VXLAN_N_VID)
d342894c 2685 return -ERANGE;
2686 }
2687
05f47d69 2688 if (data[IFLA_VXLAN_PORT_RANGE]) {
2689 const struct ifla_vxlan_port_range *p
2690 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2691
2692 if (ntohs(p->high) < ntohs(p->low)) {
2693 pr_debug("port range %u .. %u not valid\n",
2694 ntohs(p->low), ntohs(p->high));
2695 return -EINVAL;
2696 }
2697 }
2698
d342894c 2699 return 0;
2700}
2701
1b13c97f
YB		 2702static void vxlan_get_drvinfo(struct net_device *netdev,
2703 struct ethtool_drvinfo *drvinfo)
2704{
2705 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2706 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2707}
2708
2709static const struct ethtool_ops vxlan_ethtool_ops = {
2710 .get_drvinfo = vxlan_get_drvinfo,
2711 .get_link = ethtool_op_get_link,
2712};
2713
3ee64f39
TH		 2714static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2715 __be16 port, u32 flags)
553675fb 2716{
e4c7ed41 2717 struct socket *sock;
3ee64f39
TH		 2718 struct udp_port_cfg udp_conf;
2719 int err;
e4c7ed41 2720
3ee64f39 2721 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2722
3ee64f39
TH		 2723 if (ipv6) {
2724 udp_conf.family = AF_INET6;
3ee64f39 2725 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2726 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2727 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2728 } else {
2729 udp_conf.family = AF_INET;
e4c7ed41 2730 }
e4c7ed41 2731
3ee64f39 2732 udp_conf.local_udp_port = port;
553675fb 2733
3ee64f39
TH		 2734 /* Open UDP socket */
2735 err = udp_sock_create(net, &udp_conf, &sock);
2736 if (err < 0)
2737 return ERR_PTR(err);
e4c7ed41 2738
39deb2c7 2739 return sock;
e4c7ed41
CW		 2740}
2741
2742/* Create new listen socket if needed */
b1be00a6
JB		 2743static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2744 __be16 port, u32 flags)
e4c7ed41
CW		 2745{
2746 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2747 struct vxlan_sock *vs;
2748 struct socket *sock;
e4c7ed41 2749 unsigned int h;
acbf74a7 2750 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2751
dc01e7d3 2752 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2753 if (!vs)
2754 return ERR_PTR(-ENOMEM);
2755
2756 for (h = 0; h < VNI_HASH_SIZE; ++h)
2757 INIT_HLIST_HEAD(&vs->vni_list[h]);
2758
3ee64f39 2759 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2760 if (IS_ERR(sock)) {
553675fb 2761 kfree(vs);
e50fddc8 2762 return ERR_CAST(sock);
553675fb 2763 }
e4c7ed41
CW		 2764
2765 vs->sock = sock;
9c2e24e1 2766 atomic_set(&vs->refcnt, 1);
af33c1ad 2767 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2768
9c2e24e1
PS		 2769 spin_lock(&vn->sock_lock);
2770 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
e7b3db5e 2771 udp_tunnel_notify_add_rx_port(sock,
b9adcd69
AD		 2772 (vs->flags & VXLAN_F_GPE) ?
2773 UDP_TUNNEL_TYPE_VXLAN_GPE :
e7b3db5e 2774 UDP_TUNNEL_TYPE_VXLAN);
9c2e24e1 2775 spin_unlock(&vn->sock_lock);
553675fb 2776
2777 /* Mark socket as an encapsulation socket. */
5602c48c 2778 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
acbf74a7
AZ		 2779 tunnel_cfg.sk_user_data = vs;
2780 tunnel_cfg.encap_type = 1;
f2d1968e 2781 tunnel_cfg.encap_rcv = vxlan_rcv;
acbf74a7 2782 tunnel_cfg.encap_destroy = NULL;
5602c48c
TH		 2783 tunnel_cfg.gro_receive = vxlan_gro_receive;
2784 tunnel_cfg.gro_complete = vxlan_gro_complete;
acbf74a7
AZ		 2785
2786 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2787
9c2e24e1
PS		 2788 return vs;
2789}
2790
b1be00a6 2791static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2792{
205f356d
JB		 2793 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2794 struct vxlan_sock *vs = NULL;
9c2e24e1 2795
205f356d 2796 if (!vxlan->cfg.no_share) {
56ef9c90 2797 spin_lock(&vn->sock_lock);
205f356d
JB		 2798 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2799 vxlan->cfg.dst_port, vxlan->flags);
2800 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
56ef9c90 2801 spin_unlock(&vn->sock_lock);
205f356d 2802 return -EBUSY;
56ef9c90
MRL		 2803 }
2804 spin_unlock(&vn->sock_lock);
2805 }
205f356d 2806 if (!vs)
b1be00a6
JB		 2807 vs = vxlan_socket_create(vxlan->net, ipv6,
2808 vxlan->cfg.dst_port, vxlan->flags);
205f356d
JB		 2809 if (IS_ERR(vs))
2810 return PTR_ERR(vs);
b1be00a6
JB		 2811#if IS_ENABLED(CONFIG_IPV6)
2812 if (ipv6)
c6fcc4fc 2813 rcu_assign_pointer(vxlan->vn6_sock, vs);
b1be00a6
JB		 2814 else
2815#endif
c6fcc4fc 2816 rcu_assign_pointer(vxlan->vn4_sock, vs);
205f356d
JB		 2817 vxlan_vs_add_dev(vs, vxlan);
2818 return 0;
553675fb 2819}
2820
b1be00a6
JB		 2821static int vxlan_sock_add(struct vxlan_dev *vxlan)
2822{
b1be00a6 2823 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
d074bf96
JB		 2824 bool ipv6 = vxlan->flags & VXLAN_F_IPV6 || metadata;
2825 bool ipv4 = !ipv6 || metadata;
b1be00a6
JB		 2826 int ret = 0;
2827
c6fcc4fc 2828 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
b1be00a6 2829#if IS_ENABLED(CONFIG_IPV6)
c6fcc4fc 2830 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
d074bf96 2831 if (ipv6) {
b1be00a6 2832 ret = __vxlan_sock_add(vxlan, true);
d074bf96
JB		 2833 if (ret < 0 && ret != -EAFNOSUPPORT)
2834 ipv4 = false;
2835 }
b1be00a6 2836#endif
d074bf96 2837 if (ipv4)
b1be00a6
JB		 2838 ret = __vxlan_sock_add(vxlan, false);
2839 if (ret < 0)
2840 vxlan_sock_release(vxlan);
2841 return ret;
2842}
2843
0dfbdf41 2844static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
8bcdc4f3
RP		 2845 struct vxlan_config *conf,
2846 bool changelink)
d342894c 2847{
33564bbb 2848 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
07b9b37c 2849 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
c7995c43 2850 struct vxlan_rdst *dst = &vxlan->default_dst;
b1be00a6 2851 unsigned short needed_headroom = ETH_HLEN;
e4c7ed41 2852 bool use_ipv6 = false;
0dfbdf41 2853 __be16 default_port = vxlan->cfg.dst_port;
7e059158 2854 struct net_device *lowerdev = NULL;
d342894c 2855
8bcdc4f3
RP		 2856 if (!changelink) {
2857 if (conf->flags & VXLAN_F_GPE) {
2858 /* For now, allow GPE only together with
2859 * COLLECT_METADATA. This can be relaxed later; in such
2860 * case, the other side of the PtP link will have to be
2861 * provided.
2862 */
2863 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
2864 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
2865 pr_info("unsupported combination of extensions\n");
2866 return -EINVAL;
2867 }
2868 vxlan_raw_setup(dev);
2869 } else {
2870 vxlan_ether_setup(dev);
3555621d 2871 }
e1e5314d 2872
8bcdc4f3
RP		 2873 /* MTU range: 68 - 65535 */
2874 dev->min_mtu = ETH_MIN_MTU;
2875 dev->max_mtu = ETH_MAX_MTU;
2876 vxlan->net = src_net;
e1e5314d 2877 }
0c867c9b 2878
0dfbdf41 2879 dst->remote_vni = conf->vni;
d342894c 2880
0dfbdf41 2881 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
d342894c 2882
0dfbdf41
TG		 2883 /* Unless IPv6 is explicitly requested, assume IPv4 */
2884 if (!dst->remote_ip.sa.sa_family)
2885 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41 2886
0dfbdf41 2887 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
057ba29b
JB		 2888 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2889 if (!IS_ENABLED(CONFIG_IPV6))
2890 return -EPFNOSUPPORT;
e4c7ed41 2891 use_ipv6 = true;
b1be00a6 2892 vxlan->flags |= VXLAN_F_IPV6;
057ba29b 2893 }
d342894c 2894
e7f70af1
DB		 2895 if (conf->label && !use_ipv6) {
2896 pr_info("label only supported in use with IPv6\n");
2897 return -EINVAL;
2898 }
2899
8bcdc4f3
RP		 2900 if (conf->remote_ifindex &&
2901 conf->remote_ifindex != vxlan->cfg.remote_ifindex) {
7e059158 2902 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
0dfbdf41 2903 dst->remote_ifindex = conf->remote_ifindex;
34e02aa1 2904
2905 if (!lowerdev) {
8bcdc4f3
RP		 2906 pr_info("ifindex %d does not exist\n",
2907 dst->remote_ifindex);
34e02aa1 2908 return -ENODEV;
2909 }
d342894c 2910
e4c7ed41
CW		 2911#if IS_ENABLED(CONFIG_IPV6)
2912 if (use_ipv6) {
2913 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2914 if (idev && idev->cnf.disable_ipv6) {
2915 pr_info("IPv6 is disabled via sysctl\n");
2916 return -EPERM;
2917 }
e4c7ed41
CW		 2918 }
2919#endif
2920
0dfbdf41 2921 if (!conf->mtu)
91572088
JW		 2922 dev->mtu = lowerdev->mtu -
2923 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2924
b1be00a6 2925 needed_headroom = lowerdev->hard_header_len;
8bcdc4f3
RP		 2926 } else if (!conf->remote_ifindex &&
2927 vxlan_addr_multicast(&dst->remote_ip)) {
9b4cdd51
JB		 2928 pr_info("multicast destination requires interface to be specified\n");
2929 return -EINVAL;
9dc2ad10 2930 }
d342894c 2931
d6acfeb1
FM		 2932 if (lowerdev) {
2933 dev->gso_max_size = lowerdev->gso_max_size;
2934 dev->gso_max_segs = lowerdev->gso_max_segs;
2935 }
2936
7e059158 2937 if (conf->mtu) {
91572088
JW		 2938 int max_mtu = ETH_MAX_MTU;
2939
2940 if (lowerdev)
2941 max_mtu = lowerdev->mtu;
2942
2943 max_mtu -= (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2944
2945 if (conf->mtu < dev->min_mtu || conf->mtu > dev->max_mtu)
2946 return -EINVAL;
2947
2948 dev->mtu = conf->mtu;
2949
2950 if (conf->mtu > max_mtu)
2951 dev->mtu = max_mtu;
7e059158
DW		 2952 }
2953
b1be00a6
JB		 2954 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2955 needed_headroom += VXLAN6_HEADROOM;
2956 else
2957 needed_headroom += VXLAN_HEADROOM;
2958 dev->needed_headroom = needed_headroom;
2959
0dfbdf41 2960 memcpy(&vxlan->cfg, conf, sizeof(*conf));
e1e5314d
JB		 2961 if (!vxlan->cfg.dst_port) {
2962 if (conf->flags & VXLAN_F_GPE)
d5ff72d9 2963 vxlan->cfg.dst_port = htons(4790); /* IANA VXLAN-GPE port */
e1e5314d
JB		 2964 else
2965 vxlan->cfg.dst_port = default_port;
2966 }
0dfbdf41
TG		 2967 vxlan->flags |= conf->flags;
2968
2969 if (!vxlan->cfg.age_interval)
2970 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2971
8bcdc4f3
RP		 2972 if (changelink)
2973 return 0;
2974
07b9b37c
ND		 2975 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2976 if (tmp->cfg.vni == conf->vni &&
2977 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2978 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2979 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2980 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
3555621d
JB		 2981 (vxlan->flags & VXLAN_F_RCV_FLAGS)) {
2982 pr_info("duplicate VNI %u\n", be32_to_cpu(conf->vni));
2983 return -EEXIST;
2984 }
07b9b37c 2985 }
0dfbdf41 2986
0dfbdf41
TG		 2987 return 0;
2988}
2989
c80498e3
ND		 2990static int __vxlan_dev_create(struct net *net, struct net_device *dev,
2991 struct vxlan_config *conf)
2992{
2993 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2994 struct vxlan_dev *vxlan = netdev_priv(dev);
2995 int err;
2996
2997 err = vxlan_dev_configure(net, dev, conf, false);
2998 if (err)
2999 return err;
3000
3001 dev->ethtool_ops = &vxlan_ethtool_ops;
3002
3003 /* create an fdb entry for a valid default destination */
3004 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
3005 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3006 &vxlan->default_dst.remote_ip,
3007 NUD_REACHABLE | NUD_PERMANENT,
3008 NLM_F_EXCL | NLM_F_CREATE,
3009 vxlan->cfg.dst_port,
3010 vxlan->default_dst.remote_vni,
3011 vxlan->default_dst.remote_vni,
3012 vxlan->default_dst.remote_ifindex,
3013 NTF_SELF);
3014 if (err)
3015 return err;
3016 }
3017
3018 err = register_netdevice(dev);
3019 if (err) {
3020 vxlan_fdb_delete_default(vxlan, vxlan->default_dst.remote_vni);
3021 return err;
3022 }
3023
3024 list_add(&vxlan->next, &vn->vxlan_list);
3025 return 0;
3026}
3027
8bcdc4f3
RP		 3028static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3029 struct net_device *dev, struct vxlan_config *conf,
3030 bool changelink)
0dfbdf41 3031{
8bcdc4f3 3032 struct vxlan_dev *vxlan = netdev_priv(dev);
0dfbdf41 3033
8bcdc4f3 3034 memset(conf, 0, sizeof(*conf));
e277de5f 3035
8bcdc4f3
RP		 3036 /* if changelink operation, start with old existing cfg */
3037 if (changelink)
3038 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3039
3040 if (data[IFLA_VXLAN_ID]) {
3041 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3042
3043 if (changelink && (vni != conf->vni))
3044 return -EOPNOTSUPP;
3045 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3046 }
0dfbdf41
TG		 3047
3048 if (data[IFLA_VXLAN_GROUP]) {
8bcdc4f3 3049 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
0dfbdf41
TG		 3050 } else if (data[IFLA_VXLAN_GROUP6]) {
3051 if (!IS_ENABLED(CONFIG_IPV6))
3052 return -EPFNOSUPPORT;
3053
8bcdc4f3
RP		 3054 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3055 conf->remote_ip.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3056 }
3057
3058 if (data[IFLA_VXLAN_LOCAL]) {
8bcdc4f3
RP		 3059 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3060 conf->saddr.sa.sa_family = AF_INET;
0dfbdf41
TG		 3061 } else if (data[IFLA_VXLAN_LOCAL6]) {
3062 if (!IS_ENABLED(CONFIG_IPV6))
3063 return -EPFNOSUPPORT;
3064
3065 /* TODO: respect scope id */
8bcdc4f3
RP		 3066 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3067 conf->saddr.sa.sa_family = AF_INET6;
0dfbdf41
TG		 3068 }
3069
3070 if (data[IFLA_VXLAN_LINK])
8bcdc4f3 3071 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
0dfbdf41 3072
d342894c 3073 if (data[IFLA_VXLAN_TOS])
8bcdc4f3 3074 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 3075
afb97186 3076 if (data[IFLA_VXLAN_TTL])
8bcdc4f3 3077 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 3078
e7f70af1 3079 if (data[IFLA_VXLAN_LABEL])
8bcdc4f3 3080 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
e7f70af1
DB		 3081 IPV6_FLOWLABEL_MASK;
3082
8bcdc4f3
RP		 3083 if (data[IFLA_VXLAN_LEARNING]) {
3084 if (nla_get_u8(data[IFLA_VXLAN_LEARNING])) {
3085 conf->flags |= VXLAN_F_LEARN;
3086 } else {
3087 conf->flags &= ~VXLAN_F_LEARN;
3088 vxlan->flags &= ~VXLAN_F_LEARN;
3089 }
3090 } else if (!changelink) {
3091 /* default to learn on a new device */
3092 conf->flags |= VXLAN_F_LEARN;
3093 }
d342894c 3094
8bcdc4f3
RP		 3095 if (data[IFLA_VXLAN_AGEING]) {
3096 if (changelink)
3097 return -EOPNOTSUPP;
3098 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3099 }
d342894c 3100
8bcdc4f3
RP		 3101 if (data[IFLA_VXLAN_PROXY]) {
3102 if (changelink)
3103 return -EOPNOTSUPP;
3104 if (nla_get_u8(data[IFLA_VXLAN_PROXY]))
3105 conf->flags |= VXLAN_F_PROXY;
3106 }
e4f67add 3107
8bcdc4f3
RP		 3108 if (data[IFLA_VXLAN_RSC]) {
3109 if (changelink)
3110 return -EOPNOTSUPP;
3111 if (nla_get_u8(data[IFLA_VXLAN_RSC]))
3112 conf->flags |= VXLAN_F_RSC;
3113 }
e4f67add 3114
8bcdc4f3
RP		 3115 if (data[IFLA_VXLAN_L2MISS]) {
3116 if (changelink)
3117 return -EOPNOTSUPP;
3118 if (nla_get_u8(data[IFLA_VXLAN_L2MISS]))
3119 conf->flags |= VXLAN_F_L2MISS;
3120 }
e4f67add 3121
8bcdc4f3
RP		 3122 if (data[IFLA_VXLAN_L3MISS]) {
3123 if (changelink)
3124 return -EOPNOTSUPP;
3125 if (nla_get_u8(data[IFLA_VXLAN_L3MISS]))
3126 conf->flags |= VXLAN_F_L3MISS;
3127 }
e4f67add 3128
8bcdc4f3
RP		 3129 if (data[IFLA_VXLAN_LIMIT]) {
3130 if (changelink)
3131 return -EOPNOTSUPP;
3132 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3133 }
d342894c 3134
8bcdc4f3
RP		 3135 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3136 if (changelink)
3137 return -EOPNOTSUPP;
3138 if (nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3139 conf->flags |= VXLAN_F_COLLECT_METADATA;
3140 }
f8a9b1bc 3141
05f47d69 3142 if (data[IFLA_VXLAN_PORT_RANGE]) {
8bcdc4f3
RP		 3143 if (!changelink) {
3144 const struct ifla_vxlan_port_range *p
3145 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3146 conf->port_min = ntohs(p->low);
3147 conf->port_max = ntohs(p->high);
3148 } else {
3149 return -EOPNOTSUPP;
3150 }
05f47d69 3151 }
3152
8bcdc4f3
RP		 3153 if (data[IFLA_VXLAN_PORT]) {
3154 if (changelink)
3155 return -EOPNOTSUPP;
3156 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3157 }
823aa873 3158
8bcdc4f3
RP		 3159 if (data[IFLA_VXLAN_UDP_CSUM]) {
3160 if (changelink)
3161 return -EOPNOTSUPP;
3162 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3163 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3164 }
359a0ea9 3165
8bcdc4f3
RP		 3166 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3167 if (changelink)
3168 return -EOPNOTSUPP;
3169 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
3170 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
3171 }
359a0ea9 3172
8bcdc4f3
RP		 3173 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3174 if (changelink)
3175 return -EOPNOTSUPP;
3176 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
3177 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
3178 }
359a0ea9 3179
8bcdc4f3
RP		 3180 if (data[IFLA_VXLAN_REMCSUM_TX]) {
3181 if (changelink)
3182 return -EOPNOTSUPP;
3183 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
3184 conf->flags |= VXLAN_F_REMCSUM_TX;
3185 }
dfd8645e 3186
8bcdc4f3
RP		 3187 if (data[IFLA_VXLAN_REMCSUM_RX]) {
3188 if (changelink)
3189 return -EOPNOTSUPP;
3190 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3191 conf->flags |= VXLAN_F_REMCSUM_RX;
3192 }
3193
3194 if (data[IFLA_VXLAN_GBP]) {
3195 if (changelink)
3196 return -EOPNOTSUPP;
3197 conf->flags |= VXLAN_F_GBP;
3198 }
3199
3200 if (data[IFLA_VXLAN_GPE]) {
3201 if (changelink)
3202 return -EOPNOTSUPP;
3203 conf->flags |= VXLAN_F_GPE;
3204 }
3205
3206 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3207 if (changelink)
3208 return -EOPNOTSUPP;
3209 conf->flags |= VXLAN_F_REMCSUM_NOPARTIAL;
3210 }
3211
3212 if (tb[IFLA_MTU]) {
3213 if (changelink)
3214 return -EOPNOTSUPP;
3215 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3216 }
3217
3218 return 0;
3219}
3220
3221static int vxlan_newlink(struct net *src_net, struct net_device *dev,
3222 struct nlattr *tb[], struct nlattr *data[])
3223{
8bcdc4f3
RP		 3224 struct vxlan_config conf;
3225 int err;
3226
3227 err = vxlan_nl2conf(tb, data, dev, &conf, false);
3228 if (err)
3229 return err;
3230
c80498e3 3231 return __vxlan_dev_create(src_net, dev, &conf);
8bcdc4f3 3232}
dfd8645e 3233
8bcdc4f3
RP		 3234static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
3235 struct nlattr *data[])
3236{
3237 struct vxlan_dev *vxlan = netdev_priv(dev);
3238 struct vxlan_rdst *dst = &vxlan->default_dst;
3239 struct vxlan_rdst old_dst;
3240 struct vxlan_config conf;
3241 int err;
3242
3243 err = vxlan_nl2conf(tb, data,
3244 dev, &conf, true);
3245 if (err)
3246 return err;
3511494c 3247
8bcdc4f3 3248 memcpy(&old_dst, dst, sizeof(struct vxlan_rdst));
e1e5314d 3249
8bcdc4f3
RP		 3250 err = vxlan_dev_configure(vxlan->net, dev, &conf, true);
3251 if (err)
3252 return err;
0ace2ca8 3253
8bcdc4f3
RP		 3254 /* handle default dst entry */
3255 if (!vxlan_addr_equal(&dst->remote_ip, &old_dst.remote_ip)) {
3256 spin_lock_bh(&vxlan->hash_lock);
3257 if (!vxlan_addr_any(&old_dst.remote_ip))
3258 __vxlan_fdb_delete(vxlan, all_zeros_mac,
3259 old_dst.remote_ip,
3260 vxlan->cfg.dst_port,
3261 old_dst.remote_vni,
3262 old_dst.remote_vni,
3263 old_dst.remote_ifindex, 0);
3264
3265 if (!vxlan_addr_any(&dst->remote_ip)) {
3266 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3267 &dst->remote_ip,
3268 NUD_REACHABLE | NUD_PERMANENT,
3269 NLM_F_CREATE | NLM_F_APPEND,
3270 vxlan->cfg.dst_port,
3271 dst->remote_vni,
3272 dst->remote_vni,
3273 dst->remote_ifindex,
3274 NTF_SELF);
3275 if (err) {
3276 spin_unlock_bh(&vxlan->hash_lock);
3277 return err;
3278 }
3279 }
3280 spin_unlock_bh(&vxlan->hash_lock);
3281 }
ce577668 3282
8bcdc4f3 3283 return 0;
d342894c 3284}
3285
3286static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3287{
3288 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 3289 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 3290
8b3f9337
RP		 3291 vxlan_flush(vxlan, true);
3292
fe5c3561 3293 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 3294 if (!hlist_unhashed(&vxlan->hlist))
3295 hlist_del_rcu(&vxlan->hlist);
fe5c3561 3296 spin_unlock(&vn->sock_lock);
3297
58ce31cc 3298 gro_cells_destroy(&vxlan->gro_cells);
553675fb 3299 list_del(&vxlan->next);
d342894c 3300 unregister_netdevice_queue(dev, head);
3301}
3302
3303static size_t vxlan_get_size(const struct net_device *dev)
3304{
3305
3306 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 3307 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 3308 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 3309 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 3310 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3311 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
e7f70af1 3312 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
d342894c 3313 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 3314 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3315 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3316 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3317 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3318 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3319 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3320 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3321 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3322 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3323 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3324 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3325 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3326 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3327 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3328 0;
3329}
3330
3331static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3332{
3333 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3334 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3335 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3336 .low = htons(vxlan->cfg.port_min),
3337 .high = htons(vxlan->cfg.port_max),
05f47d69 3338 };
d342894c 3339
54bfd872 3340 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
d342894c 3341 goto nla_put_failure;
3342
e4c7ed41
CW		 3343 if (!vxlan_addr_any(&dst->remote_ip)) {
3344 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3345 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3346 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3347 goto nla_put_failure;
3348#if IS_ENABLED(CONFIG_IPV6)
3349 } else {
930345ea
JB		 3350 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3351 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3352 goto nla_put_failure;
3353#endif
3354 }
3355 }
d342894c 3356
c7995c43 3357 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3358 goto nla_put_failure;
3359
0dfbdf41
TG		 3360 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3361 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3362 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3363 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3364 goto nla_put_failure;
3365#if IS_ENABLED(CONFIG_IPV6)
3366 } else {
930345ea 3367 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3368 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3369 goto nla_put_failure;
3370#endif
3371 }
3372 }
d342894c 3373
0dfbdf41
TG		 3374 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3375 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e7f70af1 3376 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
e4f67add
DS		 3377 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3378 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3379 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3380 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3381 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3382 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3383 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3384 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3385 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
da8b43c0
AS		 3386 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3387 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3388 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3389 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3390 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9 3391 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
6ceb31ca 3392 !(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
359a0ea9
TH		 3393 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3394 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3395 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 3396 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3397 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3398 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3399 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3400 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 3401 goto nla_put_failure;
3402
05f47d69 3403 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3404 goto nla_put_failure;
3405
3511494c
TG		 3406 if (vxlan->flags & VXLAN_F_GBP &&
3407 nla_put_flag(skb, IFLA_VXLAN_GBP))
3408 goto nla_put_failure;
3409
e1e5314d
JB		 3410 if (vxlan->flags & VXLAN_F_GPE &&
3411 nla_put_flag(skb, IFLA_VXLAN_GPE))
3412 goto nla_put_failure;
3413
0ace2ca8
TH		 3414 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3415 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3416 goto nla_put_failure;
3417
d342894c 3418 return 0;
3419
3420nla_put_failure:
3421 return -EMSGSIZE;
3422}
3423
1728d4fa
ND		 3424static struct net *vxlan_get_link_net(const struct net_device *dev)
3425{
3426 struct vxlan_dev *vxlan = netdev_priv(dev);
3427
3428 return vxlan->net;
3429}
3430
d342894c 3431static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3432 .kind = "vxlan",
3433 .maxtype = IFLA_VXLAN_MAX,
3434 .policy = vxlan_policy,
3435 .priv_size = sizeof(struct vxlan_dev),
3436 .setup = vxlan_setup,
3437 .validate = vxlan_validate,
3438 .newlink = vxlan_newlink,
8bcdc4f3 3439 .changelink = vxlan_changelink,
d342894c 3440 .dellink = vxlan_dellink,
3441 .get_size = vxlan_get_size,
3442 .fill_info = vxlan_fill_info,
1728d4fa 3443 .get_link_net = vxlan_get_link_net,
d342894c 3444};
3445
cf5da330
ND		 3446struct net_device *vxlan_dev_create(struct net *net, const char *name,
3447 u8 name_assign_type,
3448 struct vxlan_config *conf)
3449{
3450 struct nlattr *tb[IFLA_MAX + 1];
3451 struct net_device *dev;
3452 int err;
3453
3454 memset(&tb, 0, sizeof(tb));
3455
3456 dev = rtnl_create_link(net, name, name_assign_type,
3457 &vxlan_link_ops, tb);
3458 if (IS_ERR(dev))
3459 return dev;
3460
c80498e3 3461 err = __vxlan_dev_create(net, dev, conf);
cf5da330
ND		 3462 if (err < 0) {
3463 free_netdev(dev);
3464 return ERR_PTR(err);
3465 }
3466
3467 err = rtnl_configure_link(dev, NULL);
3468 if (err < 0) {
3469 LIST_HEAD(list_kill);
3470
3471 vxlan_dellink(dev, &list_kill);
3472 unregister_netdevice_many(&list_kill);
3473 return ERR_PTR(err);
3474 }
3475
3476 return dev;
3477}
3478EXPORT_SYMBOL_GPL(vxlan_dev_create);
3479
acaf4e70
DB		 3480static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3481 struct net_device *dev)
3482{
3483 struct vxlan_dev *vxlan, *next;
3484 LIST_HEAD(list_kill);
3485
3486 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3487 struct vxlan_rdst *dst = &vxlan->default_dst;
3488
3489 /* In case we created vxlan device with carrier
3490 * and we loose the carrier due to module unload
3491 * we also need to remove vxlan device. In other
3492 * cases, it's not necessary and remote_ifindex
3493 * is 0 here, so no matches.
3494 */
3495 if (dst->remote_ifindex == dev->ifindex)
3496 vxlan_dellink(vxlan->dev, &list_kill);
3497 }
3498
3499 unregister_netdevice_many(&list_kill);
3500}
3501
b7aade15
HFS		 3502static int vxlan_netdevice_event(struct notifier_block *unused,
3503 unsigned long event, void *ptr)
acaf4e70
DB		 3504{
3505 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3506 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3507
783c1463 3508 if (event == NETDEV_UNREGISTER)
acaf4e70 3509 vxlan_handle_lowerdev_unregister(vn, dev);
7c46a640 3510 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
b7aade15 3511 vxlan_push_rx_ports(dev);
acaf4e70
DB		 3512
3513 return NOTIFY_DONE;
3514}
3515
3516static struct notifier_block vxlan_notifier_block __read_mostly = {
b7aade15 3517 .notifier_call = vxlan_netdevice_event,
acaf4e70
DB		 3518};
3519
d342894c 3520static __net_init int vxlan_init_net(struct net *net)
3521{
3522 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3523 unsigned int h;
d342894c 3524
553675fb 3525 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3526 spin_lock_init(&vn->sock_lock);
d342894c 3527
553675fb 3528 for (h = 0; h < PORT_HASH_SIZE; ++h)
3529 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3530
3531 return 0;
3532}
3533
f01ec1c0
ND		 3534static void __net_exit vxlan_exit_net(struct net *net)
3535{
3536 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3537 struct vxlan_dev *vxlan, *next;
3538 struct net_device *dev, *aux;
3539 LIST_HEAD(list);
3540
3541 rtnl_lock();
3542 for_each_netdev_safe(net, dev, aux)
3543 if (dev->rtnl_link_ops == &vxlan_link_ops)
3544 unregister_netdevice_queue(dev, &list);
3545
3546 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3547 /* If vxlan->dev is in the same netns, it has already been added
3548 * to the list by the previous loop.
3549 */
58ce31cc
TH		 3550 if (!net_eq(dev_net(vxlan->dev), net)) {
3551 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3552 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3553 }
f01ec1c0
ND		 3554 }
3555
3556 unregister_netdevice_many(&list);
3557 rtnl_unlock();
3558}
3559
d342894c 3560static struct pernet_operations vxlan_net_ops = {
3561 .init = vxlan_init_net,
f01ec1c0 3562 .exit = vxlan_exit_net,
d342894c 3563 .id = &vxlan_net_id,
3564 .size = sizeof(struct vxlan_net),
3565};
3566
3567static int __init vxlan_init_module(void)
3568{
3569 int rc;
3570
3571 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3572
783c1463 3573 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3574 if (rc)
3575 goto out1;
3576
acaf4e70 3577 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3578 if (rc)
3579 goto out2;
3580
acaf4e70
DB		 3581 rc = rtnl_link_register(&vxlan_link_ops);
3582 if (rc)
3583 goto out3;
d342894c 3584
acaf4e70
DB		 3585 return 0;
3586out3:
3587 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3588out2:
783c1463 3589 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3590out1:
3591 return rc;
3592}
7332a13b 3593late_initcall(vxlan_init_module);
d342894c 3594
3595static void __exit vxlan_cleanup_module(void)
3596{
b7153984 3597 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3598 unregister_netdevice_notifier(&vxlan_notifier_block);
783c1463
DB		 3599 unregister_pernet_subsys(&vxlan_net_ops);
3600 /* rcu_barrier() is called by netns */
d342894c 3601}
3602module_exit(vxlan_cleanup_module);
3603
3604MODULE_LICENSE("GPL");
3605MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3606MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3607MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3608MODULE_ALIAS_RTNL_LINK("vxlan");
Ubuntu Artful kernel mirror