]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
vxlan: introduce vxlan_hdr
[mirror_ubuntu-bionic-kernel.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
d342894c 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/skbuff.h>
19#include <linux/rculist.h>
20#include <linux/netdevice.h>
21#include <linux/in.h>
22#include <linux/ip.h>
23#include <linux/udp.h>
24#include <linux/igmp.h>
25#include <linux/etherdevice.h>
26#include <linux/if_ether.h>
1eaa8178 27#include <linux/if_vlan.h>
d342894c 28#include <linux/hash.h>
1b13c97f 29#include <linux/ethtool.h>
e4f67add
DS		 30#include <net/arp.h>
31#include <net/ndisc.h>
d342894c 32#include <net/ip.h>
c5441932 33#include <net/ip_tunnels.h>
d342894c 34#include <net/icmp.h>
35#include <net/udp.h>
3ee64f39 36#include <net/udp_tunnel.h>
d342894c 37#include <net/rtnetlink.h>
38#include <net/route.h>
39#include <net/dsfield.h>
40#include <net/inet_ecn.h>
41#include <net/net_namespace.h>
42#include <net/netns/generic.h>
012a5729 43#include <net/vxlan.h>
dc01e7d3 44#include <net/protocol.h>
40d29af0 45
e4c7ed41
CW		 46#if IS_ENABLED(CONFIG_IPV6)
47#include <net/ipv6.h>
48#include <net/addrconf.h>
49#include <net/ip6_tunnel.h>
660d98ca 50#include <net/ip6_checksum.h>
e4c7ed41 51#endif
ee122c79 52#include <net/dst_metadata.h>
d342894c 53
54#define VXLAN_VERSION "0.1"
55
553675fb 56#define PORT_HASH_BITS 8
57#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 58#define FDB_AGE_DEFAULT 300 /* 5 min */
59#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
60
23c578bf 61/* UDP port for VXLAN traffic.
62 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 63 * for compatibility with early adopters.
23c578bf 64 */
9daaa397
SH		 65static unsigned short vxlan_port __read_mostly = 8472;
66module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 67MODULE_PARM_DESC(udp_port, "Destination UDP port");
68
69static bool log_ecn_error = true;
70module_param(log_ecn_error, bool, 0644);
71MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
72
60d9d4c6 73static int vxlan_net_id;
0dfbdf41 74static struct rtnl_link_ops vxlan_link_ops;
553675fb 75
7256eac1 76static const u8 all_zeros_mac[ETH_ALEN + 2];
afbd8bae 77
205f356d 78static int vxlan_sock_add(struct vxlan_dev *vxlan);
614732ea 79
553675fb 80/* per-network namespace private data for this module */
81struct vxlan_net {
82 struct list_head vxlan_list;
83 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 84 spinlock_t sock_lock;
553675fb 85};
86
d342894c 87/* Forwarding table entry */
88struct vxlan_fdb {
89 struct hlist_node hlist; /* linked list of entries */
90 struct rcu_head rcu;
91 unsigned long updated; /* jiffies */
92 unsigned long used;
3e61aa8f 93 struct list_head remotes;
7177a3b0 94 u8 eth_addr[ETH_ALEN];
d342894c 95 u16 state; /* see ndm_state */
ae884082 96 u8 flags; /* see ndm_flags */
d342894c 97};
98
d342894c 99/* salt for hash table */
100static u32 vxlan_salt __read_mostly;
758c57d1 101static struct workqueue_struct *vxlan_wq;
d342894c 102
ee122c79
TG		 103static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
104{
e7030878
TG		 105 return vs->flags & VXLAN_F_COLLECT_METADATA ||
106 ip_tunnel_collect_metadata();
ee122c79
TG		 107}
108
e4c7ed41
CW		 109#if IS_ENABLED(CONFIG_IPV6)
110static inline
111bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
112{
f0ef3126
JB		 113 if (a->sa.sa_family != b->sa.sa_family)
114 return false;
115 if (a->sa.sa_family == AF_INET6)
116 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
117 else
118 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 119}
120
121static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
122{
f0ef3126
JB		 123 if (ipa->sa.sa_family == AF_INET6)
124 return ipv6_addr_any(&ipa->sin6.sin6_addr);
125 else
126 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 127}
128
129static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
130{
f0ef3126
JB		 131 if (ipa->sa.sa_family == AF_INET6)
132 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
133 else
134 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 135}
136
137static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
138{
f0ef3126 139 if (nla_len(nla) >= sizeof(struct in6_addr)) {
67b61f6c 140 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
f0ef3126
JB		 141 ip->sa.sa_family = AF_INET6;
142 return 0;
143 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 144 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 145 ip->sa.sa_family = AF_INET;
146 return 0;
147 } else {
148 return -EAFNOSUPPORT;
149 }
e4c7ed41
CW		 150}
151
152static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 153 const union vxlan_addr *ip)
e4c7ed41 154{
f0ef3126 155 if (ip->sa.sa_family == AF_INET6)
930345ea 156 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
f0ef3126 157 else
930345ea 158 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 159}
160
161#else /* !CONFIG_IPV6 */
162
163static inline
164bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
165{
f0ef3126 166 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
e4c7ed41
CW		 167}
168
169static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
170{
f0ef3126 171 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
e4c7ed41
CW		 172}
173
174static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
175{
f0ef3126 176 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
e4c7ed41
CW		 177}
178
179static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
180{
f0ef3126
JB		 181 if (nla_len(nla) >= sizeof(struct in6_addr)) {
182 return -EAFNOSUPPORT;
183 } else if (nla_len(nla) >= sizeof(__be32)) {
67b61f6c 184 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
f0ef3126
JB		 185 ip->sa.sa_family = AF_INET;
186 return 0;
187 } else {
188 return -EAFNOSUPPORT;
189 }
e4c7ed41
CW		 190}
191
192static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
f0ef3126 193 const union vxlan_addr *ip)
e4c7ed41 194{
930345ea 195 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
e4c7ed41
CW		 196}
197#endif
198
553675fb 199/* Virtual Network hash table head */
200static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
201{
202 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
203}
204
205/* Socket hash table head */
206static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 207{
208 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
209
553675fb 210 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
211}
212
3e61aa8f
SH		 213/* First remote destination for a forwarding entry.
214 * Guaranteed to be non-NULL because remotes are never deleted.
215 */
5ca5461c 216static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 217{
5ca5461c 218 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
219}
220
221static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
222{
223 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 224}
225
ac5132d1
TG		 226/* Find VXLAN socket based on network namespace, address family and UDP port
227 * and enabled unshareable flags.
228 */
229static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
230 __be16 port, u32 flags)
553675fb 231{
232 struct vxlan_sock *vs;
af33c1ad
TH		 233
234 flags &= VXLAN_F_RCV_FLAGS;
553675fb 235
236 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
19ca9fc1 237 if (inet_sk(vs->sock->sk)->inet_sport == port &&
705cc62f 238 vxlan_get_sk_family(vs) == family &&
af33c1ad 239 vs->flags == flags)
553675fb 240 return vs;
241 }
242 return NULL;
d342894c 243}
244
5cfccc5a 245static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
d342894c 246{
247 struct vxlan_dev *vxlan;
d342894c 248
553675fb 249 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
c7995c43 250 if (vxlan->default_dst.remote_vni == id)
d342894c 251 return vxlan;
252 }
253
254 return NULL;
255}
256
5cfccc5a 257/* Look up VNI in a per net namespace table */
19ca9fc1 258static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id,
ac5132d1
TG		 259 sa_family_t family, __be16 port,
260 u32 flags)
5cfccc5a
PS		 261{
262 struct vxlan_sock *vs;
263
ac5132d1 264 vs = vxlan_find_sock(net, family, port, flags);
5cfccc5a
PS		 265 if (!vs)
266 return NULL;
267
268 return vxlan_vs_find_vni(vs, id);
269}
270
d342894c 271/* Fill in neighbour message in skbuff. */
272static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 273 const struct vxlan_fdb *fdb,
274 u32 portid, u32 seq, int type, unsigned int flags,
275 const struct vxlan_rdst *rdst)
d342894c 276{
277 unsigned long now = jiffies;
278 struct nda_cacheinfo ci;
279 struct nlmsghdr *nlh;
280 struct ndmsg *ndm;
e4f67add 281 bool send_ip, send_eth;
d342894c 282
283 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
284 if (nlh == NULL)
285 return -EMSGSIZE;
286
287 ndm = nlmsg_data(nlh);
288 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 289
290 send_eth = send_ip = true;
291
292 if (type == RTM_GETNEIGH) {
293 ndm->ndm_family = AF_INET;
e4c7ed41 294 send_ip = !vxlan_addr_any(&rdst->remote_ip);
e4f67add
DS		 295 send_eth = !is_zero_ether_addr(fdb->eth_addr);
296 } else
297 ndm->ndm_family = AF_BRIDGE;
d342894c 298 ndm->ndm_state = fdb->state;
299 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 300 ndm->ndm_flags = fdb->flags;
545469f7 301 ndm->ndm_type = RTN_UNICAST;
d342894c 302
193523bf 303 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
4967082b 304 nla_put_s32(skb, NDA_LINK_NETNSID,
7a0877d4 305 peernet2id_alloc(dev_net(vxlan->dev), vxlan->net)))
193523bf
ND		 306 goto nla_put_failure;
307
e4f67add 308 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 309 goto nla_put_failure;
310
e4c7ed41 311 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
6681712d
DS		 312 goto nla_put_failure;
313
0dfbdf41 314 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
6681712d
DS		 315 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
316 goto nla_put_failure;
c7995c43 317 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 318 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 319 goto nla_put_failure;
320 if (rdst->remote_ifindex &&
321 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 322 goto nla_put_failure;
323
324 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
325 ci.ndm_confirmed = 0;
326 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
327 ci.ndm_refcnt = 0;
328
329 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
330 goto nla_put_failure;
331
053c095a
JB		 332 nlmsg_end(skb, nlh);
333 return 0;
d342894c 334
335nla_put_failure:
336 nlmsg_cancel(skb, nlh);
337 return -EMSGSIZE;
338}
339
340static inline size_t vxlan_nlmsg_size(void)
341{
342 return NLMSG_ALIGN(sizeof(struct ndmsg))
343 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
e4c7ed41 344 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
73cf3317 345 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 346 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
347 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
4967082b 348 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
d342894c 349 + nla_total_size(sizeof(struct nda_cacheinfo));
350}
351
9e4b93f9
ND		 352static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
353 struct vxlan_rdst *rd, int type)
d342894c 354{
355 struct net *net = dev_net(vxlan->dev);
356 struct sk_buff *skb;
357 int err = -ENOBUFS;
358
359 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
360 if (skb == NULL)
361 goto errout;
362
9e4b93f9 363 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
d342894c 364 if (err < 0) {
365 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
366 WARN_ON(err == -EMSGSIZE);
367 kfree_skb(skb);
368 goto errout;
369 }
370
371 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
372 return;
373errout:
374 if (err < 0)
375 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
376}
377
e4c7ed41 378static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
e4f67add
DS		 379{
380 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 381 struct vxlan_fdb f = {
382 .state = NUD_STALE,
383 };
384 struct vxlan_rdst remote = {
e4c7ed41 385 .remote_ip = *ipa, /* goes to NDA_DST */
bb3fd687
SH		 386 .remote_vni = VXLAN_N_VID,
387 };
3e61aa8f 388
9e4b93f9 389 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 390}
391
392static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
393{
bb3fd687
SH		 394 struct vxlan_fdb f = {
395 .state = NUD_STALE,
396 };
9e4b93f9 397 struct vxlan_rdst remote = { };
e4f67add 398
e4f67add
DS		 399 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
400
9e4b93f9 401 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
e4f67add
DS		 402}
403
d342894c 404/* Hash Ethernet address */
405static u32 eth_hash(const unsigned char *addr)
406{
407 u64 value = get_unaligned((u64 *)addr);
408
409 /* only want 6 bytes */
410#ifdef __BIG_ENDIAN
d342894c 411 value >>= 16;
321fb991 412#else
413 value <<= 16;
d342894c 414#endif
415 return hash_64(value, FDB_HASH_BITS);
416}
417
418/* Hash chain to use given mac address */
419static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
420 const u8 *mac)
421{
422 return &vxlan->fdb_head[eth_hash(mac)];
423}
424
425/* Look up Ethernet address in forwarding table */
014be2c8 426static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 427 const u8 *mac)
d342894c 428{
429 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
430 struct vxlan_fdb *f;
d342894c 431
b67bfe0d 432 hlist_for_each_entry_rcu(f, head, hlist) {
7367d0b5 433 if (ether_addr_equal(mac, f->eth_addr))
d342894c 434 return f;
435 }
436
437 return NULL;
438}
439
014be2c8
SS		 440static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
441 const u8 *mac)
442{
443 struct vxlan_fdb *f;
444
445 f = __vxlan_find_mac(vxlan, mac);
446 if (f)
447 f->used = jiffies;
448
449 return f;
450}
451
a5e7c10a
MR		 452/* caller should hold vxlan->hash_lock */
453static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
e4c7ed41 454 union vxlan_addr *ip, __be16 port,
a5e7c10a 455 __u32 vni, __u32 ifindex)
6681712d 456{
3e61aa8f 457 struct vxlan_rdst *rd;
6681712d 458
3e61aa8f 459 list_for_each_entry(rd, &f->remotes, list) {
e4c7ed41 460 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
6681712d
DS		 461 rd->remote_port == port &&
462 rd->remote_vni == vni &&
463 rd->remote_ifindex == ifindex)
a5e7c10a 464 return rd;
6681712d 465 }
3e61aa8f 466
a5e7c10a
MR		 467 return NULL;
468}
469
906dc186
TR		 470/* Replace destination of unicast mac */
471static int vxlan_fdb_replace(struct vxlan_fdb *f,
e4c7ed41 472 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
906dc186
TR		 473{
474 struct vxlan_rdst *rd;
475
476 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
477 if (rd)
478 return 0;
479
480 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
481 if (!rd)
482 return 0;
0c1d70af
PA		 483
484 dst_cache_reset(&rd->dst_cache);
e4c7ed41 485 rd->remote_ip = *ip;
906dc186
TR		 486 rd->remote_port = port;
487 rd->remote_vni = vni;
488 rd->remote_ifindex = ifindex;
489 return 1;
490}
491
a5e7c10a
MR		 492/* Add/update destinations for multicast */
493static int vxlan_fdb_append(struct vxlan_fdb *f,
9e4b93f9
ND		 494 union vxlan_addr *ip, __be16 port, __u32 vni,
495 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 496{
497 struct vxlan_rdst *rd;
498
499 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
500 if (rd)
501 return 0;
502
6681712d
DS		 503 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
504 if (rd == NULL)
505 return -ENOBUFS;
0c1d70af
PA		 506
507 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
508 kfree(rd);
509 return -ENOBUFS;
510 }
511
e4c7ed41 512 rd->remote_ip = *ip;
6681712d
DS		 513 rd->remote_port = port;
514 rd->remote_vni = vni;
515 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 516
517 list_add_tail_rcu(&rd->list, &f->remotes);
518
9e4b93f9 519 *rdp = rd;
6681712d
DS		 520 return 1;
521}
522
dfd8645e
TH		 523static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
524 unsigned int off,
525 struct vxlanhdr *vh, size_t hdrlen,
0ace2ca8
TH		 526 u32 data, struct gro_remcsum *grc,
527 bool nopartial)
dfd8645e 528{
b7fe10e5 529 size_t start, offset;
dfd8645e
TH		 530
531 if (skb->remcsum_offload)
b7fe10e5 532 return vh;
dfd8645e
TH		 533
534 if (!NAPI_GRO_CB(skb)->csum_valid)
535 return NULL;
536
537 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
538 offset = start + ((data & VXLAN_RCO_UDP) ?
539 offsetof(struct udphdr, check) :
540 offsetof(struct tcphdr, check));
541
b7fe10e5
TH		 542 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
543 start, offset, grc, nopartial);
dfd8645e
TH		 544
545 skb->remcsum_offload = 1;
546
547 return vh;
548}
549
a2b12f3c
TH		 550static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
551 struct sk_buff *skb,
552 struct udp_offload *uoff)
dc01e7d3
OG		 553{
554 struct sk_buff *p, **pp = NULL;
555 struct vxlanhdr *vh, *vh2;
9b174d88 556 unsigned int hlen, off_vx;
dc01e7d3 557 int flush = 1;
dfd8645e
TH		 558 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
559 udp_offloads);
560 u32 flags;
26c4f7da
TH		 561 struct gro_remcsum grc;
562
563 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 564
565 off_vx = skb_gro_offset(skb);
566 hlen = off_vx + sizeof(*vh);
567 vh = skb_gro_header_fast(skb, off_vx);
568 if (skb_gro_header_hard(skb, hlen)) {
569 vh = skb_gro_header_slow(skb, hlen, off_vx);
570 if (unlikely(!vh))
571 goto out;
572 }
dc01e7d3 573
dfd8645e
TH		 574 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
575
576 flags = ntohl(vh->vx_flags);
577
578 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
579 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
0ace2ca8
TH		 580 ntohl(vh->vx_vni), &grc,
581 !!(vs->flags &
582 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 583
584 if (!vh)
585 goto out;
586 }
587
b7fe10e5
TH		 588 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
589
dc01e7d3
OG		 590 flush = 0;
591
592 for (p = *head; p; p = p->next) {
593 if (!NAPI_GRO_CB(p)->same_flow)
594 continue;
595
596 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 597 if (vh->vx_flags != vh2->vx_flags ||
598 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 599 NAPI_GRO_CB(p)->same_flow = 0;
600 continue;
601 }
dc01e7d3
OG		 602 }
603
9b174d88 604 pp = eth_gro_receive(head, skb);
dc01e7d3 605
dc01e7d3 606out:
26c4f7da 607 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 608 NAPI_GRO_CB(skb)->flush |= flush;
609
610 return pp;
611}
612
a2b12f3c
TH		 613static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
614 struct udp_offload *uoff)
dc01e7d3 615{
cfdf1e1b
JG		 616 udp_tunnel_gro_complete(skb, nhoff);
617
9b174d88 618 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 619}
620
53cf5275 621/* Notify netdevs that UDP port started listening */
dc01e7d3 622static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 623{
624 struct net_device *dev;
dc01e7d3 625 struct sock *sk = vs->sock->sk;
53cf5275 626 struct net *net = sock_net(sk);
705cc62f 627 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 628 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG		 629 int err;
630
631 if (sa_family == AF_INET) {
787d7ac3 632 err = udp_add_offload(net, &vs->udp_offloads);
dc01e7d3
OG		 633 if (err)
634 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
635 }
53cf5275
JG		 636
637 rcu_read_lock();
638 for_each_netdev_rcu(net, dev) {
639 if (dev->netdev_ops->ndo_add_vxlan_port)
640 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
641 port);
642 }
643 rcu_read_unlock();
644}
645
646/* Notify netdevs that UDP port is no more listening */
dc01e7d3 647static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 648{
649 struct net_device *dev;
dc01e7d3 650 struct sock *sk = vs->sock->sk;
53cf5275 651 struct net *net = sock_net(sk);
705cc62f 652 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 653 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 654
655 rcu_read_lock();
656 for_each_netdev_rcu(net, dev) {
657 if (dev->netdev_ops->ndo_del_vxlan_port)
658 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
659 port);
660 }
661 rcu_read_unlock();
dc01e7d3
OG		 662
663 if (sa_family == AF_INET)
664 udp_del_offload(&vs->udp_offloads);
53cf5275
JG		 665}
666
d342894c 667/* Add new entry to forwarding table -- assumes lock held */
668static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 669 const u8 *mac, union vxlan_addr *ip,
6681712d 670 __u16 state, __u16 flags,
73cf3317 671 __be16 port, __u32 vni, __u32 ifindex,
ae884082 672 __u8 ndm_flags)
d342894c 673{
9e4b93f9 674 struct vxlan_rdst *rd = NULL;
d342894c 675 struct vxlan_fdb *f;
676 int notify = 0;
677
014be2c8 678 f = __vxlan_find_mac(vxlan, mac);
d342894c 679 if (f) {
680 if (flags & NLM_F_EXCL) {
681 netdev_dbg(vxlan->dev,
682 "lost race to create %pM\n", mac);
683 return -EEXIST;
684 }
685 if (f->state != state) {
686 f->state = state;
687 f->updated = jiffies;
688 notify = 1;
689 }
ae884082
DS		 690 if (f->flags != ndm_flags) {
691 f->flags = ndm_flags;
692 f->updated = jiffies;
693 notify = 1;
694 }
906dc186
TR		 695 if ((flags & NLM_F_REPLACE)) {
696 /* Only change unicasts */
697 if (!(is_multicast_ether_addr(f->eth_addr) ||
698 is_zero_ether_addr(f->eth_addr))) {
60840429 699 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 700 ifindex);
906dc186
TR		 701 } else
702 return -EOPNOTSUPP;
703 }
6681712d 704 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 705 (is_multicast_ether_addr(f->eth_addr) ||
706 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 707 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
708 &rd);
6681712d
DS		 709
710 if (rc < 0)
711 return rc;
712 notify |= rc;
713 }
d342894c 714 } else {
715 if (!(flags & NLM_F_CREATE))
716 return -ENOENT;
717
0dfbdf41
TG		 718 if (vxlan->cfg.addrmax &&
719 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 720 return -ENOSPC;
721
906dc186
TR		 722 /* Disallow replace to add a multicast entry */
723 if ((flags & NLM_F_REPLACE) &&
724 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
725 return -EOPNOTSUPP;
726
e4c7ed41 727 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 728 f = kmalloc(sizeof(*f), GFP_ATOMIC);
729 if (!f)
730 return -ENOMEM;
731
732 notify = 1;
d342894c 733 f->state = state;
ae884082 734 f->flags = ndm_flags;
d342894c 735 f->updated = f->used = jiffies;
3e61aa8f 736 INIT_LIST_HEAD(&f->remotes);
d342894c 737 memcpy(f->eth_addr, mac, ETH_ALEN);
738
9e4b93f9 739 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 740
d342894c 741 ++vxlan->addrcnt;
742 hlist_add_head_rcu(&f->hlist,
743 vxlan_fdb_head(vxlan, mac));
744 }
745
9e4b93f9
ND		 746 if (notify) {
747 if (rd == NULL)
748 rd = first_remote_rtnl(f);
749 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
750 }
d342894c 751
752 return 0;
753}
754
6706c82e 755static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 756{
757 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 758 struct vxlan_rdst *rd, *nd;
6681712d 759
0c1d70af
PA		 760 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
761 dst_cache_destroy(&rd->dst_cache);
6681712d 762 kfree(rd);
0c1d70af 763 }
6681712d
DS		 764 kfree(f);
765}
766
d342894c 767static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
768{
769 netdev_dbg(vxlan->dev,
770 "delete %pM\n", f->eth_addr);
771
772 --vxlan->addrcnt;
9e4b93f9 773 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 774
775 hlist_del_rcu(&f->hlist);
6681712d 776 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 777}
778
f0b074be 779static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 780 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 781{
6681712d 782 struct net *net = dev_net(vxlan->dev);
e4c7ed41 783 int err;
d342894c 784
f0b074be 785 if (tb[NDA_DST]) {
e4c7ed41
CW		 786 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
787 if (err)
788 return err;
f0b074be 789 } else {
e4c7ed41
CW		 790 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
791 if (remote->sa.sa_family == AF_INET) {
792 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
793 ip->sa.sa_family = AF_INET;
794#if IS_ENABLED(CONFIG_IPV6)
795 } else {
796 ip->sin6.sin6_addr = in6addr_any;
797 ip->sa.sa_family = AF_INET6;
798#endif
799 }
f0b074be 800 }
d342894c 801
6681712d 802 if (tb[NDA_PORT]) {
73cf3317 803 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 804 return -EINVAL;
f0b074be
MR		 805 *port = nla_get_be16(tb[NDA_PORT]);
806 } else {
0dfbdf41 807 *port = vxlan->cfg.dst_port;
f0b074be 808 }
6681712d
DS		 809
810 if (tb[NDA_VNI]) {
811 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
812 return -EINVAL;
f0b074be
MR		 813 *vni = nla_get_u32(tb[NDA_VNI]);
814 } else {
815 *vni = vxlan->default_dst.remote_vni;
816 }
6681712d
DS		 817
818 if (tb[NDA_IFINDEX]) {
5abb0029 819 struct net_device *tdev;
6681712d
DS		 820
821 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
822 return -EINVAL;
f0b074be 823 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 824 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 825 if (!tdev)
6681712d 826 return -EADDRNOTAVAIL;
f0b074be
MR		 827 } else {
828 *ifindex = 0;
829 }
830
831 return 0;
832}
833
834/* Add static entry (via netlink) */
835static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
836 struct net_device *dev,
f6f6424b 837 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 838{
839 struct vxlan_dev *vxlan = netdev_priv(dev);
840 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 841 union vxlan_addr ip;
f0b074be
MR		 842 __be16 port;
843 u32 vni, ifindex;
844 int err;
845
846 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
847 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
848 ndm->ndm_state);
849 return -EINVAL;
850 }
851
852 if (tb[NDA_DST] == NULL)
853 return -EINVAL;
854
855 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
856 if (err)
857 return err;
6681712d 858
5933a7bb
MR		 859 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
860 return -EAFNOSUPPORT;
861
d342894c 862 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 863 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 864 port, vni, ifindex, ndm->ndm_flags);
d342894c 865 spin_unlock_bh(&vxlan->hash_lock);
866
867 return err;
868}
869
870/* Delete entry (via netlink) */
1690be63
VY		 871static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
872 struct net_device *dev,
f6f6424b 873 const unsigned char *addr, u16 vid)
d342894c 874{
875 struct vxlan_dev *vxlan = netdev_priv(dev);
876 struct vxlan_fdb *f;
bc7892ba 877 struct vxlan_rdst *rd = NULL;
e4c7ed41 878 union vxlan_addr ip;
bc7892ba
MR		 879 __be16 port;
880 u32 vni, ifindex;
881 int err;
882
883 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
884 if (err)
885 return err;
886
887 err = -ENOENT;
d342894c 888
889 spin_lock_bh(&vxlan->hash_lock);
890 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 891 if (!f)
892 goto out;
893
e4c7ed41
CW		 894 if (!vxlan_addr_any(&ip)) {
895 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 896 if (!rd)
897 goto out;
898 }
899
900 err = 0;
901
902 /* remove a destination if it's not the only one on the list,
903 * otherwise destroy the fdb entry
904 */
905 if (rd && !list_is_singular(&f->remotes)) {
906 list_del_rcu(&rd->list);
9e4b93f9 907 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 908 kfree_rcu(rd, rcu);
bc7892ba 909 goto out;
d342894c 910 }
bc7892ba
MR		 911
912 vxlan_fdb_destroy(vxlan, f);
913
914out:
d342894c 915 spin_unlock_bh(&vxlan->hash_lock);
916
917 return err;
918}
919
920/* Dump forwarding table */
921static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 922 struct net_device *dev,
923 struct net_device *filter_dev, int idx)
d342894c 924{
925 struct vxlan_dev *vxlan = netdev_priv(dev);
926 unsigned int h;
927
928 for (h = 0; h < FDB_HASH_SIZE; ++h) {
929 struct vxlan_fdb *f;
d342894c 930 int err;
931
b67bfe0d 932 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 933 struct vxlan_rdst *rd;
6681712d 934
3e61aa8f 935 list_for_each_entry_rcu(rd, &f->remotes, list) {
07a51cd3
AW		 936 if (idx < cb->args[0])
937 goto skip;
938
6681712d
DS		 939 err = vxlan_fdb_info(skb, vxlan, f,
940 NETLINK_CB(cb->skb).portid,
941 cb->nlh->nlmsg_seq,
942 RTM_NEWNEIGH,
943 NLM_F_MULTI, rd);
944 if (err < 0)
3e61aa8f 945 goto out;
3e61aa8f 946skip:
07a51cd3
AW		 947 ++idx;
948 }
d342894c 949 }
950 }
3e61aa8f 951out:
d342894c 952 return idx;
953}
954
955/* Watch incoming packets to learn mapping between Ethernet address
956 * and Tunnel endpoint.
c4b49512 957 * Return true if packet is bogus and should be dropped.
d342894c 958 */
26a41ae6 959static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 960 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 961{
962 struct vxlan_dev *vxlan = netdev_priv(dev);
963 struct vxlan_fdb *f;
d342894c 964
965 f = vxlan_find_mac(vxlan, src_mac);
966 if (likely(f)) {
5ca5461c 967 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 968
e4c7ed41 969 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 970 return false;
971
972 /* Don't migrate static entries, drop packets */
eb064c3b 973 if (f->state & NUD_NOARP)
26a41ae6 974 return true;
d342894c 975
976 if (net_ratelimit())
977 netdev_info(dev,
e4c7ed41 978 "%pM migrated from %pIS to %pIS\n",
a4870f79 979 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 980
e4c7ed41 981 rdst->remote_ip = *src_ip;
d342894c 982 f->updated = jiffies;
9e4b93f9 983 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 984 } else {
985 /* learned new entry */
986 spin_lock(&vxlan->hash_lock);
3bf74b1a 987
988 /* close off race between vxlan_flush and incoming packets */
989 if (netif_running(dev))
990 vxlan_fdb_create(vxlan, src_mac, src_ip,
991 NUD_REACHABLE,
992 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 993 vxlan->cfg.dst_port,
3bf74b1a 994 vxlan->default_dst.remote_vni,
995 0, NTF_SELF);
d342894c 996 spin_unlock(&vxlan->hash_lock);
997 }
26a41ae6 998
999 return false;
d342894c 1000}
1001
d342894c 1002/* See if multicast group is already in use by other ID */
95ab0991 1003static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 1004{
553675fb 1005 struct vxlan_dev *vxlan;
b1be00a6 1006 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 1007
95ab0991
G		 1008 /* The vxlan_sock is only used by dev, leaving group has
1009 * no effect on other vxlan devices.
1010 */
b1be00a6
JB		 1011 if (family == AF_INET && dev->vn4_sock &&
1012 atomic_read(&dev->vn4_sock->refcnt) == 1)
95ab0991 1013 return false;
b1be00a6
JB		 1014#if IS_ENABLED(CONFIG_IPV6)
1015 if (family == AF_INET6 && dev->vn6_sock &&
1016 atomic_read(&dev->vn6_sock->refcnt) == 1)
1017 return false;
1018#endif
95ab0991 1019
553675fb 1020 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1021 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1022 continue;
d342894c 1023
b1be00a6 1024 if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock)
95ab0991 1025 continue;
b1be00a6
JB		 1026#if IS_ENABLED(CONFIG_IPV6)
1027 if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock)
1028 continue;
1029#endif
95ab0991
G		 1030
1031 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1032 &dev->default_dst.remote_ip))
1033 continue;
1034
1035 if (vxlan->default_dst.remote_ifindex !=
1036 dev->default_dst.remote_ifindex)
1037 continue;
1038
1039 return true;
553675fb 1040 }
d342894c 1041
1042 return false;
1043}
1044
b1be00a6 1045static void __vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1046{
b1be00a6 1047 struct vxlan_net *vn;
012a5729 1048
b1be00a6
JB		 1049 if (!vs)
1050 return;
7c47cedf
SH		 1051 if (!atomic_dec_and_test(&vs->refcnt))
1052 return;
d342894c 1053
b1be00a6 1054 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1055 spin_lock(&vn->sock_lock);
7c47cedf 1056 hlist_del_rcu(&vs->hlist);
dc01e7d3 1057 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1058 spin_unlock(&vn->sock_lock);
1059
7c47cedf 1060 queue_work(vxlan_wq, &vs->del_work);
d342894c 1061}
1062
b1be00a6
JB		 1063static void vxlan_sock_release(struct vxlan_dev *vxlan)
1064{
1065 __vxlan_sock_release(vxlan->vn4_sock);
1066#if IS_ENABLED(CONFIG_IPV6)
1067 __vxlan_sock_release(vxlan->vn6_sock);
1068#endif
1069}
1070
56ef9c90 1071/* Update multicast group membership when first VNI on
c4b49512 1072 * multicast address is brought up
7c47cedf 1073 */
56ef9c90 1074static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1075{
b1be00a6 1076 struct sock *sk;
e4c7ed41
CW		 1077 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1078 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1079 int ret = -EINVAL;
d342894c 1080
e4c7ed41
CW		 1081 if (ip->sa.sa_family == AF_INET) {
1082 struct ip_mreqn mreq = {
1083 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1084 .imr_ifindex = ifindex,
1085 };
1086
b1be00a6
JB		 1087 sk = vxlan->vn4_sock->sock->sk;
1088 lock_sock(sk);
56ef9c90 1089 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1090 release_sock(sk);
e4c7ed41
CW		 1091#if IS_ENABLED(CONFIG_IPV6)
1092 } else {
b1be00a6
JB		 1093 sk = vxlan->vn6_sock->sock->sk;
1094 lock_sock(sk);
56ef9c90
MRL		 1095 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1096 &ip->sin6.sin6_addr);
b1be00a6 1097 release_sock(sk);
e4c7ed41
CW		 1098#endif
1099 }
3fc2de2f 1100
56ef9c90 1101 return ret;
3fc2de2f 1102}
1103
1104/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1105static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1106{
b1be00a6 1107 struct sock *sk;
e4c7ed41
CW		 1108 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1109 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1110 int ret = -EINVAL;
3fc2de2f 1111
e4c7ed41
CW		 1112 if (ip->sa.sa_family == AF_INET) {
1113 struct ip_mreqn mreq = {
1114 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1115 .imr_ifindex = ifindex,
1116 };
1117
b1be00a6
JB		 1118 sk = vxlan->vn4_sock->sock->sk;
1119 lock_sock(sk);
56ef9c90 1120 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1121 release_sock(sk);
e4c7ed41
CW		 1122#if IS_ENABLED(CONFIG_IPV6)
1123 } else {
b1be00a6
JB		 1124 sk = vxlan->vn6_sock->sock->sk;
1125 lock_sock(sk);
56ef9c90
MRL		 1126 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1127 &ip->sin6.sin6_addr);
b1be00a6 1128 release_sock(sk);
e4c7ed41
CW		 1129#endif
1130 }
d342894c 1131
56ef9c90 1132 return ret;
d342894c 1133}
1134
dfd8645e 1135static struct vxlanhdr *vxlan_remcsum(struct sk_buff *skb, struct vxlanhdr *vh,
0ace2ca8 1136 size_t hdrlen, u32 data, bool nopartial)
dfd8645e
TH		 1137{
1138 size_t start, offset, plen;
dfd8645e 1139
b7fe10e5
TH		 1140 if (skb->remcsum_offload)
1141 return vh;
1142
dfd8645e
TH		 1143 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
1144 offset = start + ((data & VXLAN_RCO_UDP) ?
1145 offsetof(struct udphdr, check) :
1146 offsetof(struct tcphdr, check));
1147
1148 plen = hdrlen + offset + sizeof(u16);
1149
1150 if (!pskb_may_pull(skb, plen))
1151 return NULL;
1152
1153 vh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1154
0ace2ca8
TH		 1155 skb_remcsum_process(skb, (void *)vh + hdrlen, start, offset,
1156 nopartial);
dfd8645e
TH		 1157
1158 return vh;
1159}
1160
614732ea
TG		 1161static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
1162 struct vxlan_metadata *md, u32 vni,
1163 struct metadata_dst *tun_dst)
1164{
1165 struct iphdr *oip = NULL;
1166 struct ipv6hdr *oip6 = NULL;
1167 struct vxlan_dev *vxlan;
1168 struct pcpu_sw_netstats *stats;
1169 union vxlan_addr saddr;
1170 int err = 0;
614732ea
TG		 1171
1172 /* For flow based devices, map all packets to VNI 0 */
da8b43c0 1173 if (vs->flags & VXLAN_F_COLLECT_METADATA)
614732ea
TG		 1174 vni = 0;
1175
1176 /* Is this VNI defined? */
1177 vxlan = vxlan_vs_find_vni(vs, vni);
1178 if (!vxlan)
1179 goto drop;
1180
614732ea
TG		 1181 skb_reset_mac_header(skb);
1182 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
1183 skb->protocol = eth_type_trans(skb, vxlan->dev);
1184 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1185
1186 /* Ignore packet loops (and multicast echo) */
1187 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1188 goto drop;
1189
ce212d0f
JB		 1190 /* Get data from the outer IP header */
1191 if (vxlan_get_sk_family(vs) == AF_INET) {
614732ea
TG		 1192 oip = ip_hdr(skb);
1193 saddr.sin.sin_addr.s_addr = oip->saddr;
1194 saddr.sa.sa_family = AF_INET;
1195#if IS_ENABLED(CONFIG_IPV6)
1196 } else {
1197 oip6 = ipv6_hdr(skb);
1198 saddr.sin6.sin6_addr = oip6->saddr;
1199 saddr.sa.sa_family = AF_INET6;
1200#endif
1201 }
1202
1203 if (tun_dst) {
1204 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1205 tun_dst = NULL;
1206 }
1207
1208 if ((vxlan->flags & VXLAN_F_LEARN) &&
1209 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1210 goto drop;
1211
1212 skb_reset_network_header(skb);
1213 /* In flow-based mode, GBP is carried in dst_metadata */
da8b43c0 1214 if (!(vs->flags & VXLAN_F_COLLECT_METADATA))
614732ea
TG		 1215 skb->mark = md->gbp;
1216
1217 if (oip6)
1218 err = IP6_ECN_decapsulate(oip6, skb);
1219 if (oip)
1220 err = IP_ECN_decapsulate(oip, skb);
1221
1222 if (unlikely(err)) {
1223 if (log_ecn_error) {
1224 if (oip6)
1225 net_info_ratelimited("non-ECT from %pI6\n",
1226 &oip6->saddr);
1227 if (oip)
1228 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1229 &oip->saddr, oip->tos);
1230 }
1231 if (err > 1) {
1232 ++vxlan->dev->stats.rx_frame_errors;
1233 ++vxlan->dev->stats.rx_errors;
1234 goto drop;
1235 }
1236 }
1237
1238 stats = this_cpu_ptr(vxlan->dev->tstats);
1239 u64_stats_update_begin(&stats->syncp);
1240 stats->rx_packets++;
1241 stats->rx_bytes += skb->len;
1242 u64_stats_update_end(&stats->syncp);
1243
58ce31cc 1244 gro_cells_receive(&vxlan->gro_cells, skb);
614732ea
TG		 1245
1246 return;
1247drop:
1248 if (tun_dst)
1249 dst_release((struct dst_entry *)tun_dst);
1250
1251 /* Consume bad packet */
1252 kfree_skb(skb);
1253}
1254
d342894c 1255/* Callback from net/ipv4/udp.c to receive packets */
1256static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1257{
ee122c79 1258 struct metadata_dst *tun_dst = NULL;
5cfccc5a 1259 struct vxlan_sock *vs;
3bf39475 1260 u32 flags, vni;
ee122c79
TG		 1261 struct vxlan_metadata _md;
1262 struct vxlan_metadata *md = &_md;
d342894c 1263
d342894c 1264 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1265 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 1266 goto error;
1267
d4ac05ff
JB		 1268 flags = ntohl(vxlan_hdr(skb)->vx_flags);
1269 vni = ntohl(vxlan_hdr(skb)->vx_vni);
3bf39475
TH		 1270
1271 if (flags & VXLAN_HF_VNI) {
1272 flags &= ~VXLAN_HF_VNI;
1273 } else {
1274 /* VNI flag always required to be set */
1275 goto bad_flags;
d342894c 1276 }
1277
5cfccc5a
PS		 1278 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1279 goto drop;
1280
559835ea 1281 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1282 if (!vs)
d342894c 1283 goto drop;
d342894c 1284
dfd8645e 1285 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
d4ac05ff
JB		 1286 if (!vxlan_remcsum(skb, vxlan_hdr(skb), sizeof(struct vxlanhdr), vni,
1287 !!(vs->flags & VXLAN_F_REMCSUM_NOPARTIAL)))
dfd8645e
TH		 1288 goto drop;
1289
1290 flags &= ~VXLAN_HF_RCO;
40fb70f3 1291 vni &= VXLAN_VNI_MASK;
dfd8645e
TH		 1292 }
1293
ee122c79 1294 if (vxlan_collect_metadata(vs)) {
c29a70d2
PS		 1295 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1296 cpu_to_be64(vni >> 8), sizeof(*md));
1297
ee122c79
TG		 1298 if (!tun_dst)
1299 goto drop;
1300
0f1b7354 1301 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
ee122c79
TG		 1302 } else {
1303 memset(md, 0, sizeof(*md));
1304 }
1305
3511494c
TG		 1306 /* For backwards compatibility, only allow reserved fields to be
1307 * used by VXLAN extensions if explicitly requested.
1308 */
1309 if ((flags & VXLAN_HF_GBP) && (vs->flags & VXLAN_F_GBP)) {
1310 struct vxlanhdr_gbp *gbp;
1311
d4ac05ff 1312 gbp = (struct vxlanhdr_gbp *)vxlan_hdr(skb);
ee122c79
TG		 1313 md->gbp = ntohs(gbp->policy_id);
1314
1315 if (tun_dst)
0f1b7354 1316 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
3511494c
TG		 1317
1318 if (gbp->dont_learn)
ee122c79 1319 md->gbp |= VXLAN_GBP_DONT_LEARN;
3511494c
TG		 1320
1321 if (gbp->policy_applied)
ee122c79 1322 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
3511494c
TG		 1323
1324 flags &= ~VXLAN_GBP_USED_BITS;
1325 }
1326
40fb70f3 1327 if (flags || vni & ~VXLAN_VNI_MASK) {
3bf39475
TH		 1328 /* If there are any unprocessed flags remaining treat
1329 * this as a malformed packet. This behavior diverges from
1330 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1331 * in reserved fields are to be ignored. The approach here
c4b49512 1332 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1333 * is more robust and provides a little more security in
1334 * adding extensions to VXLAN.
1335 */
1336
1337 goto bad_flags;
1338 }
1339
614732ea 1340 vxlan_rcv(vs, skb, md, vni >> 8, tun_dst);
5cfccc5a
PS		 1341 return 0;
1342
1343drop:
1344 /* Consume bad packet */
1345 kfree_skb(skb);
1346 return 0;
1347
3bf39475
TH		 1348bad_flags:
1349 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
d4ac05ff
JB		 1350 ntohl(vxlan_hdr(skb)->vx_flags),
1351 ntohl(vxlan_hdr(skb)->vx_vni));
3bf39475 1352
5cfccc5a 1353error:
ee122c79
TG		 1354 if (tun_dst)
1355 dst_release((struct dst_entry *)tun_dst);
1356
5cfccc5a
PS		 1357 /* Return non vxlan pkt */
1358 return 1;
1359}
1360
e4f67add
DS		 1361static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1362{
1363 struct vxlan_dev *vxlan = netdev_priv(dev);
1364 struct arphdr *parp;
1365 u8 *arpptr, *sha;
1366 __be32 sip, tip;
1367 struct neighbour *n;
1368
1369 if (dev->flags & IFF_NOARP)
1370 goto out;
1371
1372 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1373 dev->stats.tx_dropped++;
1374 goto out;
1375 }
1376 parp = arp_hdr(skb);
1377
1378 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1379 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1380 parp->ar_pro != htons(ETH_P_IP) ||
1381 parp->ar_op != htons(ARPOP_REQUEST) ||
1382 parp->ar_hln != dev->addr_len ||
1383 parp->ar_pln != 4)
1384 goto out;
1385 arpptr = (u8 *)parp + sizeof(struct arphdr);
1386 sha = arpptr;
1387 arpptr += dev->addr_len; /* sha */
1388 memcpy(&sip, arpptr, sizeof(sip));
1389 arpptr += sizeof(sip);
1390 arpptr += dev->addr_len; /* tha */
1391 memcpy(&tip, arpptr, sizeof(tip));
1392
1393 if (ipv4_is_loopback(tip) ||
1394 ipv4_is_multicast(tip))
1395 goto out;
1396
1397 n = neigh_lookup(&arp_tbl, &tip, dev);
1398
1399 if (n) {
e4f67add
DS		 1400 struct vxlan_fdb *f;
1401 struct sk_buff *reply;
1402
1403 if (!(n->nud_state & NUD_CONNECTED)) {
1404 neigh_release(n);
1405 goto out;
1406 }
1407
1408 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1409 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1410 /* bridge-local neighbor */
1411 neigh_release(n);
1412 goto out;
1413 }
1414
1415 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1416 n->ha, sha);
1417
1418 neigh_release(n);
1419
7346135d
DS		 1420 if (reply == NULL)
1421 goto out;
1422
e4f67add
DS		 1423 skb_reset_mac_header(reply);
1424 __skb_pull(reply, skb_network_offset(reply));
1425 reply->ip_summed = CHECKSUM_UNNECESSARY;
1426 reply->pkt_type = PACKET_HOST;
1427
1428 if (netif_rx_ni(reply) == NET_RX_DROP)
1429 dev->stats.rx_dropped++;
e4c7ed41
CW		 1430 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1431 union vxlan_addr ipa = {
1432 .sin.sin_addr.s_addr = tip,
a45e92a5 1433 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1434 };
1435
1436 vxlan_ip_miss(dev, &ipa);
1437 }
e4f67add
DS		 1438out:
1439 consume_skb(skb);
1440 return NETDEV_TX_OK;
1441}
1442
f564f45c 1443#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1444static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1445 struct neighbour *n, bool isrouter)
1446{
1447 struct net_device *dev = request->dev;
1448 struct sk_buff *reply;
1449 struct nd_msg *ns, *na;
1450 struct ipv6hdr *pip6;
1451 u8 *daddr;
1452 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1453 int ns_olen;
1454 int i, len;
1455
1456 if (dev == NULL)
1457 return NULL;
1458
1459 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1460 sizeof(*na) + na_olen + dev->needed_tailroom;
1461 reply = alloc_skb(len, GFP_ATOMIC);
1462 if (reply == NULL)
1463 return NULL;
1464
1465 reply->protocol = htons(ETH_P_IPV6);
1466 reply->dev = dev;
1467 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1468 skb_push(reply, sizeof(struct ethhdr));
1469 skb_set_mac_header(reply, 0);
1470
1471 ns = (struct nd_msg *)skb_transport_header(request);
1472
1473 daddr = eth_hdr(request)->h_source;
1474 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1475 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1476 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1477 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1478 break;
1479 }
1480 }
1481
1482 /* Ethernet header */
1483 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1484 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1485 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1486 reply->protocol = htons(ETH_P_IPV6);
1487
1488 skb_pull(reply, sizeof(struct ethhdr));
1489 skb_set_network_header(reply, 0);
1490 skb_put(reply, sizeof(struct ipv6hdr));
1491
1492 /* IPv6 header */
1493
1494 pip6 = ipv6_hdr(reply);
1495 memset(pip6, 0, sizeof(struct ipv6hdr));
1496 pip6->version = 6;
1497 pip6->priority = ipv6_hdr(request)->priority;
1498 pip6->nexthdr = IPPROTO_ICMPV6;
1499 pip6->hop_limit = 255;
1500 pip6->daddr = ipv6_hdr(request)->saddr;
1501 pip6->saddr = *(struct in6_addr *)n->primary_key;
1502
1503 skb_pull(reply, sizeof(struct ipv6hdr));
1504 skb_set_transport_header(reply, 0);
1505
1506 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1507
1508 /* Neighbor Advertisement */
1509 memset(na, 0, sizeof(*na)+na_olen);
1510 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1511 na->icmph.icmp6_router = isrouter;
1512 na->icmph.icmp6_override = 1;
1513 na->icmph.icmp6_solicited = 1;
1514 na->target = ns->target;
1515 ether_addr_copy(&na->opt[2], n->ha);
1516 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1517 na->opt[1] = na_olen >> 3;
1518
1519 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1520 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1521 csum_partial(na, sizeof(*na)+na_olen, 0));
1522
1523 pip6->payload_len = htons(sizeof(*na)+na_olen);
1524
1525 skb_push(reply, sizeof(struct ipv6hdr));
1526
1527 reply->ip_summed = CHECKSUM_UNNECESSARY;
1528
1529 return reply;
1530}
1531
f564f45c
CW		 1532static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1533{
1534 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1535 struct nd_msg *msg;
f564f45c
CW		 1536 const struct ipv6hdr *iphdr;
1537 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1538 struct neighbour *n;
1539 struct inet6_dev *in6_dev;
f564f45c
CW		 1540
1541 in6_dev = __in6_dev_get(dev);
1542 if (!in6_dev)
1543 goto out;
1544
f564f45c
CW		 1545 iphdr = ipv6_hdr(skb);
1546 saddr = &iphdr->saddr;
1547 daddr = &iphdr->daddr;
1548
f564f45c
CW		 1549 msg = (struct nd_msg *)skb_transport_header(skb);
1550 if (msg->icmph.icmp6_code != 0 ||
1551 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1552 goto out;
1553
4b29dba9
DS		 1554 if (ipv6_addr_loopback(daddr) ||
1555 ipv6_addr_is_multicast(&msg->target))
1556 goto out;
1557
1558 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1559
1560 if (n) {
1561 struct vxlan_fdb *f;
4b29dba9 1562 struct sk_buff *reply;
f564f45c
CW		 1563
1564 if (!(n->nud_state & NUD_CONNECTED)) {
1565 neigh_release(n);
1566 goto out;
1567 }
1568
1569 f = vxlan_find_mac(vxlan, n->ha);
1570 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1571 /* bridge-local neighbor */
1572 neigh_release(n);
1573 goto out;
1574 }
1575
4b29dba9
DS		 1576 reply = vxlan_na_create(skb, n,
1577 !!(f ? f->flags & NTF_ROUTER : 0));
1578
f564f45c 1579 neigh_release(n);
4b29dba9
DS		 1580
1581 if (reply == NULL)
1582 goto out;
1583
1584 if (netif_rx_ni(reply) == NET_RX_DROP)
1585 dev->stats.rx_dropped++;
1586
f564f45c 1587 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1588 union vxlan_addr ipa = {
1589 .sin6.sin6_addr = msg->target,
a45e92a5 1590 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1591 };
1592
f564f45c
CW		 1593 vxlan_ip_miss(dev, &ipa);
1594 }
1595
1596out:
1597 consume_skb(skb);
1598 return NETDEV_TX_OK;
1599}
1600#endif
1601
e4f67add
DS		 1602static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1603{
1604 struct vxlan_dev *vxlan = netdev_priv(dev);
1605 struct neighbour *n;
e4f67add
DS		 1606
1607 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1608 return false;
1609
1610 n = NULL;
1611 switch (ntohs(eth_hdr(skb)->h_proto)) {
1612 case ETH_P_IP:
e15a00aa
CW		 1613 {
1614 struct iphdr *pip;
1615
e4f67add
DS		 1616 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1617 return false;
1618 pip = ip_hdr(skb);
1619 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1620 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1621 union vxlan_addr ipa = {
1622 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1623 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1624 };
1625
1626 vxlan_ip_miss(dev, &ipa);
1627 return false;
1628 }
1629
e4f67add 1630 break;
e15a00aa
CW		 1631 }
1632#if IS_ENABLED(CONFIG_IPV6)
1633 case ETH_P_IPV6:
1634 {
1635 struct ipv6hdr *pip6;
1636
1637 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1638 return false;
1639 pip6 = ipv6_hdr(skb);
1640 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1641 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1642 union vxlan_addr ipa = {
1643 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1644 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1645 };
1646
1647 vxlan_ip_miss(dev, &ipa);
1648 return false;
1649 }
1650
1651 break;
1652 }
1653#endif
e4f67add
DS		 1654 default:
1655 return false;
1656 }
1657
1658 if (n) {
1659 bool diff;
1660
7367d0b5 1661 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1662 if (diff) {
1663 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1664 dev->addr_len);
1665 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1666 }
1667 neigh_release(n);
1668 return diff;
e4c7ed41
CW		 1669 }
1670
e4f67add
DS		 1671 return false;
1672}
1673
af33c1ad 1674static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1675 struct vxlan_metadata *md)
1676{
1677 struct vxlanhdr_gbp *gbp;
1678
db79a621
TG		 1679 if (!md->gbp)
1680 return;
1681
3511494c
TG		 1682 gbp = (struct vxlanhdr_gbp *)vxh;
1683 vxh->vx_flags |= htonl(VXLAN_HF_GBP);
1684
1685 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1686 gbp->dont_learn = 1;
1687
1688 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1689 gbp->policy_applied = 1;
1690
1691 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1692}
1693
f491e56d
JB		 1694static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1695 int iphdr_len, __be32 vni,
1696 struct vxlan_metadata *md, u32 vxflags,
b4ed5cad 1697 bool udp_sum)
e4c7ed41 1698{
e4c7ed41 1699 struct vxlanhdr *vxh;
e4c7ed41
CW		 1700 int min_headroom;
1701 int err;
dfd8645e
TH		 1702 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1703 u16 hdrlen = sizeof(struct vxlanhdr);
1704
af33c1ad 1705 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1706 skb->ip_summed == CHECKSUM_PARTIAL) {
1707 int csum_start = skb_checksum_start_offset(skb);
1708
1709 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1710 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1711 (skb->csum_offset == offsetof(struct udphdr, check) ||
b5708501 1712 skb->csum_offset == offsetof(struct tcphdr, check)))
dfd8645e 1713 type |= SKB_GSO_TUNNEL_REMCSUM;
dfd8645e 1714 }
e4c7ed41 1715
e4c7ed41 1716 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
f491e56d 1717 + VXLAN_HLEN + iphdr_len
df8a39de 1718 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1719
1720 /* Need space for new headers (invalidates iph ptr) */
1721 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1722 if (unlikely(err)) {
1723 kfree_skb(skb);
649c5b8b 1724 return err;
74f47278 1725 }
649c5b8b 1726
5968250c
JP		 1727 skb = vlan_hwaccel_push_inside(skb);
1728 if (WARN_ON(!skb))
1729 return -ENOMEM;
1eaa8178 1730
6fa79666 1731 skb = iptunnel_handle_offloads(skb, type);
b736a623
JG		 1732 if (IS_ERR(skb))
1733 return PTR_ERR(skb);
1734
49560532 1735 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1736 vxh->vx_flags = htonl(VXLAN_HF_VNI);
614732ea 1737 vxh->vx_vni = vni;
49560532 1738
dfd8645e
TH		 1739 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1740 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1741 VXLAN_RCO_SHIFT;
1742
1743 if (skb->csum_offset == offsetof(struct udphdr, check))
1744 data |= VXLAN_RCO_UDP;
1745
1746 vxh->vx_vni |= htonl(data);
1747 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1748
1749 if (!skb_is_gso(skb)) {
1750 skb->ip_summed = CHECKSUM_NONE;
1751 skb->encapsulation = 0;
1752 }
1753 }
1754
af33c1ad
TH		 1755 if (vxflags & VXLAN_F_GBP)
1756 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1757
996c9fd1 1758 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
039f5062 1759 return 0;
49560532 1760}
49560532 1761
1a8496ba
JB		 1762static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan,
1763 struct sk_buff *skb, int oif, u8 tos,
0c1d70af
PA		 1764 __be32 daddr, __be32 *saddr,
1765 struct dst_cache *dst_cache,
1766 struct ip_tunnel_info *info)
1a8496ba
JB		 1767{
1768 struct rtable *rt = NULL;
0c1d70af 1769 bool use_cache = false;
1a8496ba
JB		 1770 struct flowi4 fl4;
1771
0c1d70af
PA		 1772 /* when the ip_tunnel_info is availble, the tos used for lookup is
1773 * packet independent, so we can use the cache
1774 */
d71785ff 1775 if (!skb->mark && (!tos || info)) {
0c1d70af
PA		 1776 use_cache = true;
1777 rt = dst_cache_get_ip4(dst_cache, saddr);
1778 if (rt)
1779 return rt;
1780 }
1781
1a8496ba
JB		 1782 memset(&fl4, 0, sizeof(fl4));
1783 fl4.flowi4_oif = oif;
1784 fl4.flowi4_tos = RT_TOS(tos);
1785 fl4.flowi4_mark = skb->mark;
1786 fl4.flowi4_proto = IPPROTO_UDP;
1787 fl4.daddr = daddr;
1788 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
1789
1790 rt = ip_route_output_key(vxlan->net, &fl4);
0c1d70af 1791 if (!IS_ERR(rt)) {
1a8496ba 1792 *saddr = fl4.saddr;
0c1d70af
PA		 1793 if (use_cache)
1794 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
1795 }
1a8496ba
JB		 1796 return rt;
1797}
1798
e5d4b29f
JB		 1799#if IS_ENABLED(CONFIG_IPV6)
1800static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1801 struct sk_buff *skb, int oif,
1802 const struct in6_addr *daddr,
0c1d70af
PA		 1803 struct in6_addr *saddr,
1804 struct dst_cache *dst_cache)
e5d4b29f
JB		 1805{
1806 struct dst_entry *ndst;
1807 struct flowi6 fl6;
1808 int err;
1809
d71785ff 1810 if (!skb->mark) {
0c1d70af
PA		 1811 ndst = dst_cache_get_ip6(dst_cache, saddr);
1812 if (ndst)
1813 return ndst;
1814 }
1815
e5d4b29f
JB		 1816 memset(&fl6, 0, sizeof(fl6));
1817 fl6.flowi6_oif = oif;
1818 fl6.daddr = *daddr;
1819 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
1820 fl6.flowi6_mark = skb->mark;
1821 fl6.flowi6_proto = IPPROTO_UDP;
1822
1823 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1824 vxlan->vn6_sock->sock->sk,
1825 &ndst, &fl6);
1826 if (err < 0)
1827 return ERR_PTR(err);
1828
1829 *saddr = fl6.saddr;
d71785ff 1830 if (!skb->mark)
0c1d70af 1831 dst_cache_set_ip6(dst_cache, ndst, saddr);
e5d4b29f
JB		 1832 return ndst;
1833}
1834#endif
1835
9dcc71e1
SS		 1836/* Bypass encapsulation if the destination is local */
1837static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1838 struct vxlan_dev *dst_vxlan)
1839{
8f84985f 1840 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1841 union vxlan_addr loopback;
1842 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1843 struct net_device *dev = skb->dev;
1844 int len = skb->len;
9dcc71e1 1845
8f84985f
LR		 1846 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1847 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1848 skb->pkt_type = PACKET_HOST;
1849 skb->encapsulation = 0;
1850 skb->dev = dst_vxlan->dev;
1851 __skb_pull(skb, skb_network_offset(skb));
1852
e4c7ed41
CW		 1853 if (remote_ip->sa.sa_family == AF_INET) {
1854 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1855 loopback.sa.sa_family = AF_INET;
1856#if IS_ENABLED(CONFIG_IPV6)
1857 } else {
1858 loopback.sin6.sin6_addr = in6addr_loopback;
1859 loopback.sa.sa_family = AF_INET6;
1860#endif
1861 }
1862
9dcc71e1 1863 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1864 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1865
1866 u64_stats_update_begin(&tx_stats->syncp);
1867 tx_stats->tx_packets++;
ce6502a8 1868 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1869 u64_stats_update_end(&tx_stats->syncp);
1870
1871 if (netif_rx(skb) == NET_RX_SUCCESS) {
1872 u64_stats_update_begin(&rx_stats->syncp);
1873 rx_stats->rx_packets++;
ce6502a8 1874 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1875 u64_stats_update_end(&rx_stats->syncp);
1876 } else {
ce6502a8 1877 dev->stats.rx_dropped++;
9dcc71e1
SS		 1878 }
1879}
1880
4ad16930
SH		 1881static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1882 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1883{
d71785ff 1884 struct dst_cache *dst_cache;
3093fbe7 1885 struct ip_tunnel_info *info;
d342894c 1886 struct vxlan_dev *vxlan = netdev_priv(dev);
b1be00a6 1887 struct sock *sk;
e4c7ed41 1888 struct rtable *rt = NULL;
d342894c 1889 const struct iphdr *old_iph;
e4c7ed41 1890 union vxlan_addr *dst;
ee122c79
TG		 1891 union vxlan_addr remote_ip;
1892 struct vxlan_metadata _md;
1893 struct vxlan_metadata *md = &_md;
e4c7ed41 1894 __be16 src_port = 0, dst_port;
234f5b73 1895 u32 vni;
d342894c 1896 __be16 df = 0;
1897 __u8 tos, ttl;
0e6fbc5b 1898 int err;
ee122c79 1899 u32 flags = vxlan->flags;
b4ed5cad 1900 bool udp_sum = false;
f491e56d 1901 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
d342894c 1902
61adedf3 1903 info = skb_tunnel_info(skb);
3093fbe7 1904
ee122c79 1905 if (rdst) {
0dfbdf41 1906 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
ee122c79
TG		 1907 vni = rdst->remote_vni;
1908 dst = &rdst->remote_ip;
d71785ff 1909 dst_cache = &rdst->dst_cache;
ee122c79
TG		 1910 } else {
1911 if (!info) {
1912 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1913 dev->name);
1914 goto drop;
1915 }
0dfbdf41 1916 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
ee122c79 1917 vni = be64_to_cpu(info->key.tun_id);
b1be00a6
JB		 1918 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
1919 if (remote_ip.sa.sa_family == AF_INET)
a725e514
JB		 1920 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1921 else
1922 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
ee122c79 1923 dst = &remote_ip;
d71785ff 1924 dst_cache = &info->dst_cache;
ee122c79 1925 }
e4f67add 1926
e4c7ed41 1927 if (vxlan_addr_any(dst)) {
e4f67add 1928 if (did_rsc) {
e4f67add 1929 /* short-circuited back to local bridge */
9dcc71e1 1930 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1931 return;
e4f67add 1932 }
ef59febe 1933 goto drop;
e4f67add 1934 }
ef59febe 1935
d342894c 1936 old_iph = ip_hdr(skb);
1937
0dfbdf41 1938 ttl = vxlan->cfg.ttl;
e4c7ed41 1939 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1940 ttl = 1;
1941
0dfbdf41 1942 tos = vxlan->cfg.tos;
d342894c 1943 if (tos == 1)
206aaafc 1944 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1945
0dfbdf41
TG		 1946 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1947 vxlan->cfg.port_max, true);
d342894c 1948
a725e514 1949 if (info) {
a725e514
JB		 1950 ttl = info->key.ttl;
1951 tos = info->key.tos;
b4ed5cad 1952 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
a725e514
JB		 1953
1954 if (info->options_len)
4c222798 1955 md = ip_tunnel_info_opts(info);
a725e514
JB		 1956 } else {
1957 md->gbp = skb->mark;
1958 }
1959
e4c7ed41 1960 if (dst->sa.sa_family == AF_INET) {
1a8496ba
JB		 1961 __be32 saddr;
1962
b1be00a6
JB		 1963 if (!vxlan->vn4_sock)
1964 goto drop;
1965 sk = vxlan->vn4_sock->sock->sk;
1966
35e2d115
JG		 1967 if (info) {
1968 if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
1969 df = htons(IP_DF);
b4ed5cad
JB		 1970 } else {
1971 udp_sum = !!(flags & VXLAN_F_UDP_CSUM);
35e2d115 1972 }
ee122c79 1973
1a8496ba
JB		 1974 rt = vxlan_get_route(vxlan, skb,
1975 rdst ? rdst->remote_ifindex : 0, tos,
0c1d70af 1976 dst->sin.sin_addr.s_addr, &saddr,
d71785ff 1977 dst_cache, info);
e4c7ed41
CW		 1978 if (IS_ERR(rt)) {
1979 netdev_dbg(dev, "no route to %pI4\n",
1980 &dst->sin.sin_addr.s_addr);
1981 dev->stats.tx_carrier_errors++;
1982 goto tx_error;
1983 }
d342894c 1984
e4c7ed41
CW		 1985 if (rt->dst.dev == dev) {
1986 netdev_dbg(dev, "circular route to %pI4\n",
1987 &dst->sin.sin_addr.s_addr);
1988 dev->stats.collisions++;
fffc15a5 1989 goto rt_tx_error;
e4c7ed41
CW		 1990 }
1991
1992 /* Bypass encapsulation if the destination is local */
1993 if (rt->rt_flags & RTCF_LOCAL &&
1994 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1995 struct vxlan_dev *dst_vxlan;
1996
1997 ip_rt_put(rt);
19ca9fc1 1998 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 1999 dst->sa.sa_family, dst_port,
2000 vxlan->flags);
e4c7ed41
CW		 2001 if (!dst_vxlan)
2002 goto tx_error;
2003 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2004 return;
2005 }
2006
2007 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2008 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
f491e56d
JB		 2009 err = vxlan_build_skb(skb, &rt->dst, sizeof(struct iphdr),
2010 htonl(vni << 8), md, flags, udp_sum);
2011 if (err < 0)
2012 goto xmit_tx_error;
2013
2014 udp_tunnel_xmit_skb(rt, sk, skb, saddr,
2015 dst->sin.sin_addr.s_addr, tos, ttl, df,
2016 src_port, dst_port, xnet, !udp_sum);
e4c7ed41
CW		 2017#if IS_ENABLED(CONFIG_IPV6)
2018 } else {
e4c7ed41 2019 struct dst_entry *ndst;
e5d4b29f 2020 struct in6_addr saddr;
6f264e47 2021 u32 rt6i_flags;
e4c7ed41 2022
b1be00a6
JB		 2023 if (!vxlan->vn6_sock)
2024 goto drop;
2025 sk = vxlan->vn6_sock->sock->sk;
2026
e5d4b29f
JB		 2027 ndst = vxlan6_get_route(vxlan, skb,
2028 rdst ? rdst->remote_ifindex : 0,
0c1d70af 2029 &dst->sin6.sin6_addr, &saddr,
d71785ff 2030 dst_cache);
e5d4b29f 2031 if (IS_ERR(ndst)) {
e4c7ed41
CW		 2032 netdev_dbg(dev, "no route to %pI6\n",
2033 &dst->sin6.sin6_addr);
2034 dev->stats.tx_carrier_errors++;
9dcc71e1 2035 goto tx_error;
e4c7ed41 2036 }
d342894c 2037
e4c7ed41
CW		 2038 if (ndst->dev == dev) {
2039 netdev_dbg(dev, "circular route to %pI6\n",
2040 &dst->sin6.sin6_addr);
2041 dst_release(ndst);
2042 dev->stats.collisions++;
2043 goto tx_error;
2044 }
0e6fbc5b 2045
e4c7ed41 2046 /* Bypass encapsulation if the destination is local */
6f264e47
JB		 2047 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2048 if (rt6i_flags & RTF_LOCAL &&
2049 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
e4c7ed41
CW		 2050 struct vxlan_dev *dst_vxlan;
2051
2052 dst_release(ndst);
19ca9fc1 2053 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2054 dst->sa.sa_family, dst_port,
2055 vxlan->flags);
e4c7ed41
CW		 2056 if (!dst_vxlan)
2057 goto tx_error;
2058 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2059 return;
2060 }
49560532 2061
b4ed5cad
JB		 2062 if (!info)
2063 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
35e2d115 2064
e4c7ed41 2065 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f491e56d
JB		 2066 skb_scrub_packet(skb, xnet);
2067 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2068 htonl(vni << 8), md, flags, udp_sum);
2069 if (err < 0) {
2070 dst_release(ndst);
2071 return;
2072 }
2073 udp_tunnel6_xmit_skb(ndst, sk, skb, dev,
2074 &saddr, &dst->sin6.sin6_addr,
2075 0, ttl, src_port, dst_port, !udp_sum);
e4c7ed41
CW		 2076#endif
2077 }
0e6fbc5b 2078
4ad16930 2079 return;
d342894c 2080
2081drop:
2082 dev->stats.tx_dropped++;
2083 goto tx_free;
2084
f491e56d
JB		 2085xmit_tx_error:
2086 /* skb is already freed. */
2087 skb = NULL;
49560532
PS		 2088rt_tx_error:
2089 ip_rt_put(rt);
d342894c 2090tx_error:
2091 dev->stats.tx_errors++;
2092tx_free:
2093 dev_kfree_skb(skb);
d342894c 2094}
2095
6681712d
DS		 2096/* Transmit local packets over Vxlan
2097 *
2098 * Outer IP header inherits ECN and DF from inner header.
2099 * Outer UDP destination is the VXLAN assigned port.
2100 * source port is based on hash of flow
2101 */
2102static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2103{
2104 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2105 const struct ip_tunnel_info *info;
6681712d
DS		 2106 struct ethhdr *eth;
2107 bool did_rsc = false;
8f646c92 2108 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2109 struct vxlan_fdb *f;
6681712d 2110
61adedf3 2111 info = skb_tunnel_info(skb);
3093fbe7 2112
6681712d
DS		 2113 skb_reset_mac_header(skb);
2114 eth = eth_hdr(skb);
2115
f564f45c
CW		 2116 if ((vxlan->flags & VXLAN_F_PROXY)) {
2117 if (ntohs(eth->h_proto) == ETH_P_ARP)
2118 return arp_reduce(dev, skb);
2119#if IS_ENABLED(CONFIG_IPV6)
2120 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 2121 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2122 + sizeof(struct nd_msg)) &&
f564f45c
CW		 2123 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2124 struct nd_msg *msg;
2125
2126 msg = (struct nd_msg *)skb_transport_header(skb);
2127 if (msg->icmph.icmp6_code == 0 &&
2128 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2129 return neigh_reduce(dev, skb);
2130 }
7a9f526f 2131 eth = eth_hdr(skb);
f564f45c
CW		 2132#endif
2133 }
6681712d 2134
da8b43c0 2135 if (vxlan->flags & VXLAN_F_COLLECT_METADATA &&
46fa062a 2136 info && info->mode & IP_TUNNEL_INFO_TX) {
ee122c79
TG		 2137 vxlan_xmit_one(skb, dev, NULL, false);
2138 return NETDEV_TX_OK;
2139 }
2140
6681712d 2141 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 2142 did_rsc = false;
2143
2144 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2145 (ntohs(eth->h_proto) == ETH_P_IP ||
2146 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2147 did_rsc = route_shortcircuit(dev, skb);
2148 if (did_rsc)
2149 f = vxlan_find_mac(vxlan, eth->h_dest);
2150 }
2151
6681712d 2152 if (f == NULL) {
afbd8bae
MR		 2153 f = vxlan_find_mac(vxlan, all_zeros_mac);
2154 if (f == NULL) {
2155 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2156 !is_multicast_ether_addr(eth->h_dest))
2157 vxlan_fdb_miss(vxlan, eth->h_dest);
2158
2159 dev->stats.tx_dropped++;
8f646c92 2160 kfree_skb(skb);
afbd8bae
MR		 2161 return NETDEV_TX_OK;
2162 }
2163 }
6681712d 2164
afbd8bae
MR		 2165 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2166 struct sk_buff *skb1;
6681712d 2167
8f646c92
ED		 2168 if (!fdst) {
2169 fdst = rdst;
2170 continue;
2171 }
afbd8bae
MR		 2172 skb1 = skb_clone(skb, GFP_ATOMIC);
2173 if (skb1)
2174 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 2175 }
2176
8f646c92
ED		 2177 if (fdst)
2178 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2179 else
2180 kfree_skb(skb);
4ad16930 2181 return NETDEV_TX_OK;
6681712d
DS		 2182}
2183
d342894c 2184/* Walk the forwarding table and purge stale entries */
2185static void vxlan_cleanup(unsigned long arg)
2186{
2187 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2188 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2189 unsigned int h;
2190
2191 if (!netif_running(vxlan->dev))
2192 return;
2193
d342894c 2194 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2195 struct hlist_node *p, *n;
14e1d0fa
SD		 2196
2197 spin_lock_bh(&vxlan->hash_lock);
d342894c 2198 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2199 struct vxlan_fdb *f
2200 = container_of(p, struct vxlan_fdb, hlist);
2201 unsigned long timeout;
2202
3c172868 2203 if (f->state & NUD_PERMANENT)
d342894c 2204 continue;
2205
0dfbdf41 2206 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2207 if (time_before_eq(timeout, jiffies)) {
2208 netdev_dbg(vxlan->dev,
2209 "garbage collect %pM\n",
2210 f->eth_addr);
2211 f->state = NUD_STALE;
2212 vxlan_fdb_destroy(vxlan, f);
2213 } else if (time_before(timeout, next_timer))
2214 next_timer = timeout;
2215 }
14e1d0fa 2216 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2217 }
d342894c 2218
2219 mod_timer(&vxlan->age_timer, next_timer);
2220}
2221
9c2e24e1
PS		 2222static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2223{
56ef9c90 2224 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
9c2e24e1
PS		 2225 __u32 vni = vxlan->default_dst.remote_vni;
2226
56ef9c90 2227 spin_lock(&vn->sock_lock);
9c2e24e1 2228 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2229 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2230}
2231
d342894c 2232/* Setup stats when device is created */
2233static int vxlan_init(struct net_device *dev)
2234{
1c213bd2 2235 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2236 if (!dev->tstats)
d342894c 2237 return -ENOMEM;
2238
2239 return 0;
2240}
2241
ba609e9b 2242static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2243{
2244 struct vxlan_fdb *f;
2245
2246 spin_lock_bh(&vxlan->hash_lock);
2247 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2248 if (f)
2249 vxlan_fdb_destroy(vxlan, f);
2250 spin_unlock_bh(&vxlan->hash_lock);
2251}
2252
ebf4063e
SH		 2253static void vxlan_uninit(struct net_device *dev)
2254{
2255 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2256
ba609e9b 2257 vxlan_fdb_delete_default(vxlan);
afbd8bae 2258
ebf4063e
SH		 2259 free_percpu(dev->tstats);
2260}
2261
d342894c 2262/* Start ageing timer and join group when device is brought up */
2263static int vxlan_open(struct net_device *dev)
2264{
2265 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2266 int ret;
56ef9c90 2267
205f356d
JB		 2268 ret = vxlan_sock_add(vxlan);
2269 if (ret < 0)
2270 return ret;
d342894c 2271
79d4a94f 2272 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2273 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2274 if (ret == -EADDRINUSE)
2275 ret = 0;
56ef9c90 2276 if (ret) {
205f356d 2277 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2278 return ret;
2279 }
d342894c 2280 }
2281
0dfbdf41 2282 if (vxlan->cfg.age_interval)
d342894c 2283 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2284
56ef9c90 2285 return ret;
d342894c 2286}
2287
2288/* Purge the forwarding table */
2289static void vxlan_flush(struct vxlan_dev *vxlan)
2290{
31fec5aa 2291 unsigned int h;
d342894c 2292
2293 spin_lock_bh(&vxlan->hash_lock);
2294 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2295 struct hlist_node *p, *n;
2296 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2297 struct vxlan_fdb *f
2298 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2299 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2300 if (!is_zero_ether_addr(f->eth_addr))
2301 vxlan_fdb_destroy(vxlan, f);
d342894c 2302 }
2303 }
2304 spin_unlock_bh(&vxlan->hash_lock);
2305}
2306
2307/* Cleanup timer and forwarding table on shutdown */
2308static int vxlan_stop(struct net_device *dev)
2309{
2310 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2311 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2312 int ret = 0;
d342894c 2313
24c0e683 2314 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2315 !vxlan_group_used(vn, vxlan))
56ef9c90 2316 ret = vxlan_igmp_leave(vxlan);
d342894c 2317
2318 del_timer_sync(&vxlan->age_timer);
2319
2320 vxlan_flush(vxlan);
205f356d 2321 vxlan_sock_release(vxlan);
d342894c 2322
56ef9c90 2323 return ret;
d342894c 2324}
2325
d342894c 2326/* Stub, nothing needs to be done. */
2327static void vxlan_set_multicast_list(struct net_device *dev)
2328{
2329}
2330
345010b5
DB		 2331static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2332{
2333 struct vxlan_dev *vxlan = netdev_priv(dev);
2334 struct vxlan_rdst *dst = &vxlan->default_dst;
2335 struct net_device *lowerdev;
2336 int max_mtu;
2337
f01ec1c0 2338 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
345010b5
DB		 2339 if (lowerdev == NULL)
2340 return eth_change_mtu(dev, new_mtu);
2341
2342 if (dst->remote_ip.sa.sa_family == AF_INET6)
2343 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2344 else
2345 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2346
2347 if (new_mtu < 68 || new_mtu > max_mtu)
2348 return -EINVAL;
2349
2350 dev->mtu = new_mtu;
2351 return 0;
2352}
2353
fc4099f1
PS		 2354static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2355{
2356 struct vxlan_dev *vxlan = netdev_priv(dev);
2357 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2358 __be16 sport, dport;
2359
2360 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2361 vxlan->cfg.port_max, true);
2362 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2363
239e944f 2364 if (ip_tunnel_info_af(info) == AF_INET) {
1a8496ba
JB		 2365 struct rtable *rt;
2366
239e944f
JB		 2367 if (!vxlan->vn4_sock)
2368 return -EINVAL;
1a8496ba
JB		 2369 rt = vxlan_get_route(vxlan, skb, 0, info->key.tos,
2370 info->key.u.ipv4.dst,
0c1d70af 2371 &info->key.u.ipv4.src, NULL, info);
1a8496ba
JB		 2372 if (IS_ERR(rt))
2373 return PTR_ERR(rt);
2374 ip_rt_put(rt);
239e944f
JB		 2375 } else {
2376#if IS_ENABLED(CONFIG_IPV6)
2377 struct dst_entry *ndst;
2378
2379 if (!vxlan->vn6_sock)
2380 return -EINVAL;
2381 ndst = vxlan6_get_route(vxlan, skb, 0,
2382 &info->key.u.ipv6.dst,
0c1d70af 2383 &info->key.u.ipv6.src, NULL);
239e944f
JB		 2384 if (IS_ERR(ndst))
2385 return PTR_ERR(ndst);
2386 dst_release(ndst);
239e944f
JB		 2387#else /* !CONFIG_IPV6 */
2388 return -EPFNOSUPPORT;
2389#endif
2390 }
1a8496ba
JB		 2391 info->key.tp_src = sport;
2392 info->key.tp_dst = dport;
239e944f 2393 return 0;
fc4099f1
PS		 2394}
2395
d342894c 2396static const struct net_device_ops vxlan_netdev_ops = {
2397 .ndo_init = vxlan_init,
ebf4063e 2398 .ndo_uninit = vxlan_uninit,
d342894c 2399 .ndo_open = vxlan_open,
2400 .ndo_stop = vxlan_stop,
2401 .ndo_start_xmit = vxlan_xmit,
e8171045 2402 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2403 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2404 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2405 .ndo_validate_addr = eth_validate_addr,
2406 .ndo_set_mac_address = eth_mac_addr,
2407 .ndo_fdb_add = vxlan_fdb_add,
2408 .ndo_fdb_del = vxlan_fdb_delete,
2409 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2410 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2411};
2412
2413/* Info for udev, that this is a virtual tunnel endpoint */
2414static struct device_type vxlan_type = {
2415 .name = "vxlan",
2416};
2417
53cf5275 2418/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2419 * supply the listening VXLAN udp ports. Callers are expected
2420 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2421 */
2422void vxlan_get_rx_port(struct net_device *dev)
2423{
2424 struct vxlan_sock *vs;
2425 struct net *net = dev_net(dev);
2426 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2427 sa_family_t sa_family;
35e42379
JG		 2428 __be16 port;
2429 unsigned int i;
53cf5275
JG		 2430
2431 spin_lock(&vn->sock_lock);
2432 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2433 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2434 port = inet_sk(vs->sock->sk)->inet_sport;
705cc62f 2435 sa_family = vxlan_get_sk_family(vs);
53cf5275
JG		 2436 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2437 port);
2438 }
2439 }
2440 spin_unlock(&vn->sock_lock);
2441}
2442EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2443
d342894c 2444/* Initialize the device structure. */
2445static void vxlan_setup(struct net_device *dev)
2446{
2447 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2448 unsigned int h;
d342894c 2449
2450 eth_hw_addr_random(dev);
2451 ether_setup(dev);
2452
2453 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2454 dev->destructor = free_netdev;
d342894c 2455 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2456
d342894c 2457 dev->features |= NETIF_F_LLTX;
d6727fe3 2458 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2459 dev->features |= NETIF_F_RXCSUM;
05c0db08 2460 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2461
1eaa8178
PS		 2462 dev->vlan_features = dev->features;
2463 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2464 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2465 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2466 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2467 netif_keep_dst(dev);
22dba393 2468 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
d342894c 2469
553675fb 2470 INIT_LIST_HEAD(&vxlan->next);
d342894c 2471 spin_lock_init(&vxlan->hash_lock);
2472
2473 init_timer_deferrable(&vxlan->age_timer);
2474 vxlan->age_timer.function = vxlan_cleanup;
2475 vxlan->age_timer.data = (unsigned long) vxlan;
2476
0dfbdf41 2477 vxlan->cfg.dst_port = htons(vxlan_port);
05f47d69 2478
d342894c 2479 vxlan->dev = dev;
2480
58ce31cc
TH		 2481 gro_cells_init(&vxlan->gro_cells, dev);
2482
d342894c 2483 for (h = 0; h < FDB_HASH_SIZE; ++h)
2484 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2485}
2486
2487static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2488 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2489 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2490 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2491 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2492 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2493 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2494 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2495 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2496 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2497 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2498 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2499 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2500 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2501 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2502 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2503 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2504 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2505 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2506 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2507 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2508 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2509 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2510 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2511 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
0ace2ca8 2512 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2513};
2514
2515static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2516{
2517 if (tb[IFLA_ADDRESS]) {
2518 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2519 pr_debug("invalid link address (not ethernet)\n");
2520 return -EINVAL;
2521 }
2522
2523 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2524 pr_debug("invalid all zero ethernet address\n");
2525 return -EADDRNOTAVAIL;
2526 }
2527 }
2528
2529 if (!data)
2530 return -EINVAL;
2531
2532 if (data[IFLA_VXLAN_ID]) {
2533 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2534 if (id >= VXLAN_VID_MASK)
2535 return -ERANGE;
2536 }
2537
05f47d69 2538 if (data[IFLA_VXLAN_PORT_RANGE]) {
2539 const struct ifla_vxlan_port_range *p
2540 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2541
2542 if (ntohs(p->high) < ntohs(p->low)) {
2543 pr_debug("port range %u .. %u not valid\n",
2544 ntohs(p->low), ntohs(p->high));
2545 return -EINVAL;
2546 }
2547 }
2548
d342894c 2549 return 0;
2550}
2551
1b13c97f
YB		 2552static void vxlan_get_drvinfo(struct net_device *netdev,
2553 struct ethtool_drvinfo *drvinfo)
2554{
2555 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2556 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2557}
2558
2559static const struct ethtool_ops vxlan_ethtool_ops = {
2560 .get_drvinfo = vxlan_get_drvinfo,
2561 .get_link = ethtool_op_get_link,
2562};
2563
553675fb 2564static void vxlan_del_work(struct work_struct *work)
2565{
2566 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
acbf74a7 2567 udp_tunnel_sock_release(vs->sock);
553675fb 2568 kfree_rcu(vs, rcu);
2569}
2570
3ee64f39
TH		 2571static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2572 __be16 port, u32 flags)
553675fb 2573{
e4c7ed41 2574 struct socket *sock;
3ee64f39
TH		 2575 struct udp_port_cfg udp_conf;
2576 int err;
e4c7ed41 2577
3ee64f39 2578 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2579
3ee64f39
TH		 2580 if (ipv6) {
2581 udp_conf.family = AF_INET6;
3ee64f39 2582 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2583 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2584 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2585 } else {
2586 udp_conf.family = AF_INET;
e4c7ed41 2587 }
e4c7ed41 2588
3ee64f39 2589 udp_conf.local_udp_port = port;
553675fb 2590
3ee64f39
TH		 2591 /* Open UDP socket */
2592 err = udp_sock_create(net, &udp_conf, &sock);
2593 if (err < 0)
2594 return ERR_PTR(err);
e4c7ed41 2595
39deb2c7 2596 return sock;
e4c7ed41
CW		 2597}
2598
2599/* Create new listen socket if needed */
b1be00a6
JB		 2600static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2601 __be16 port, u32 flags)
e4c7ed41
CW		 2602{
2603 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2604 struct vxlan_sock *vs;
2605 struct socket *sock;
e4c7ed41 2606 unsigned int h;
acbf74a7 2607 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2608
dc01e7d3 2609 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2610 if (!vs)
2611 return ERR_PTR(-ENOMEM);
2612
2613 for (h = 0; h < VNI_HASH_SIZE; ++h)
2614 INIT_HLIST_HEAD(&vs->vni_list[h]);
2615
2616 INIT_WORK(&vs->del_work, vxlan_del_work);
2617
3ee64f39 2618 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2619 if (IS_ERR(sock)) {
56ef9c90
MRL		 2620 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2621 PTR_ERR(sock));
553675fb 2622 kfree(vs);
e50fddc8 2623 return ERR_CAST(sock);
553675fb 2624 }
e4c7ed41
CW		 2625
2626 vs->sock = sock;
9c2e24e1 2627 atomic_set(&vs->refcnt, 1);
af33c1ad 2628 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2629
dc01e7d3
OG		 2630 /* Initialize the vxlan udp offloads structure */
2631 vs->udp_offloads.port = port;
2632 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2633 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2634
9c2e24e1
PS		 2635 spin_lock(&vn->sock_lock);
2636 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2637 vxlan_notify_add_rx_port(vs);
9c2e24e1 2638 spin_unlock(&vn->sock_lock);
553675fb 2639
2640 /* Mark socket as an encapsulation socket. */
acbf74a7
AZ		 2641 tunnel_cfg.sk_user_data = vs;
2642 tunnel_cfg.encap_type = 1;
2643 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2644 tunnel_cfg.encap_destroy = NULL;
2645
2646 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2647
9c2e24e1
PS		 2648 return vs;
2649}
2650
b1be00a6 2651static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2652{
205f356d
JB		 2653 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2654 struct vxlan_sock *vs = NULL;
9c2e24e1 2655
205f356d 2656 if (!vxlan->cfg.no_share) {
56ef9c90 2657 spin_lock(&vn->sock_lock);
205f356d
JB		 2658 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2659 vxlan->cfg.dst_port, vxlan->flags);
2660 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
56ef9c90 2661 spin_unlock(&vn->sock_lock);
205f356d 2662 return -EBUSY;
56ef9c90
MRL		 2663 }
2664 spin_unlock(&vn->sock_lock);
2665 }
205f356d 2666 if (!vs)
b1be00a6
JB		 2667 vs = vxlan_socket_create(vxlan->net, ipv6,
2668 vxlan->cfg.dst_port, vxlan->flags);
205f356d
JB		 2669 if (IS_ERR(vs))
2670 return PTR_ERR(vs);
b1be00a6
JB		 2671#if IS_ENABLED(CONFIG_IPV6)
2672 if (ipv6)
2673 vxlan->vn6_sock = vs;
2674 else
2675#endif
2676 vxlan->vn4_sock = vs;
205f356d
JB		 2677 vxlan_vs_add_dev(vs, vxlan);
2678 return 0;
553675fb 2679}
2680
b1be00a6
JB		 2681static int vxlan_sock_add(struct vxlan_dev *vxlan)
2682{
2683 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2684 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2685 int ret = 0;
2686
2687 vxlan->vn4_sock = NULL;
2688#if IS_ENABLED(CONFIG_IPV6)
2689 vxlan->vn6_sock = NULL;
2690 if (ipv6 || metadata)
2691 ret = __vxlan_sock_add(vxlan, true);
2692#endif
2693 if (!ret && (!ipv6 || metadata))
2694 ret = __vxlan_sock_add(vxlan, false);
2695 if (ret < 0)
2696 vxlan_sock_release(vxlan);
2697 return ret;
2698}
2699
0dfbdf41
TG		 2700static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2701 struct vxlan_config *conf)
d342894c 2702{
33564bbb 2703 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
07b9b37c 2704 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
c7995c43 2705 struct vxlan_rdst *dst = &vxlan->default_dst;
b1be00a6 2706 unsigned short needed_headroom = ETH_HLEN;
d342894c 2707 int err;
e4c7ed41 2708 bool use_ipv6 = false;
0dfbdf41 2709 __be16 default_port = vxlan->cfg.dst_port;
d342894c 2710
33564bbb 2711 vxlan->net = src_net;
f01ec1c0 2712
0dfbdf41 2713 dst->remote_vni = conf->vni;
d342894c 2714
0dfbdf41 2715 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
d342894c 2716
0dfbdf41
TG		 2717 /* Unless IPv6 is explicitly requested, assume IPv4 */
2718 if (!dst->remote_ip.sa.sa_family)
2719 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41 2720
0dfbdf41 2721 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
057ba29b
JB		 2722 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2723 if (!IS_ENABLED(CONFIG_IPV6))
2724 return -EPFNOSUPPORT;
e4c7ed41 2725 use_ipv6 = true;
b1be00a6 2726 vxlan->flags |= VXLAN_F_IPV6;
057ba29b 2727 }
d342894c 2728
0dfbdf41 2729 if (conf->remote_ifindex) {
34e02aa1 2730 struct net_device *lowerdev
0dfbdf41
TG		 2731 = __dev_get_by_index(src_net, conf->remote_ifindex);
2732
2733 dst->remote_ifindex = conf->remote_ifindex;
34e02aa1 2734
2735 if (!lowerdev) {
c7995c43 2736 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2737 return -ENODEV;
2738 }
d342894c 2739
e4c7ed41
CW		 2740#if IS_ENABLED(CONFIG_IPV6)
2741 if (use_ipv6) {
2742 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2743 if (idev && idev->cnf.disable_ipv6) {
2744 pr_info("IPv6 is disabled via sysctl\n");
2745 return -EPERM;
2746 }
e4c7ed41
CW		 2747 }
2748#endif
2749
0dfbdf41 2750 if (!conf->mtu)
e4c7ed41 2751 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2752
b1be00a6 2753 needed_headroom = lowerdev->hard_header_len;
9dc2ad10 2754 }
d342894c 2755
b1be00a6
JB		 2756 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2757 needed_headroom += VXLAN6_HEADROOM;
2758 else
2759 needed_headroom += VXLAN_HEADROOM;
2760 dev->needed_headroom = needed_headroom;
2761
0dfbdf41
TG		 2762 memcpy(&vxlan->cfg, conf, sizeof(*conf));
2763 if (!vxlan->cfg.dst_port)
2764 vxlan->cfg.dst_port = default_port;
2765 vxlan->flags |= conf->flags;
2766
2767 if (!vxlan->cfg.age_interval)
2768 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2769
07b9b37c
ND		 2770 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2771 if (tmp->cfg.vni == conf->vni &&
2772 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2773 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2774 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2775 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2776 (vxlan->flags & VXLAN_F_RCV_FLAGS))
0dfbdf41 2777 return -EEXIST;
07b9b37c 2778 }
0dfbdf41
TG		 2779
2780 dev->ethtool_ops = &vxlan_ethtool_ops;
2781
2782 /* create an fdb entry for a valid default destination */
2783 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2784 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2785 &vxlan->default_dst.remote_ip,
2786 NUD_REACHABLE|NUD_PERMANENT,
2787 NLM_F_EXCL|NLM_F_CREATE,
2788 vxlan->cfg.dst_port,
2789 vxlan->default_dst.remote_vni,
2790 vxlan->default_dst.remote_ifindex,
2791 NTF_SELF);
2792 if (err)
2793 return err;
2794 }
2795
2796 err = register_netdevice(dev);
2797 if (err) {
2798 vxlan_fdb_delete_default(vxlan);
2799 return err;
2800 }
2801
2802 list_add(&vxlan->next, &vn->vxlan_list);
2803
2804 return 0;
2805}
2806
2807struct net_device *vxlan_dev_create(struct net *net, const char *name,
2808 u8 name_assign_type, struct vxlan_config *conf)
2809{
2810 struct nlattr *tb[IFLA_MAX+1];
2811 struct net_device *dev;
2812 int err;
2813
2814 memset(&tb, 0, sizeof(tb));
2815
2816 dev = rtnl_create_link(net, name, name_assign_type,
2817 &vxlan_link_ops, tb);
2818 if (IS_ERR(dev))
2819 return dev;
2820
2821 err = vxlan_dev_configure(net, dev, conf);
2822 if (err < 0) {
2823 free_netdev(dev);
2824 return ERR_PTR(err);
2825 }
2826
2827 return dev;
2828}
2829EXPORT_SYMBOL_GPL(vxlan_dev_create);
2830
2831static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2832 struct nlattr *tb[], struct nlattr *data[])
2833{
2834 struct vxlan_config conf;
2835 int err;
2836
0dfbdf41 2837 memset(&conf, 0, sizeof(conf));
e277de5f
JG		 2838
2839 if (data[IFLA_VXLAN_ID])
2840 conf.vni = nla_get_u32(data[IFLA_VXLAN_ID]);
0dfbdf41
TG		 2841
2842 if (data[IFLA_VXLAN_GROUP]) {
2843 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2844 } else if (data[IFLA_VXLAN_GROUP6]) {
2845 if (!IS_ENABLED(CONFIG_IPV6))
2846 return -EPFNOSUPPORT;
2847
2848 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2849 conf.remote_ip.sa.sa_family = AF_INET6;
2850 }
2851
2852 if (data[IFLA_VXLAN_LOCAL]) {
2853 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2854 conf.saddr.sa.sa_family = AF_INET;
2855 } else if (data[IFLA_VXLAN_LOCAL6]) {
2856 if (!IS_ENABLED(CONFIG_IPV6))
2857 return -EPFNOSUPPORT;
2858
2859 /* TODO: respect scope id */
2860 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2861 conf.saddr.sa.sa_family = AF_INET6;
2862 }
2863
2864 if (data[IFLA_VXLAN_LINK])
2865 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2866
d342894c 2867 if (data[IFLA_VXLAN_TOS])
0dfbdf41 2868 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 2869
afb97186 2870 if (data[IFLA_VXLAN_TTL])
0dfbdf41 2871 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 2872
d342894c 2873 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
0dfbdf41 2874 conf.flags |= VXLAN_F_LEARN;
d342894c 2875
2876 if (data[IFLA_VXLAN_AGEING])
0dfbdf41 2877 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
d342894c 2878
e4f67add 2879 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
0dfbdf41 2880 conf.flags |= VXLAN_F_PROXY;
e4f67add
DS		 2881
2882 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
0dfbdf41 2883 conf.flags |= VXLAN_F_RSC;
e4f67add
DS		 2884
2885 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
0dfbdf41 2886 conf.flags |= VXLAN_F_L2MISS;
e4f67add
DS		 2887
2888 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
0dfbdf41 2889 conf.flags |= VXLAN_F_L3MISS;
e4f67add 2890
d342894c 2891 if (data[IFLA_VXLAN_LIMIT])
0dfbdf41 2892 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
d342894c 2893
f8a9b1bc
AS		 2894 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
2895 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
2896 conf.flags |= VXLAN_F_COLLECT_METADATA;
2897
05f47d69 2898 if (data[IFLA_VXLAN_PORT_RANGE]) {
2899 const struct ifla_vxlan_port_range *p
2900 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
0dfbdf41
TG		 2901 conf.port_min = ntohs(p->low);
2902 conf.port_max = ntohs(p->high);
05f47d69 2903 }
2904
823aa873 2905 if (data[IFLA_VXLAN_PORT])
0dfbdf41 2906 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
823aa873 2907
359a0ea9 2908 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
0dfbdf41 2909 conf.flags |= VXLAN_F_UDP_CSUM;
359a0ea9
TH		 2910
2911 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2912 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
0dfbdf41 2913 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
359a0ea9
TH		 2914
2915 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2916 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
0dfbdf41 2917 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
359a0ea9 2918
dfd8645e
TH		 2919 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2920 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
0dfbdf41 2921 conf.flags |= VXLAN_F_REMCSUM_TX;
dfd8645e
TH		 2922
2923 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2924 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
0dfbdf41 2925 conf.flags |= VXLAN_F_REMCSUM_RX;
dfd8645e 2926
3511494c 2927 if (data[IFLA_VXLAN_GBP])
0dfbdf41 2928 conf.flags |= VXLAN_F_GBP;
3511494c 2929
0ace2ca8 2930 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
0dfbdf41 2931 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
0ace2ca8 2932
0dfbdf41
TG		 2933 err = vxlan_dev_configure(src_net, dev, &conf);
2934 switch (err) {
2935 case -ENODEV:
2936 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
2937 break;
1b13c97f 2938
0dfbdf41
TG		 2939 case -EPERM:
2940 pr_info("IPv6 is disabled via sysctl\n");
2941 break;
d342894c 2942
0dfbdf41
TG		 2943 case -EEXIST:
2944 pr_info("duplicate VNI %u\n", conf.vni);
2945 break;
afbd8bae
MR		 2946 }
2947
0dfbdf41 2948 return err;
d342894c 2949}
2950
2951static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2952{
2953 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2954 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 2955
fe5c3561 2956 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2957 if (!hlist_unhashed(&vxlan->hlist))
2958 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2959 spin_unlock(&vn->sock_lock);
2960
58ce31cc 2961 gro_cells_destroy(&vxlan->gro_cells);
553675fb 2962 list_del(&vxlan->next);
d342894c 2963 unregister_netdevice_queue(dev, head);
2964}
2965
2966static size_t vxlan_get_size(const struct net_device *dev)
2967{
2968
2969 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2970 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2971 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2972 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2973 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2974 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2975 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 2976 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2977 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2978 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2979 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 2980 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 2981 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2982 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 2983 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 2984 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2985 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2986 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2987 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 2988 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
2989 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 2990 0;
2991}
2992
2993static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2994{
2995 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2996 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 2997 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 2998 .low = htons(vxlan->cfg.port_min),
2999 .high = htons(vxlan->cfg.port_max),
05f47d69 3000 };
d342894c 3001
c7995c43 3002 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 3003 goto nla_put_failure;
3004
e4c7ed41
CW		 3005 if (!vxlan_addr_any(&dst->remote_ip)) {
3006 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3007 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3008 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3009 goto nla_put_failure;
3010#if IS_ENABLED(CONFIG_IPV6)
3011 } else {
930345ea
JB		 3012 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3013 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3014 goto nla_put_failure;
3015#endif
3016 }
3017 }
d342894c 3018
c7995c43 3019 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3020 goto nla_put_failure;
3021
0dfbdf41
TG		 3022 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3023 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3024 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3025 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3026 goto nla_put_failure;
3027#if IS_ENABLED(CONFIG_IPV6)
3028 } else {
930345ea 3029 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3030 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3031 goto nla_put_failure;
3032#endif
3033 }
3034 }
d342894c 3035
0dfbdf41
TG		 3036 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3037 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e4f67add
DS		 3038 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3039 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3040 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3041 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3042 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3043 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3044 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3045 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3046 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
da8b43c0
AS		 3047 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3048 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3049 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3050 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3051 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9
TH		 3052 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
3053 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
3054 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3055 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3056 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 3057 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3058 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3059 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3060 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3061 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 3062 goto nla_put_failure;
3063
05f47d69 3064 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3065 goto nla_put_failure;
3066
3511494c
TG		 3067 if (vxlan->flags & VXLAN_F_GBP &&
3068 nla_put_flag(skb, IFLA_VXLAN_GBP))
3069 goto nla_put_failure;
3070
0ace2ca8
TH		 3071 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3072 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3073 goto nla_put_failure;
3074
d342894c 3075 return 0;
3076
3077nla_put_failure:
3078 return -EMSGSIZE;
3079}
3080
1728d4fa
ND		 3081static struct net *vxlan_get_link_net(const struct net_device *dev)
3082{
3083 struct vxlan_dev *vxlan = netdev_priv(dev);
3084
3085 return vxlan->net;
3086}
3087
d342894c 3088static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3089 .kind = "vxlan",
3090 .maxtype = IFLA_VXLAN_MAX,
3091 .policy = vxlan_policy,
3092 .priv_size = sizeof(struct vxlan_dev),
3093 .setup = vxlan_setup,
3094 .validate = vxlan_validate,
3095 .newlink = vxlan_newlink,
3096 .dellink = vxlan_dellink,
3097 .get_size = vxlan_get_size,
3098 .fill_info = vxlan_fill_info,
1728d4fa 3099 .get_link_net = vxlan_get_link_net,
d342894c 3100};
3101
acaf4e70
DB		 3102static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3103 struct net_device *dev)
3104{
3105 struct vxlan_dev *vxlan, *next;
3106 LIST_HEAD(list_kill);
3107
3108 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3109 struct vxlan_rdst *dst = &vxlan->default_dst;
3110
3111 /* In case we created vxlan device with carrier
3112 * and we loose the carrier due to module unload
3113 * we also need to remove vxlan device. In other
3114 * cases, it's not necessary and remote_ifindex
3115 * is 0 here, so no matches.
3116 */
3117 if (dst->remote_ifindex == dev->ifindex)
3118 vxlan_dellink(vxlan->dev, &list_kill);
3119 }
3120
3121 unregister_netdevice_many(&list_kill);
3122}
3123
3124static int vxlan_lowerdev_event(struct notifier_block *unused,
3125 unsigned long event, void *ptr)
3126{
3127 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3128 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3129
783c1463 3130 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 3131 vxlan_handle_lowerdev_unregister(vn, dev);
3132
3133 return NOTIFY_DONE;
3134}
3135
3136static struct notifier_block vxlan_notifier_block __read_mostly = {
3137 .notifier_call = vxlan_lowerdev_event,
3138};
3139
d342894c 3140static __net_init int vxlan_init_net(struct net *net)
3141{
3142 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3143 unsigned int h;
d342894c 3144
553675fb 3145 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3146 spin_lock_init(&vn->sock_lock);
d342894c 3147
553675fb 3148 for (h = 0; h < PORT_HASH_SIZE; ++h)
3149 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3150
3151 return 0;
3152}
3153
f01ec1c0
ND		 3154static void __net_exit vxlan_exit_net(struct net *net)
3155{
3156 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3157 struct vxlan_dev *vxlan, *next;
3158 struct net_device *dev, *aux;
3159 LIST_HEAD(list);
3160
3161 rtnl_lock();
3162 for_each_netdev_safe(net, dev, aux)
3163 if (dev->rtnl_link_ops == &vxlan_link_ops)
3164 unregister_netdevice_queue(dev, &list);
3165
3166 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3167 /* If vxlan->dev is in the same netns, it has already been added
3168 * to the list by the previous loop.
3169 */
58ce31cc
TH		 3170 if (!net_eq(dev_net(vxlan->dev), net)) {
3171 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3172 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3173 }
f01ec1c0
ND		 3174 }
3175
3176 unregister_netdevice_many(&list);
3177 rtnl_unlock();
3178}
3179
d342894c 3180static struct pernet_operations vxlan_net_ops = {
3181 .init = vxlan_init_net,
f01ec1c0 3182 .exit = vxlan_exit_net,
d342894c 3183 .id = &vxlan_net_id,
3184 .size = sizeof(struct vxlan_net),
3185};
3186
3187static int __init vxlan_init_module(void)
3188{
3189 int rc;
3190
758c57d1
SH		 3191 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3192 if (!vxlan_wq)
3193 return -ENOMEM;
3194
d342894c 3195 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3196
783c1463 3197 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3198 if (rc)
3199 goto out1;
3200
acaf4e70 3201 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3202 if (rc)
3203 goto out2;
3204
acaf4e70
DB		 3205 rc = rtnl_link_register(&vxlan_link_ops);
3206 if (rc)
3207 goto out3;
d342894c 3208
acaf4e70
DB		 3209 return 0;
3210out3:
3211 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3212out2:
783c1463 3213 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3214out1:
758c57d1 3215 destroy_workqueue(vxlan_wq);
d342894c 3216 return rc;
3217}
7332a13b 3218late_initcall(vxlan_init_module);
d342894c 3219
3220static void __exit vxlan_cleanup_module(void)
3221{
b7153984 3222 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3223 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 3224 destroy_workqueue(vxlan_wq);
783c1463
DB		 3225 unregister_pernet_subsys(&vxlan_net_ops);
3226 /* rcu_barrier() is called by netns */
d342894c 3227}
3228module_exit(vxlan_cleanup_module);
3229
3230MODULE_LICENSE("GPL");
3231MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3232MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3233MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3234MODULE_ALIAS_RTNL_LINK("vxlan");
ubuntu-bionic-kernel mirror