projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
vxlan: move IPv6 outpute route calculation to a function
[deliverable/linux.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>
acbf74a7 45#include <net/udp_tunnel.h>
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
afbd8bae
MR		 76static const u8 all_zeros_mac[ETH_ALEN];
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;
e4c7ed41 483 rd->remote_ip = *ip;
906dc186
TR		 484 rd->remote_port = port;
485 rd->remote_vni = vni;
486 rd->remote_ifindex = ifindex;
487 return 1;
488}
489
a5e7c10a
MR		 490/* Add/update destinations for multicast */
491static int vxlan_fdb_append(struct vxlan_fdb *f,
9e4b93f9
ND		 492 union vxlan_addr *ip, __be16 port, __u32 vni,
493 __u32 ifindex, struct vxlan_rdst **rdp)
a5e7c10a
MR		 494{
495 struct vxlan_rdst *rd;
496
497 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
498 if (rd)
499 return 0;
500
6681712d
DS		 501 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
502 if (rd == NULL)
503 return -ENOBUFS;
e4c7ed41 504 rd->remote_ip = *ip;
6681712d
DS		 505 rd->remote_port = port;
506 rd->remote_vni = vni;
507 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 508
509 list_add_tail_rcu(&rd->list, &f->remotes);
510
9e4b93f9 511 *rdp = rd;
6681712d
DS		 512 return 1;
513}
514
dfd8645e
TH		 515static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
516 unsigned int off,
517 struct vxlanhdr *vh, size_t hdrlen,
0ace2ca8
TH		 518 u32 data, struct gro_remcsum *grc,
519 bool nopartial)
dfd8645e 520{
b7fe10e5 521 size_t start, offset;
dfd8645e
TH		 522
523 if (skb->remcsum_offload)
b7fe10e5 524 return vh;
dfd8645e
TH		 525
526 if (!NAPI_GRO_CB(skb)->csum_valid)
527 return NULL;
528
529 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
530 offset = start + ((data & VXLAN_RCO_UDP) ?
531 offsetof(struct udphdr, check) :
532 offsetof(struct tcphdr, check));
533
b7fe10e5
TH		 534 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
535 start, offset, grc, nopartial);
dfd8645e
TH		 536
537 skb->remcsum_offload = 1;
538
539 return vh;
540}
541
a2b12f3c
TH		 542static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
543 struct sk_buff *skb,
544 struct udp_offload *uoff)
dc01e7d3
OG		 545{
546 struct sk_buff *p, **pp = NULL;
547 struct vxlanhdr *vh, *vh2;
9b174d88 548 unsigned int hlen, off_vx;
dc01e7d3 549 int flush = 1;
dfd8645e
TH		 550 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
551 udp_offloads);
552 u32 flags;
26c4f7da
TH		 553 struct gro_remcsum grc;
554
555 skb_gro_remcsum_init(&grc);
dc01e7d3
OG		 556
557 off_vx = skb_gro_offset(skb);
558 hlen = off_vx + sizeof(*vh);
559 vh = skb_gro_header_fast(skb, off_vx);
560 if (skb_gro_header_hard(skb, hlen)) {
561 vh = skb_gro_header_slow(skb, hlen, off_vx);
562 if (unlikely(!vh))
563 goto out;
564 }
dc01e7d3 565
dfd8645e
TH		 566 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
567
568 flags = ntohl(vh->vx_flags);
569
570 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
571 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
0ace2ca8
TH		 572 ntohl(vh->vx_vni), &grc,
573 !!(vs->flags &
574 VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 575
576 if (!vh)
577 goto out;
578 }
579
b7fe10e5
TH		 580 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
581
dc01e7d3
OG		 582 flush = 0;
583
584 for (p = *head; p; p = p->next) {
585 if (!NAPI_GRO_CB(p)->same_flow)
586 continue;
587
588 vh2 = (struct vxlanhdr *)(p->data + off_vx);
3511494c
TG		 589 if (vh->vx_flags != vh2->vx_flags ||
590 vh->vx_vni != vh2->vx_vni) {
dc01e7d3
OG		 591 NAPI_GRO_CB(p)->same_flow = 0;
592 continue;
593 }
dc01e7d3
OG		 594 }
595
9b174d88 596 pp = eth_gro_receive(head, skb);
dc01e7d3 597
dc01e7d3 598out:
26c4f7da 599 skb_gro_remcsum_cleanup(skb, &grc);
dc01e7d3
OG		 600 NAPI_GRO_CB(skb)->flush |= flush;
601
602 return pp;
603}
604
a2b12f3c
TH		 605static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
606 struct udp_offload *uoff)
dc01e7d3 607{
cfdf1e1b
JG		 608 udp_tunnel_gro_complete(skb, nhoff);
609
9b174d88 610 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
dc01e7d3
OG		 611}
612
53cf5275 613/* Notify netdevs that UDP port started listening */
dc01e7d3 614static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 615{
616 struct net_device *dev;
dc01e7d3 617 struct sock *sk = vs->sock->sk;
53cf5275 618 struct net *net = sock_net(sk);
705cc62f 619 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 620 __be16 port = inet_sk(sk)->inet_sport;
dc01e7d3
OG		 621 int err;
622
623 if (sa_family == AF_INET) {
624 err = udp_add_offload(&vs->udp_offloads);
625 if (err)
626 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
627 }
53cf5275
JG		 628
629 rcu_read_lock();
630 for_each_netdev_rcu(net, dev) {
631 if (dev->netdev_ops->ndo_add_vxlan_port)
632 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
633 port);
634 }
635 rcu_read_unlock();
636}
637
638/* Notify netdevs that UDP port is no more listening */
dc01e7d3 639static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
53cf5275
JG		 640{
641 struct net_device *dev;
dc01e7d3 642 struct sock *sk = vs->sock->sk;
53cf5275 643 struct net *net = sock_net(sk);
705cc62f 644 sa_family_t sa_family = vxlan_get_sk_family(vs);
35e42379 645 __be16 port = inet_sk(sk)->inet_sport;
53cf5275
JG		 646
647 rcu_read_lock();
648 for_each_netdev_rcu(net, dev) {
649 if (dev->netdev_ops->ndo_del_vxlan_port)
650 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
651 port);
652 }
653 rcu_read_unlock();
dc01e7d3
OG		 654
655 if (sa_family == AF_INET)
656 udp_del_offload(&vs->udp_offloads);
53cf5275
JG		 657}
658
d342894c 659/* Add new entry to forwarding table -- assumes lock held */
660static int vxlan_fdb_create(struct vxlan_dev *vxlan,
e4c7ed41 661 const u8 *mac, union vxlan_addr *ip,
6681712d 662 __u16 state, __u16 flags,
73cf3317 663 __be16 port, __u32 vni, __u32 ifindex,
ae884082 664 __u8 ndm_flags)
d342894c 665{
9e4b93f9 666 struct vxlan_rdst *rd = NULL;
d342894c 667 struct vxlan_fdb *f;
668 int notify = 0;
669
014be2c8 670 f = __vxlan_find_mac(vxlan, mac);
d342894c 671 if (f) {
672 if (flags & NLM_F_EXCL) {
673 netdev_dbg(vxlan->dev,
674 "lost race to create %pM\n", mac);
675 return -EEXIST;
676 }
677 if (f->state != state) {
678 f->state = state;
679 f->updated = jiffies;
680 notify = 1;
681 }
ae884082
DS		 682 if (f->flags != ndm_flags) {
683 f->flags = ndm_flags;
684 f->updated = jiffies;
685 notify = 1;
686 }
906dc186
TR		 687 if ((flags & NLM_F_REPLACE)) {
688 /* Only change unicasts */
689 if (!(is_multicast_ether_addr(f->eth_addr) ||
690 is_zero_ether_addr(f->eth_addr))) {
60840429 691 notify |= vxlan_fdb_replace(f, ip, port, vni,
906dc186 692 ifindex);
906dc186
TR		 693 } else
694 return -EOPNOTSUPP;
695 }
6681712d 696 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 697 (is_multicast_ether_addr(f->eth_addr) ||
698 is_zero_ether_addr(f->eth_addr))) {
9e4b93f9
ND		 699 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
700 &rd);
6681712d
DS		 701
702 if (rc < 0)
703 return rc;
704 notify |= rc;
705 }
d342894c 706 } else {
707 if (!(flags & NLM_F_CREATE))
708 return -ENOENT;
709
0dfbdf41
TG		 710 if (vxlan->cfg.addrmax &&
711 vxlan->addrcnt >= vxlan->cfg.addrmax)
d342894c 712 return -ENOSPC;
713
906dc186
TR		 714 /* Disallow replace to add a multicast entry */
715 if ((flags & NLM_F_REPLACE) &&
716 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
717 return -EOPNOTSUPP;
718
e4c7ed41 719 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
d342894c 720 f = kmalloc(sizeof(*f), GFP_ATOMIC);
721 if (!f)
722 return -ENOMEM;
723
724 notify = 1;
d342894c 725 f->state = state;
ae884082 726 f->flags = ndm_flags;
d342894c 727 f->updated = f->used = jiffies;
3e61aa8f 728 INIT_LIST_HEAD(&f->remotes);
d342894c 729 memcpy(f->eth_addr, mac, ETH_ALEN);
730
9e4b93f9 731 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
3e61aa8f 732
d342894c 733 ++vxlan->addrcnt;
734 hlist_add_head_rcu(&f->hlist,
735 vxlan_fdb_head(vxlan, mac));
736 }
737
9e4b93f9
ND		 738 if (notify) {
739 if (rd == NULL)
740 rd = first_remote_rtnl(f);
741 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
742 }
d342894c 743
744 return 0;
745}
746
6706c82e 747static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 748{
749 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 750 struct vxlan_rdst *rd, *nd;
6681712d 751
3e61aa8f 752 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 753 kfree(rd);
6681712d
DS		 754 kfree(f);
755}
756
d342894c 757static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
758{
759 netdev_dbg(vxlan->dev,
760 "delete %pM\n", f->eth_addr);
761
762 --vxlan->addrcnt;
9e4b93f9 763 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
d342894c 764
765 hlist_del_rcu(&f->hlist);
6681712d 766 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 767}
768
f0b074be 769static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
e4c7ed41 770 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 771{
6681712d 772 struct net *net = dev_net(vxlan->dev);
e4c7ed41 773 int err;
d342894c 774
f0b074be 775 if (tb[NDA_DST]) {
e4c7ed41
CW		 776 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
777 if (err)
778 return err;
f0b074be 779 } else {
e4c7ed41
CW		 780 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
781 if (remote->sa.sa_family == AF_INET) {
782 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
783 ip->sa.sa_family = AF_INET;
784#if IS_ENABLED(CONFIG_IPV6)
785 } else {
786 ip->sin6.sin6_addr = in6addr_any;
787 ip->sa.sa_family = AF_INET6;
788#endif
789 }
f0b074be 790 }
d342894c 791
6681712d 792 if (tb[NDA_PORT]) {
73cf3317 793 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 794 return -EINVAL;
f0b074be
MR		 795 *port = nla_get_be16(tb[NDA_PORT]);
796 } else {
0dfbdf41 797 *port = vxlan->cfg.dst_port;
f0b074be 798 }
6681712d
DS		 799
800 if (tb[NDA_VNI]) {
801 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
802 return -EINVAL;
f0b074be
MR		 803 *vni = nla_get_u32(tb[NDA_VNI]);
804 } else {
805 *vni = vxlan->default_dst.remote_vni;
806 }
6681712d
DS		 807
808 if (tb[NDA_IFINDEX]) {
5abb0029 809 struct net_device *tdev;
6681712d
DS		 810
811 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
812 return -EINVAL;
f0b074be 813 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
73763949 814 tdev = __dev_get_by_index(net, *ifindex);
5abb0029 815 if (!tdev)
6681712d 816 return -EADDRNOTAVAIL;
f0b074be
MR		 817 } else {
818 *ifindex = 0;
819 }
820
821 return 0;
822}
823
824/* Add static entry (via netlink) */
825static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
826 struct net_device *dev,
f6f6424b 827 const unsigned char *addr, u16 vid, u16 flags)
f0b074be
MR		 828{
829 struct vxlan_dev *vxlan = netdev_priv(dev);
830 /* struct net *net = dev_net(vxlan->dev); */
e4c7ed41 831 union vxlan_addr ip;
f0b074be
MR		 832 __be16 port;
833 u32 vni, ifindex;
834 int err;
835
836 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
837 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
838 ndm->ndm_state);
839 return -EINVAL;
840 }
841
842 if (tb[NDA_DST] == NULL)
843 return -EINVAL;
844
845 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
846 if (err)
847 return err;
6681712d 848
5933a7bb
MR		 849 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
850 return -EAFNOSUPPORT;
851
d342894c 852 spin_lock_bh(&vxlan->hash_lock);
e4c7ed41 853 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
73cf3317 854 port, vni, ifindex, ndm->ndm_flags);
d342894c 855 spin_unlock_bh(&vxlan->hash_lock);
856
857 return err;
858}
859
860/* Delete entry (via netlink) */
1690be63
VY		 861static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
862 struct net_device *dev,
f6f6424b 863 const unsigned char *addr, u16 vid)
d342894c 864{
865 struct vxlan_dev *vxlan = netdev_priv(dev);
866 struct vxlan_fdb *f;
bc7892ba 867 struct vxlan_rdst *rd = NULL;
e4c7ed41 868 union vxlan_addr ip;
bc7892ba
MR		 869 __be16 port;
870 u32 vni, ifindex;
871 int err;
872
873 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
874 if (err)
875 return err;
876
877 err = -ENOENT;
d342894c 878
879 spin_lock_bh(&vxlan->hash_lock);
880 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 881 if (!f)
882 goto out;
883
e4c7ed41
CW		 884 if (!vxlan_addr_any(&ip)) {
885 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
bc7892ba
MR		 886 if (!rd)
887 goto out;
888 }
889
890 err = 0;
891
892 /* remove a destination if it's not the only one on the list,
893 * otherwise destroy the fdb entry
894 */
895 if (rd && !list_is_singular(&f->remotes)) {
896 list_del_rcu(&rd->list);
9e4b93f9 897 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
dcdc7a59 898 kfree_rcu(rd, rcu);
bc7892ba 899 goto out;
d342894c 900 }
bc7892ba
MR		 901
902 vxlan_fdb_destroy(vxlan, f);
903
904out:
d342894c 905 spin_unlock_bh(&vxlan->hash_lock);
906
907 return err;
908}
909
910/* Dump forwarding table */
911static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
5d5eacb3
JHS		 912 struct net_device *dev,
913 struct net_device *filter_dev, int idx)
d342894c 914{
915 struct vxlan_dev *vxlan = netdev_priv(dev);
916 unsigned int h;
917
918 for (h = 0; h < FDB_HASH_SIZE; ++h) {
919 struct vxlan_fdb *f;
d342894c 920 int err;
921
b67bfe0d 922 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 923 struct vxlan_rdst *rd;
6681712d 924
3e61aa8f 925 list_for_each_entry_rcu(rd, &f->remotes, list) {
07a51cd3
AW		 926 if (idx < cb->args[0])
927 goto skip;
928
6681712d
DS		 929 err = vxlan_fdb_info(skb, vxlan, f,
930 NETLINK_CB(cb->skb).portid,
931 cb->nlh->nlmsg_seq,
932 RTM_NEWNEIGH,
933 NLM_F_MULTI, rd);
934 if (err < 0)
3e61aa8f 935 goto out;
3e61aa8f 936skip:
07a51cd3
AW		 937 ++idx;
938 }
d342894c 939 }
940 }
3e61aa8f 941out:
d342894c 942 return idx;
943}
944
945/* Watch incoming packets to learn mapping between Ethernet address
946 * and Tunnel endpoint.
c4b49512 947 * Return true if packet is bogus and should be dropped.
d342894c 948 */
26a41ae6 949static bool vxlan_snoop(struct net_device *dev,
e4c7ed41 950 union vxlan_addr *src_ip, const u8 *src_mac)
d342894c 951{
952 struct vxlan_dev *vxlan = netdev_priv(dev);
953 struct vxlan_fdb *f;
d342894c 954
955 f = vxlan_find_mac(vxlan, src_mac);
956 if (likely(f)) {
5ca5461c 957 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f 958
e4c7ed41 959 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
26a41ae6 960 return false;
961
962 /* Don't migrate static entries, drop packets */
eb064c3b 963 if (f->state & NUD_NOARP)
26a41ae6 964 return true;
d342894c 965
966 if (net_ratelimit())
967 netdev_info(dev,
e4c7ed41 968 "%pM migrated from %pIS to %pIS\n",
a4870f79 969 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
d342894c 970
e4c7ed41 971 rdst->remote_ip = *src_ip;
d342894c 972 f->updated = jiffies;
9e4b93f9 973 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
d342894c 974 } else {
975 /* learned new entry */
976 spin_lock(&vxlan->hash_lock);
3bf74b1a 977
978 /* close off race between vxlan_flush and incoming packets */
979 if (netif_running(dev))
980 vxlan_fdb_create(vxlan, src_mac, src_ip,
981 NUD_REACHABLE,
982 NLM_F_EXCL|NLM_F_CREATE,
0dfbdf41 983 vxlan->cfg.dst_port,
3bf74b1a 984 vxlan->default_dst.remote_vni,
985 0, NTF_SELF);
d342894c 986 spin_unlock(&vxlan->hash_lock);
987 }
26a41ae6 988
989 return false;
d342894c 990}
991
d342894c 992/* See if multicast group is already in use by other ID */
95ab0991 993static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
d342894c 994{
553675fb 995 struct vxlan_dev *vxlan;
b1be00a6 996 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
d342894c 997
95ab0991
G		 998 /* The vxlan_sock is only used by dev, leaving group has
999 * no effect on other vxlan devices.
1000 */
b1be00a6
JB		 1001 if (family == AF_INET && dev->vn4_sock &&
1002 atomic_read(&dev->vn4_sock->refcnt) == 1)
95ab0991 1003 return false;
b1be00a6
JB		 1004#if IS_ENABLED(CONFIG_IPV6)
1005 if (family == AF_INET6 && dev->vn6_sock &&
1006 atomic_read(&dev->vn6_sock->refcnt) == 1)
1007 return false;
1008#endif
95ab0991 1009
553675fb 1010 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
95ab0991 1011 if (!netif_running(vxlan->dev) || vxlan == dev)
553675fb 1012 continue;
d342894c 1013
b1be00a6 1014 if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock)
95ab0991 1015 continue;
b1be00a6
JB		 1016#if IS_ENABLED(CONFIG_IPV6)
1017 if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock)
1018 continue;
1019#endif
95ab0991
G		 1020
1021 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1022 &dev->default_dst.remote_ip))
1023 continue;
1024
1025 if (vxlan->default_dst.remote_ifindex !=
1026 dev->default_dst.remote_ifindex)
1027 continue;
1028
1029 return true;
553675fb 1030 }
d342894c 1031
1032 return false;
1033}
1034
b1be00a6 1035static void __vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 1036{
b1be00a6 1037 struct vxlan_net *vn;
012a5729 1038
b1be00a6
JB		 1039 if (!vs)
1040 return;
7c47cedf
SH		 1041 if (!atomic_dec_and_test(&vs->refcnt))
1042 return;
d342894c 1043
b1be00a6 1044 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1c51a915 1045 spin_lock(&vn->sock_lock);
7c47cedf 1046 hlist_del_rcu(&vs->hlist);
dc01e7d3 1047 vxlan_notify_del_rx_port(vs);
1c51a915
SH		 1048 spin_unlock(&vn->sock_lock);
1049
7c47cedf 1050 queue_work(vxlan_wq, &vs->del_work);
d342894c 1051}
1052
b1be00a6
JB		 1053static void vxlan_sock_release(struct vxlan_dev *vxlan)
1054{
1055 __vxlan_sock_release(vxlan->vn4_sock);
1056#if IS_ENABLED(CONFIG_IPV6)
1057 __vxlan_sock_release(vxlan->vn6_sock);
1058#endif
1059}
1060
56ef9c90 1061/* Update multicast group membership when first VNI on
c4b49512 1062 * multicast address is brought up
7c47cedf 1063 */
56ef9c90 1064static int vxlan_igmp_join(struct vxlan_dev *vxlan)
d342894c 1065{
b1be00a6 1066 struct sock *sk;
e4c7ed41
CW		 1067 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1068 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1069 int ret = -EINVAL;
d342894c 1070
e4c7ed41
CW		 1071 if (ip->sa.sa_family == AF_INET) {
1072 struct ip_mreqn mreq = {
1073 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1074 .imr_ifindex = ifindex,
1075 };
1076
b1be00a6
JB		 1077 sk = vxlan->vn4_sock->sock->sk;
1078 lock_sock(sk);
56ef9c90 1079 ret = ip_mc_join_group(sk, &mreq);
b1be00a6 1080 release_sock(sk);
e4c7ed41
CW		 1081#if IS_ENABLED(CONFIG_IPV6)
1082 } else {
b1be00a6
JB		 1083 sk = vxlan->vn6_sock->sock->sk;
1084 lock_sock(sk);
56ef9c90
MRL		 1085 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1086 &ip->sin6.sin6_addr);
b1be00a6 1087 release_sock(sk);
e4c7ed41
CW		 1088#endif
1089 }
3fc2de2f 1090
56ef9c90 1091 return ret;
3fc2de2f 1092}
1093
1094/* Inverse of vxlan_igmp_join when last VNI is brought down */
56ef9c90 1095static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
3fc2de2f 1096{
b1be00a6 1097 struct sock *sk;
e4c7ed41
CW		 1098 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1099 int ifindex = vxlan->default_dst.remote_ifindex;
149d7549 1100 int ret = -EINVAL;
3fc2de2f 1101
e4c7ed41
CW		 1102 if (ip->sa.sa_family == AF_INET) {
1103 struct ip_mreqn mreq = {
1104 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1105 .imr_ifindex = ifindex,
1106 };
1107
b1be00a6
JB		 1108 sk = vxlan->vn4_sock->sock->sk;
1109 lock_sock(sk);
56ef9c90 1110 ret = ip_mc_leave_group(sk, &mreq);
b1be00a6 1111 release_sock(sk);
e4c7ed41
CW		 1112#if IS_ENABLED(CONFIG_IPV6)
1113 } else {
b1be00a6
JB		 1114 sk = vxlan->vn6_sock->sock->sk;
1115 lock_sock(sk);
56ef9c90
MRL		 1116 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1117 &ip->sin6.sin6_addr);
b1be00a6 1118 release_sock(sk);
e4c7ed41
CW		 1119#endif
1120 }
d342894c 1121
56ef9c90 1122 return ret;
d342894c 1123}
1124
dfd8645e 1125static struct vxlanhdr *vxlan_remcsum(struct sk_buff *skb, struct vxlanhdr *vh,
0ace2ca8 1126 size_t hdrlen, u32 data, bool nopartial)
dfd8645e
TH		 1127{
1128 size_t start, offset, plen;
dfd8645e 1129
b7fe10e5
TH		 1130 if (skb->remcsum_offload)
1131 return vh;
1132
dfd8645e
TH		 1133 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
1134 offset = start + ((data & VXLAN_RCO_UDP) ?
1135 offsetof(struct udphdr, check) :
1136 offsetof(struct tcphdr, check));
1137
1138 plen = hdrlen + offset + sizeof(u16);
1139
1140 if (!pskb_may_pull(skb, plen))
1141 return NULL;
1142
1143 vh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1144
0ace2ca8
TH		 1145 skb_remcsum_process(skb, (void *)vh + hdrlen, start, offset,
1146 nopartial);
dfd8645e
TH		 1147
1148 return vh;
1149}
1150
614732ea
TG		 1151static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
1152 struct vxlan_metadata *md, u32 vni,
1153 struct metadata_dst *tun_dst)
1154{
1155 struct iphdr *oip = NULL;
1156 struct ipv6hdr *oip6 = NULL;
1157 struct vxlan_dev *vxlan;
1158 struct pcpu_sw_netstats *stats;
1159 union vxlan_addr saddr;
1160 int err = 0;
1161 union vxlan_addr *remote_ip;
1162
1163 /* For flow based devices, map all packets to VNI 0 */
da8b43c0 1164 if (vs->flags & VXLAN_F_COLLECT_METADATA)
614732ea
TG		 1165 vni = 0;
1166
1167 /* Is this VNI defined? */
1168 vxlan = vxlan_vs_find_vni(vs, vni);
1169 if (!vxlan)
1170 goto drop;
1171
1172 remote_ip = &vxlan->default_dst.remote_ip;
1173 skb_reset_mac_header(skb);
1174 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
1175 skb->protocol = eth_type_trans(skb, vxlan->dev);
1176 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1177
1178 /* Ignore packet loops (and multicast echo) */
1179 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1180 goto drop;
1181
1182 /* Re-examine inner Ethernet packet */
1183 if (remote_ip->sa.sa_family == AF_INET) {
1184 oip = ip_hdr(skb);
1185 saddr.sin.sin_addr.s_addr = oip->saddr;
1186 saddr.sa.sa_family = AF_INET;
1187#if IS_ENABLED(CONFIG_IPV6)
1188 } else {
1189 oip6 = ipv6_hdr(skb);
1190 saddr.sin6.sin6_addr = oip6->saddr;
1191 saddr.sa.sa_family = AF_INET6;
1192#endif
1193 }
1194
1195 if (tun_dst) {
1196 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1197 tun_dst = NULL;
1198 }
1199
1200 if ((vxlan->flags & VXLAN_F_LEARN) &&
1201 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1202 goto drop;
1203
1204 skb_reset_network_header(skb);
1205 /* In flow-based mode, GBP is carried in dst_metadata */
da8b43c0 1206 if (!(vs->flags & VXLAN_F_COLLECT_METADATA))
614732ea
TG		 1207 skb->mark = md->gbp;
1208
1209 if (oip6)
1210 err = IP6_ECN_decapsulate(oip6, skb);
1211 if (oip)
1212 err = IP_ECN_decapsulate(oip, skb);
1213
1214 if (unlikely(err)) {
1215 if (log_ecn_error) {
1216 if (oip6)
1217 net_info_ratelimited("non-ECT from %pI6\n",
1218 &oip6->saddr);
1219 if (oip)
1220 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1221 &oip->saddr, oip->tos);
1222 }
1223 if (err > 1) {
1224 ++vxlan->dev->stats.rx_frame_errors;
1225 ++vxlan->dev->stats.rx_errors;
1226 goto drop;
1227 }
1228 }
1229
1230 stats = this_cpu_ptr(vxlan->dev->tstats);
1231 u64_stats_update_begin(&stats->syncp);
1232 stats->rx_packets++;
1233 stats->rx_bytes += skb->len;
1234 u64_stats_update_end(&stats->syncp);
1235
58ce31cc 1236 gro_cells_receive(&vxlan->gro_cells, skb);
614732ea
TG		 1237
1238 return;
1239drop:
1240 if (tun_dst)
1241 dst_release((struct dst_entry *)tun_dst);
1242
1243 /* Consume bad packet */
1244 kfree_skb(skb);
1245}
1246
d342894c 1247/* Callback from net/ipv4/udp.c to receive packets */
1248static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1249{
ee122c79 1250 struct metadata_dst *tun_dst = NULL;
5cfccc5a 1251 struct vxlan_sock *vs;
d342894c 1252 struct vxlanhdr *vxh;
3bf39475 1253 u32 flags, vni;
ee122c79
TG		 1254 struct vxlan_metadata _md;
1255 struct vxlan_metadata *md = &_md;
d342894c 1256
d342894c 1257 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 1258 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 1259 goto error;
1260
7ce04758 1261 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
3bf39475
TH		 1262 flags = ntohl(vxh->vx_flags);
1263 vni = ntohl(vxh->vx_vni);
1264
1265 if (flags & VXLAN_HF_VNI) {
1266 flags &= ~VXLAN_HF_VNI;
1267 } else {
1268 /* VNI flag always required to be set */
1269 goto bad_flags;
d342894c 1270 }
1271
5cfccc5a
PS		 1272 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1273 goto drop;
dfd8645e 1274 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
5cfccc5a 1275
559835ea 1276 vs = rcu_dereference_sk_user_data(sk);
5cfccc5a 1277 if (!vs)
d342894c 1278 goto drop;
d342894c 1279
dfd8645e 1280 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
0ace2ca8
TH		 1281 vxh = vxlan_remcsum(skb, vxh, sizeof(struct vxlanhdr), vni,
1282 !!(vs->flags & VXLAN_F_REMCSUM_NOPARTIAL));
dfd8645e
TH		 1283 if (!vxh)
1284 goto drop;
1285
1286 flags &= ~VXLAN_HF_RCO;
40fb70f3 1287 vni &= VXLAN_VNI_MASK;
dfd8645e
TH		 1288 }
1289
ee122c79 1290 if (vxlan_collect_metadata(vs)) {
c29a70d2
PS		 1291 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1292 cpu_to_be64(vni >> 8), sizeof(*md));
1293
ee122c79
TG		 1294 if (!tun_dst)
1295 goto drop;
1296
0f1b7354 1297 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
ee122c79
TG		 1298 } else {
1299 memset(md, 0, sizeof(*md));
1300 }
1301
3511494c
TG		 1302 /* For backwards compatibility, only allow reserved fields to be
1303 * used by VXLAN extensions if explicitly requested.
1304 */
1305 if ((flags & VXLAN_HF_GBP) && (vs->flags & VXLAN_F_GBP)) {
1306 struct vxlanhdr_gbp *gbp;
1307
1308 gbp = (struct vxlanhdr_gbp *)vxh;
ee122c79
TG		 1309 md->gbp = ntohs(gbp->policy_id);
1310
1311 if (tun_dst)
0f1b7354 1312 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
3511494c
TG		 1313
1314 if (gbp->dont_learn)
ee122c79 1315 md->gbp |= VXLAN_GBP_DONT_LEARN;
3511494c
TG		 1316
1317 if (gbp->policy_applied)
ee122c79 1318 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
3511494c
TG		 1319
1320 flags &= ~VXLAN_GBP_USED_BITS;
1321 }
1322
40fb70f3 1323 if (flags || vni & ~VXLAN_VNI_MASK) {
3bf39475
TH		 1324 /* If there are any unprocessed flags remaining treat
1325 * this as a malformed packet. This behavior diverges from
1326 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1327 * in reserved fields are to be ignored. The approach here
c4b49512 1328 * maintains compatibility with previous stack code, and also
3bf39475
TH		 1329 * is more robust and provides a little more security in
1330 * adding extensions to VXLAN.
1331 */
1332
1333 goto bad_flags;
1334 }
1335
614732ea 1336 vxlan_rcv(vs, skb, md, vni >> 8, tun_dst);
5cfccc5a
PS		 1337 return 0;
1338
1339drop:
1340 /* Consume bad packet */
1341 kfree_skb(skb);
1342 return 0;
1343
3bf39475
TH		 1344bad_flags:
1345 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1346 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1347
5cfccc5a 1348error:
ee122c79
TG		 1349 if (tun_dst)
1350 dst_release((struct dst_entry *)tun_dst);
1351
5cfccc5a
PS		 1352 /* Return non vxlan pkt */
1353 return 1;
1354}
1355
e4f67add
DS		 1356static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1357{
1358 struct vxlan_dev *vxlan = netdev_priv(dev);
1359 struct arphdr *parp;
1360 u8 *arpptr, *sha;
1361 __be32 sip, tip;
1362 struct neighbour *n;
1363
1364 if (dev->flags & IFF_NOARP)
1365 goto out;
1366
1367 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1368 dev->stats.tx_dropped++;
1369 goto out;
1370 }
1371 parp = arp_hdr(skb);
1372
1373 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1374 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1375 parp->ar_pro != htons(ETH_P_IP) ||
1376 parp->ar_op != htons(ARPOP_REQUEST) ||
1377 parp->ar_hln != dev->addr_len ||
1378 parp->ar_pln != 4)
1379 goto out;
1380 arpptr = (u8 *)parp + sizeof(struct arphdr);
1381 sha = arpptr;
1382 arpptr += dev->addr_len; /* sha */
1383 memcpy(&sip, arpptr, sizeof(sip));
1384 arpptr += sizeof(sip);
1385 arpptr += dev->addr_len; /* tha */
1386 memcpy(&tip, arpptr, sizeof(tip));
1387
1388 if (ipv4_is_loopback(tip) ||
1389 ipv4_is_multicast(tip))
1390 goto out;
1391
1392 n = neigh_lookup(&arp_tbl, &tip, dev);
1393
1394 if (n) {
e4f67add
DS		 1395 struct vxlan_fdb *f;
1396 struct sk_buff *reply;
1397
1398 if (!(n->nud_state & NUD_CONNECTED)) {
1399 neigh_release(n);
1400 goto out;
1401 }
1402
1403 f = vxlan_find_mac(vxlan, n->ha);
e4c7ed41 1404 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
e4f67add
DS		 1405 /* bridge-local neighbor */
1406 neigh_release(n);
1407 goto out;
1408 }
1409
1410 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1411 n->ha, sha);
1412
1413 neigh_release(n);
1414
7346135d
DS		 1415 if (reply == NULL)
1416 goto out;
1417
e4f67add
DS		 1418 skb_reset_mac_header(reply);
1419 __skb_pull(reply, skb_network_offset(reply));
1420 reply->ip_summed = CHECKSUM_UNNECESSARY;
1421 reply->pkt_type = PACKET_HOST;
1422
1423 if (netif_rx_ni(reply) == NET_RX_DROP)
1424 dev->stats.rx_dropped++;
e4c7ed41
CW		 1425 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1426 union vxlan_addr ipa = {
1427 .sin.sin_addr.s_addr = tip,
a45e92a5 1428 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1429 };
1430
1431 vxlan_ip_miss(dev, &ipa);
1432 }
e4f67add
DS		 1433out:
1434 consume_skb(skb);
1435 return NETDEV_TX_OK;
1436}
1437
f564f45c 1438#if IS_ENABLED(CONFIG_IPV6)
4b29dba9
DS		 1439static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1440 struct neighbour *n, bool isrouter)
1441{
1442 struct net_device *dev = request->dev;
1443 struct sk_buff *reply;
1444 struct nd_msg *ns, *na;
1445 struct ipv6hdr *pip6;
1446 u8 *daddr;
1447 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1448 int ns_olen;
1449 int i, len;
1450
1451 if (dev == NULL)
1452 return NULL;
1453
1454 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1455 sizeof(*na) + na_olen + dev->needed_tailroom;
1456 reply = alloc_skb(len, GFP_ATOMIC);
1457 if (reply == NULL)
1458 return NULL;
1459
1460 reply->protocol = htons(ETH_P_IPV6);
1461 reply->dev = dev;
1462 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1463 skb_push(reply, sizeof(struct ethhdr));
1464 skb_set_mac_header(reply, 0);
1465
1466 ns = (struct nd_msg *)skb_transport_header(request);
1467
1468 daddr = eth_hdr(request)->h_source;
1469 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1470 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1471 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1472 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1473 break;
1474 }
1475 }
1476
1477 /* Ethernet header */
1478 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1479 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1480 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1481 reply->protocol = htons(ETH_P_IPV6);
1482
1483 skb_pull(reply, sizeof(struct ethhdr));
1484 skb_set_network_header(reply, 0);
1485 skb_put(reply, sizeof(struct ipv6hdr));
1486
1487 /* IPv6 header */
1488
1489 pip6 = ipv6_hdr(reply);
1490 memset(pip6, 0, sizeof(struct ipv6hdr));
1491 pip6->version = 6;
1492 pip6->priority = ipv6_hdr(request)->priority;
1493 pip6->nexthdr = IPPROTO_ICMPV6;
1494 pip6->hop_limit = 255;
1495 pip6->daddr = ipv6_hdr(request)->saddr;
1496 pip6->saddr = *(struct in6_addr *)n->primary_key;
1497
1498 skb_pull(reply, sizeof(struct ipv6hdr));
1499 skb_set_transport_header(reply, 0);
1500
1501 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1502
1503 /* Neighbor Advertisement */
1504 memset(na, 0, sizeof(*na)+na_olen);
1505 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1506 na->icmph.icmp6_router = isrouter;
1507 na->icmph.icmp6_override = 1;
1508 na->icmph.icmp6_solicited = 1;
1509 na->target = ns->target;
1510 ether_addr_copy(&na->opt[2], n->ha);
1511 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1512 na->opt[1] = na_olen >> 3;
1513
1514 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1515 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1516 csum_partial(na, sizeof(*na)+na_olen, 0));
1517
1518 pip6->payload_len = htons(sizeof(*na)+na_olen);
1519
1520 skb_push(reply, sizeof(struct ipv6hdr));
1521
1522 reply->ip_summed = CHECKSUM_UNNECESSARY;
1523
1524 return reply;
1525}
1526
f564f45c
CW		 1527static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1528{
1529 struct vxlan_dev *vxlan = netdev_priv(dev);
4b29dba9 1530 struct nd_msg *msg;
f564f45c
CW		 1531 const struct ipv6hdr *iphdr;
1532 const struct in6_addr *saddr, *daddr;
4b29dba9
DS		 1533 struct neighbour *n;
1534 struct inet6_dev *in6_dev;
f564f45c
CW		 1535
1536 in6_dev = __in6_dev_get(dev);
1537 if (!in6_dev)
1538 goto out;
1539
f564f45c
CW		 1540 iphdr = ipv6_hdr(skb);
1541 saddr = &iphdr->saddr;
1542 daddr = &iphdr->daddr;
1543
f564f45c
CW		 1544 msg = (struct nd_msg *)skb_transport_header(skb);
1545 if (msg->icmph.icmp6_code != 0 ||
1546 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1547 goto out;
1548
4b29dba9
DS		 1549 if (ipv6_addr_loopback(daddr) ||
1550 ipv6_addr_is_multicast(&msg->target))
1551 goto out;
1552
1553 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
f564f45c
CW		 1554
1555 if (n) {
1556 struct vxlan_fdb *f;
4b29dba9 1557 struct sk_buff *reply;
f564f45c
CW		 1558
1559 if (!(n->nud_state & NUD_CONNECTED)) {
1560 neigh_release(n);
1561 goto out;
1562 }
1563
1564 f = vxlan_find_mac(vxlan, n->ha);
1565 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1566 /* bridge-local neighbor */
1567 neigh_release(n);
1568 goto out;
1569 }
1570
4b29dba9
DS		 1571 reply = vxlan_na_create(skb, n,
1572 !!(f ? f->flags & NTF_ROUTER : 0));
1573
f564f45c 1574 neigh_release(n);
4b29dba9
DS		 1575
1576 if (reply == NULL)
1577 goto out;
1578
1579 if (netif_rx_ni(reply) == NET_RX_DROP)
1580 dev->stats.rx_dropped++;
1581
f564f45c 1582 } else if (vxlan->flags & VXLAN_F_L3MISS) {
4b29dba9
DS		 1583 union vxlan_addr ipa = {
1584 .sin6.sin6_addr = msg->target,
a45e92a5 1585 .sin6.sin6_family = AF_INET6,
4b29dba9
DS		 1586 };
1587
f564f45c
CW		 1588 vxlan_ip_miss(dev, &ipa);
1589 }
1590
1591out:
1592 consume_skb(skb);
1593 return NETDEV_TX_OK;
1594}
1595#endif
1596
e4f67add
DS		 1597static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1598{
1599 struct vxlan_dev *vxlan = netdev_priv(dev);
1600 struct neighbour *n;
e4f67add
DS		 1601
1602 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1603 return false;
1604
1605 n = NULL;
1606 switch (ntohs(eth_hdr(skb)->h_proto)) {
1607 case ETH_P_IP:
e15a00aa
CW		 1608 {
1609 struct iphdr *pip;
1610
e4f67add
DS		 1611 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1612 return false;
1613 pip = ip_hdr(skb);
1614 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
e4c7ed41
CW		 1615 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1616 union vxlan_addr ipa = {
1617 .sin.sin_addr.s_addr = pip->daddr,
a45e92a5 1618 .sin.sin_family = AF_INET,
e4c7ed41
CW		 1619 };
1620
1621 vxlan_ip_miss(dev, &ipa);
1622 return false;
1623 }
1624
e4f67add 1625 break;
e15a00aa
CW		 1626 }
1627#if IS_ENABLED(CONFIG_IPV6)
1628 case ETH_P_IPV6:
1629 {
1630 struct ipv6hdr *pip6;
1631
1632 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1633 return false;
1634 pip6 = ipv6_hdr(skb);
1635 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1636 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1637 union vxlan_addr ipa = {
1638 .sin6.sin6_addr = pip6->daddr,
a45e92a5 1639 .sin6.sin6_family = AF_INET6,
e15a00aa
CW		 1640 };
1641
1642 vxlan_ip_miss(dev, &ipa);
1643 return false;
1644 }
1645
1646 break;
1647 }
1648#endif
e4f67add
DS		 1649 default:
1650 return false;
1651 }
1652
1653 if (n) {
1654 bool diff;
1655
7367d0b5 1656 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
e4f67add
DS		 1657 if (diff) {
1658 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1659 dev->addr_len);
1660 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1661 }
1662 neigh_release(n);
1663 return diff;
e4c7ed41
CW		 1664 }
1665
e4f67add
DS		 1666 return false;
1667}
1668
af33c1ad 1669static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
3511494c
TG		 1670 struct vxlan_metadata *md)
1671{
1672 struct vxlanhdr_gbp *gbp;
1673
db79a621
TG		 1674 if (!md->gbp)
1675 return;
1676
3511494c
TG		 1677 gbp = (struct vxlanhdr_gbp *)vxh;
1678 vxh->vx_flags |= htonl(VXLAN_HF_GBP);
1679
1680 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1681 gbp->dont_learn = 1;
1682
1683 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1684 gbp->policy_applied = 1;
1685
1686 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1687}
1688
e4c7ed41 1689#if IS_ENABLED(CONFIG_IPV6)
79b16aad
DM		 1690static int vxlan6_xmit_skb(struct dst_entry *dst, struct sock *sk,
1691 struct sk_buff *skb,
e4c7ed41
CW		 1692 struct net_device *dev, struct in6_addr *saddr,
1693 struct in6_addr *daddr, __u8 prio, __u8 ttl,
6b6948dd 1694 __be16 src_port, __be16 dst_port, __be32 vni,
af33c1ad 1695 struct vxlan_metadata *md, bool xnet, u32 vxflags)
e4c7ed41 1696{
e4c7ed41 1697 struct vxlanhdr *vxh;
e4c7ed41
CW		 1698 int min_headroom;
1699 int err;
af33c1ad 1700 bool udp_sum = !(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX);
dfd8645e
TH		 1701 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1702 u16 hdrlen = sizeof(struct vxlanhdr);
1703
af33c1ad 1704 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1705 skb->ip_summed == CHECKSUM_PARTIAL) {
1706 int csum_start = skb_checksum_start_offset(skb);
1707
1708 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1709 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1710 (skb->csum_offset == offsetof(struct udphdr, check) ||
1711 skb->csum_offset == offsetof(struct tcphdr, check))) {
1712 udp_sum = false;
1713 type |= SKB_GSO_TUNNEL_REMCSUM;
1714 }
1715 }
e4c7ed41 1716
f01ec1c0 1717 skb_scrub_packet(skb, xnet);
963a88b3 1718
e4c7ed41
CW		 1719 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1720 + VXLAN_HLEN + sizeof(struct ipv6hdr)
df8a39de 1721 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
e4c7ed41
CW		 1722
1723 /* Need space for new headers (invalidates iph ptr) */
1724 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1725 if (unlikely(err)) {
1726 kfree_skb(skb);
1727 goto err;
1728 }
e4c7ed41 1729
5968250c 1730 skb = vlan_hwaccel_push_inside(skb);
74f47278
PS		 1731 if (WARN_ON(!skb)) {
1732 err = -ENOMEM;
1733 goto err;
1734 }
e4c7ed41 1735
b736a623
JG		 1736 skb = iptunnel_handle_offloads(skb, udp_sum, type);
1737 if (IS_ERR(skb)) {
1738 err = -EINVAL;
1739 goto err;
1740 }
1741
e4c7ed41 1742 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1743 vxh->vx_flags = htonl(VXLAN_HF_VNI);
614732ea 1744 vxh->vx_vni = vni;
e4c7ed41 1745
dfd8645e
TH		 1746 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1747 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1748 VXLAN_RCO_SHIFT;
1749
1750 if (skb->csum_offset == offsetof(struct udphdr, check))
1751 data |= VXLAN_RCO_UDP;
1752
1753 vxh->vx_vni |= htonl(data);
1754 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1755
1756 if (!skb_is_gso(skb)) {
1757 skb->ip_summed = CHECKSUM_NONE;
1758 skb->encapsulation = 0;
1759 }
1760 }
1761
af33c1ad
TH		 1762 if (vxflags & VXLAN_F_GBP)
1763 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1764
996c9fd1
TH		 1765 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1766
79b16aad 1767 udp_tunnel6_xmit_skb(dst, sk, skb, dev, saddr, daddr, prio,
d998f8ef 1768 ttl, src_port, dst_port,
af33c1ad 1769 !!(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX));
e4c7ed41 1770 return 0;
74f47278
PS		 1771err:
1772 dst_release(dst);
1773 return err;
e4c7ed41
CW		 1774}
1775#endif
1776
614732ea
TG		 1777static int vxlan_xmit_skb(struct rtable *rt, struct sock *sk, struct sk_buff *skb,
1778 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
6b6948dd 1779 __be16 src_port, __be16 dst_port, __be32 vni,
614732ea 1780 struct vxlan_metadata *md, bool xnet, u32 vxflags)
49560532
PS		 1781{
1782 struct vxlanhdr *vxh;
649c5b8b 1783 int min_headroom;
49560532 1784 int err;
af33c1ad 1785 bool udp_sum = !!(vxflags & VXLAN_F_UDP_CSUM);
dfd8645e
TH		 1786 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1787 u16 hdrlen = sizeof(struct vxlanhdr);
1788
af33c1ad 1789 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
dfd8645e
TH		 1790 skb->ip_summed == CHECKSUM_PARTIAL) {
1791 int csum_start = skb_checksum_start_offset(skb);
1792
1793 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1794 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1795 (skb->csum_offset == offsetof(struct udphdr, check) ||
1796 skb->csum_offset == offsetof(struct tcphdr, check))) {
1797 udp_sum = false;
1798 type |= SKB_GSO_TUNNEL_REMCSUM;
1799 }
1800 }
49560532 1801
649c5b8b 1802 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1eaa8178 1803 + VXLAN_HLEN + sizeof(struct iphdr)
df8a39de 1804 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
649c5b8b
PS		 1805
1806 /* Need space for new headers (invalidates iph ptr) */
1807 err = skb_cow_head(skb, min_headroom);
74f47278
PS		 1808 if (unlikely(err)) {
1809 kfree_skb(skb);
649c5b8b 1810 return err;
74f47278 1811 }
649c5b8b 1812
5968250c
JP		 1813 skb = vlan_hwaccel_push_inside(skb);
1814 if (WARN_ON(!skb))
1815 return -ENOMEM;
1eaa8178 1816
b736a623
JG		 1817 skb = iptunnel_handle_offloads(skb, udp_sum, type);
1818 if (IS_ERR(skb))
1819 return PTR_ERR(skb);
1820
49560532 1821 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
3bf39475 1822 vxh->vx_flags = htonl(VXLAN_HF_VNI);
614732ea 1823 vxh->vx_vni = vni;
49560532 1824
dfd8645e
TH		 1825 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1826 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1827 VXLAN_RCO_SHIFT;
1828
1829 if (skb->csum_offset == offsetof(struct udphdr, check))
1830 data |= VXLAN_RCO_UDP;
1831
1832 vxh->vx_vni |= htonl(data);
1833 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1834
1835 if (!skb_is_gso(skb)) {
1836 skb->ip_summed = CHECKSUM_NONE;
1837 skb->encapsulation = 0;
1838 }
1839 }
1840
af33c1ad
TH		 1841 if (vxflags & VXLAN_F_GBP)
1842 vxlan_build_gbp_hdr(vxh, vxflags, md);
3511494c 1843
996c9fd1
TH		 1844 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1845
79b16aad 1846 return udp_tunnel_xmit_skb(rt, sk, skb, src, dst, tos,
d998f8ef 1847 ttl, df, src_port, dst_port, xnet,
af33c1ad 1848 !(vxflags & VXLAN_F_UDP_CSUM));
49560532 1849}
49560532 1850
e5d4b29f
JB		 1851#if IS_ENABLED(CONFIG_IPV6)
1852static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1853 struct sk_buff *skb, int oif,
1854 const struct in6_addr *daddr,
1855 struct in6_addr *saddr)
1856{
1857 struct dst_entry *ndst;
1858 struct flowi6 fl6;
1859 int err;
1860
1861 memset(&fl6, 0, sizeof(fl6));
1862 fl6.flowi6_oif = oif;
1863 fl6.daddr = *daddr;
1864 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
1865 fl6.flowi6_mark = skb->mark;
1866 fl6.flowi6_proto = IPPROTO_UDP;
1867
1868 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1869 vxlan->vn6_sock->sock->sk,
1870 &ndst, &fl6);
1871 if (err < 0)
1872 return ERR_PTR(err);
1873
1874 *saddr = fl6.saddr;
1875 return ndst;
1876}
1877#endif
1878
9dcc71e1
SS		 1879/* Bypass encapsulation if the destination is local */
1880static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1881 struct vxlan_dev *dst_vxlan)
1882{
8f84985f 1883 struct pcpu_sw_netstats *tx_stats, *rx_stats;
e4c7ed41
CW		 1884 union vxlan_addr loopback;
1885 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
ce6502a8
LR		 1886 struct net_device *dev = skb->dev;
1887 int len = skb->len;
9dcc71e1 1888
8f84985f
LR		 1889 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1890 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
9dcc71e1
SS		 1891 skb->pkt_type = PACKET_HOST;
1892 skb->encapsulation = 0;
1893 skb->dev = dst_vxlan->dev;
1894 __skb_pull(skb, skb_network_offset(skb));
1895
e4c7ed41
CW		 1896 if (remote_ip->sa.sa_family == AF_INET) {
1897 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1898 loopback.sa.sa_family = AF_INET;
1899#if IS_ENABLED(CONFIG_IPV6)
1900 } else {
1901 loopback.sin6.sin6_addr = in6addr_loopback;
1902 loopback.sa.sa_family = AF_INET6;
1903#endif
1904 }
1905
9dcc71e1 1906 if (dst_vxlan->flags & VXLAN_F_LEARN)
e4c7ed41 1907 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
9dcc71e1
SS		 1908
1909 u64_stats_update_begin(&tx_stats->syncp);
1910 tx_stats->tx_packets++;
ce6502a8 1911 tx_stats->tx_bytes += len;
9dcc71e1
SS		 1912 u64_stats_update_end(&tx_stats->syncp);
1913
1914 if (netif_rx(skb) == NET_RX_SUCCESS) {
1915 u64_stats_update_begin(&rx_stats->syncp);
1916 rx_stats->rx_packets++;
ce6502a8 1917 rx_stats->rx_bytes += len;
9dcc71e1
SS		 1918 u64_stats_update_end(&rx_stats->syncp);
1919 } else {
ce6502a8 1920 dev->stats.rx_dropped++;
9dcc71e1
SS		 1921 }
1922}
1923
4ad16930
SH		 1924static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1925 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1926{
3093fbe7 1927 struct ip_tunnel_info *info;
d342894c 1928 struct vxlan_dev *vxlan = netdev_priv(dev);
b1be00a6 1929 struct sock *sk;
e4c7ed41 1930 struct rtable *rt = NULL;
d342894c 1931 const struct iphdr *old_iph;
d342894c 1932 struct flowi4 fl4;
e4c7ed41 1933 union vxlan_addr *dst;
ee122c79
TG		 1934 union vxlan_addr remote_ip;
1935 struct vxlan_metadata _md;
1936 struct vxlan_metadata *md = &_md;
e4c7ed41 1937 __be16 src_port = 0, dst_port;
234f5b73 1938 u32 vni;
d342894c 1939 __be16 df = 0;
1940 __u8 tos, ttl;
0e6fbc5b 1941 int err;
ee122c79 1942 u32 flags = vxlan->flags;
d342894c 1943
61adedf3 1944 info = skb_tunnel_info(skb);
3093fbe7 1945
ee122c79 1946 if (rdst) {
0dfbdf41 1947 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
ee122c79
TG		 1948 vni = rdst->remote_vni;
1949 dst = &rdst->remote_ip;
1950 } else {
1951 if (!info) {
1952 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1953 dev->name);
1954 goto drop;
1955 }
0dfbdf41 1956 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
ee122c79 1957 vni = be64_to_cpu(info->key.tun_id);
b1be00a6
JB		 1958 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
1959 if (remote_ip.sa.sa_family == AF_INET)
a725e514
JB		 1960 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1961 else
1962 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
ee122c79
TG		 1963 dst = &remote_ip;
1964 }
e4f67add 1965
e4c7ed41 1966 if (vxlan_addr_any(dst)) {
e4f67add 1967 if (did_rsc) {
e4f67add 1968 /* short-circuited back to local bridge */
9dcc71e1 1969 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1970 return;
e4f67add 1971 }
ef59febe 1972 goto drop;
e4f67add 1973 }
ef59febe 1974
d342894c 1975 old_iph = ip_hdr(skb);
1976
0dfbdf41 1977 ttl = vxlan->cfg.ttl;
e4c7ed41 1978 if (!ttl && vxlan_addr_multicast(dst))
d342894c 1979 ttl = 1;
1980
0dfbdf41 1981 tos = vxlan->cfg.tos;
d342894c 1982 if (tos == 1)
206aaafc 1983 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1984
0dfbdf41
TG		 1985 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1986 vxlan->cfg.port_max, true);
d342894c 1987
a725e514
JB		 1988 if (info) {
1989 if (info->key.tun_flags & TUNNEL_CSUM)
1990 flags |= VXLAN_F_UDP_CSUM;
1991 else
1992 flags &= ~VXLAN_F_UDP_CSUM;
1993
1994 ttl = info->key.ttl;
1995 tos = info->key.tos;
1996
1997 if (info->options_len)
4c222798 1998 md = ip_tunnel_info_opts(info);
a725e514
JB		 1999 } else {
2000 md->gbp = skb->mark;
2001 }
2002
e4c7ed41 2003 if (dst->sa.sa_family == AF_INET) {
b1be00a6
JB		 2004 if (!vxlan->vn4_sock)
2005 goto drop;
2006 sk = vxlan->vn4_sock->sock->sk;
2007
a725e514
JB		 2008 if (info && (info->key.tun_flags & TUNNEL_DONT_FRAGMENT))
2009 df = htons(IP_DF);
ee122c79 2010
e4c7ed41 2011 memset(&fl4, 0, sizeof(fl4));
ee122c79 2012 fl4.flowi4_oif = rdst ? rdst->remote_ifindex : 0;
e4c7ed41 2013 fl4.flowi4_tos = RT_TOS(tos);
239fb791
TG		 2014 fl4.flowi4_mark = skb->mark;
2015 fl4.flowi4_proto = IPPROTO_UDP;
e4c7ed41 2016 fl4.daddr = dst->sin.sin_addr.s_addr;
0dfbdf41 2017 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
e4c7ed41 2018
f01ec1c0 2019 rt = ip_route_output_key(vxlan->net, &fl4);
e4c7ed41
CW		 2020 if (IS_ERR(rt)) {
2021 netdev_dbg(dev, "no route to %pI4\n",
2022 &dst->sin.sin_addr.s_addr);
2023 dev->stats.tx_carrier_errors++;
2024 goto tx_error;
2025 }
d342894c 2026
e4c7ed41
CW		 2027 if (rt->dst.dev == dev) {
2028 netdev_dbg(dev, "circular route to %pI4\n",
2029 &dst->sin.sin_addr.s_addr);
2030 dev->stats.collisions++;
fffc15a5 2031 goto rt_tx_error;
e4c7ed41
CW		 2032 }
2033
2034 /* Bypass encapsulation if the destination is local */
2035 if (rt->rt_flags & RTCF_LOCAL &&
2036 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2037 struct vxlan_dev *dst_vxlan;
2038
2039 ip_rt_put(rt);
19ca9fc1 2040 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2041 dst->sa.sa_family, dst_port,
2042 vxlan->flags);
e4c7ed41
CW		 2043 if (!dst_vxlan)
2044 goto tx_error;
2045 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2046 return;
2047 }
2048
2049 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2050 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
79b16aad 2051 err = vxlan_xmit_skb(rt, sk, skb, fl4.saddr,
af33c1ad 2052 dst->sin.sin_addr.s_addr, tos, ttl, df,
614732ea 2053 src_port, dst_port, htonl(vni << 8), md,
af33c1ad 2054 !net_eq(vxlan->net, dev_net(vxlan->dev)),
ee122c79 2055 flags);
74f47278
PS		 2056 if (err < 0) {
2057 /* skb is already freed. */
2058 skb = NULL;
e4c7ed41 2059 goto rt_tx_error;
74f47278
PS		 2060 }
2061
e4c7ed41
CW		 2062 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
2063#if IS_ENABLED(CONFIG_IPV6)
2064 } else {
e4c7ed41 2065 struct dst_entry *ndst;
e5d4b29f 2066 struct in6_addr saddr;
6f264e47 2067 u32 rt6i_flags;
e4c7ed41 2068
b1be00a6
JB		 2069 if (!vxlan->vn6_sock)
2070 goto drop;
2071 sk = vxlan->vn6_sock->sock->sk;
2072
e5d4b29f
JB		 2073 ndst = vxlan6_get_route(vxlan, skb,
2074 rdst ? rdst->remote_ifindex : 0,
2075 &dst->sin6.sin6_addr, &saddr);
2076 if (IS_ERR(ndst)) {
e4c7ed41
CW		 2077 netdev_dbg(dev, "no route to %pI6\n",
2078 &dst->sin6.sin6_addr);
2079 dev->stats.tx_carrier_errors++;
9dcc71e1 2080 goto tx_error;
e4c7ed41 2081 }
d342894c 2082
e4c7ed41
CW		 2083 if (ndst->dev == dev) {
2084 netdev_dbg(dev, "circular route to %pI6\n",
2085 &dst->sin6.sin6_addr);
2086 dst_release(ndst);
2087 dev->stats.collisions++;
2088 goto tx_error;
2089 }
0e6fbc5b 2090
e4c7ed41 2091 /* Bypass encapsulation if the destination is local */
6f264e47
JB		 2092 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2093 if (rt6i_flags & RTF_LOCAL &&
2094 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
e4c7ed41
CW		 2095 struct vxlan_dev *dst_vxlan;
2096
2097 dst_release(ndst);
19ca9fc1 2098 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
ac5132d1
TG		 2099 dst->sa.sa_family, dst_port,
2100 vxlan->flags);
e4c7ed41
CW		 2101 if (!dst_vxlan)
2102 goto tx_error;
2103 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2104 return;
2105 }
49560532 2106
e4c7ed41 2107 ttl = ttl ? : ip6_dst_hoplimit(ndst);
e5d4b29f 2108 err = vxlan6_xmit_skb(ndst, sk, skb, dev, &saddr, &dst->sin6.sin6_addr,
614732ea 2109 0, ttl, src_port, dst_port, htonl(vni << 8), md,
af33c1ad 2110 !net_eq(vxlan->net, dev_net(vxlan->dev)),
a725e514 2111 flags);
e4c7ed41
CW		 2112#endif
2113 }
0e6fbc5b 2114
4ad16930 2115 return;
d342894c 2116
2117drop:
2118 dev->stats.tx_dropped++;
2119 goto tx_free;
2120
49560532
PS		 2121rt_tx_error:
2122 ip_rt_put(rt);
d342894c 2123tx_error:
2124 dev->stats.tx_errors++;
2125tx_free:
2126 dev_kfree_skb(skb);
d342894c 2127}
2128
6681712d
DS		 2129/* Transmit local packets over Vxlan
2130 *
2131 * Outer IP header inherits ECN and DF from inner header.
2132 * Outer UDP destination is the VXLAN assigned port.
2133 * source port is based on hash of flow
2134 */
2135static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2136{
2137 struct vxlan_dev *vxlan = netdev_priv(dev);
3093fbe7 2138 const struct ip_tunnel_info *info;
6681712d
DS		 2139 struct ethhdr *eth;
2140 bool did_rsc = false;
8f646c92 2141 struct vxlan_rdst *rdst, *fdst = NULL;
6681712d 2142 struct vxlan_fdb *f;
6681712d 2143
61adedf3 2144 info = skb_tunnel_info(skb);
3093fbe7 2145
6681712d
DS		 2146 skb_reset_mac_header(skb);
2147 eth = eth_hdr(skb);
2148
f564f45c
CW		 2149 if ((vxlan->flags & VXLAN_F_PROXY)) {
2150 if (ntohs(eth->h_proto) == ETH_P_ARP)
2151 return arp_reduce(dev, skb);
2152#if IS_ENABLED(CONFIG_IPV6)
2153 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
91269e39
LR		 2154 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2155 + sizeof(struct nd_msg)) &&
f564f45c
CW		 2156 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2157 struct nd_msg *msg;
2158
2159 msg = (struct nd_msg *)skb_transport_header(skb);
2160 if (msg->icmph.icmp6_code == 0 &&
2161 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2162 return neigh_reduce(dev, skb);
2163 }
7a9f526f 2164 eth = eth_hdr(skb);
f564f45c
CW		 2165#endif
2166 }
6681712d 2167
da8b43c0 2168 if (vxlan->flags & VXLAN_F_COLLECT_METADATA &&
46fa062a 2169 info && info->mode & IP_TUNNEL_INFO_TX) {
ee122c79
TG		 2170 vxlan_xmit_one(skb, dev, NULL, false);
2171 return NETDEV_TX_OK;
2172 }
2173
6681712d 2174 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 2175 did_rsc = false;
2176
2177 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
e15a00aa
CW		 2178 (ntohs(eth->h_proto) == ETH_P_IP ||
2179 ntohs(eth->h_proto) == ETH_P_IPV6)) {
ae884082
DS		 2180 did_rsc = route_shortcircuit(dev, skb);
2181 if (did_rsc)
2182 f = vxlan_find_mac(vxlan, eth->h_dest);
2183 }
2184
6681712d 2185 if (f == NULL) {
afbd8bae
MR		 2186 f = vxlan_find_mac(vxlan, all_zeros_mac);
2187 if (f == NULL) {
2188 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2189 !is_multicast_ether_addr(eth->h_dest))
2190 vxlan_fdb_miss(vxlan, eth->h_dest);
2191
2192 dev->stats.tx_dropped++;
8f646c92 2193 kfree_skb(skb);
afbd8bae
MR		 2194 return NETDEV_TX_OK;
2195 }
2196 }
6681712d 2197
afbd8bae
MR		 2198 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2199 struct sk_buff *skb1;
6681712d 2200
8f646c92
ED		 2201 if (!fdst) {
2202 fdst = rdst;
2203 continue;
2204 }
afbd8bae
MR		 2205 skb1 = skb_clone(skb, GFP_ATOMIC);
2206 if (skb1)
2207 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 2208 }
2209
8f646c92
ED		 2210 if (fdst)
2211 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2212 else
2213 kfree_skb(skb);
4ad16930 2214 return NETDEV_TX_OK;
6681712d
DS		 2215}
2216
d342894c 2217/* Walk the forwarding table and purge stale entries */
2218static void vxlan_cleanup(unsigned long arg)
2219{
2220 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2221 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2222 unsigned int h;
2223
2224 if (!netif_running(vxlan->dev))
2225 return;
2226
d342894c 2227 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2228 struct hlist_node *p, *n;
14e1d0fa
SD		 2229
2230 spin_lock_bh(&vxlan->hash_lock);
d342894c 2231 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2232 struct vxlan_fdb *f
2233 = container_of(p, struct vxlan_fdb, hlist);
2234 unsigned long timeout;
2235
3c172868 2236 if (f->state & NUD_PERMANENT)
d342894c 2237 continue;
2238
0dfbdf41 2239 timeout = f->used + vxlan->cfg.age_interval * HZ;
d342894c 2240 if (time_before_eq(timeout, jiffies)) {
2241 netdev_dbg(vxlan->dev,
2242 "garbage collect %pM\n",
2243 f->eth_addr);
2244 f->state = NUD_STALE;
2245 vxlan_fdb_destroy(vxlan, f);
2246 } else if (time_before(timeout, next_timer))
2247 next_timer = timeout;
2248 }
14e1d0fa 2249 spin_unlock_bh(&vxlan->hash_lock);
d342894c 2250 }
d342894c 2251
2252 mod_timer(&vxlan->age_timer, next_timer);
2253}
2254
9c2e24e1
PS		 2255static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2256{
56ef9c90 2257 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
9c2e24e1
PS		 2258 __u32 vni = vxlan->default_dst.remote_vni;
2259
56ef9c90 2260 spin_lock(&vn->sock_lock);
9c2e24e1 2261 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
56ef9c90 2262 spin_unlock(&vn->sock_lock);
9c2e24e1
PS		 2263}
2264
d342894c 2265/* Setup stats when device is created */
2266static int vxlan_init(struct net_device *dev)
2267{
1c213bd2 2268 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e8171045 2269 if (!dev->tstats)
d342894c 2270 return -ENOMEM;
2271
2272 return 0;
2273}
2274
ba609e9b 2275static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 2276{
2277 struct vxlan_fdb *f;
2278
2279 spin_lock_bh(&vxlan->hash_lock);
2280 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2281 if (f)
2282 vxlan_fdb_destroy(vxlan, f);
2283 spin_unlock_bh(&vxlan->hash_lock);
2284}
2285
ebf4063e
SH		 2286static void vxlan_uninit(struct net_device *dev)
2287{
2288 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e 2289
ba609e9b 2290 vxlan_fdb_delete_default(vxlan);
afbd8bae 2291
ebf4063e
SH		 2292 free_percpu(dev->tstats);
2293}
2294
d342894c 2295/* Start ageing timer and join group when device is brought up */
2296static int vxlan_open(struct net_device *dev)
2297{
2298 struct vxlan_dev *vxlan = netdev_priv(dev);
205f356d 2299 int ret;
56ef9c90 2300
205f356d
JB		 2301 ret = vxlan_sock_add(vxlan);
2302 if (ret < 0)
2303 return ret;
d342894c 2304
79d4a94f 2305 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
56ef9c90 2306 ret = vxlan_igmp_join(vxlan);
bef0057b
MRL		 2307 if (ret == -EADDRINUSE)
2308 ret = 0;
56ef9c90 2309 if (ret) {
205f356d 2310 vxlan_sock_release(vxlan);
56ef9c90
MRL		 2311 return ret;
2312 }
d342894c 2313 }
2314
0dfbdf41 2315 if (vxlan->cfg.age_interval)
d342894c 2316 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2317
56ef9c90 2318 return ret;
d342894c 2319}
2320
2321/* Purge the forwarding table */
2322static void vxlan_flush(struct vxlan_dev *vxlan)
2323{
31fec5aa 2324 unsigned int h;
d342894c 2325
2326 spin_lock_bh(&vxlan->hash_lock);
2327 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2328 struct hlist_node *p, *n;
2329 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2330 struct vxlan_fdb *f
2331 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 2332 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2333 if (!is_zero_ether_addr(f->eth_addr))
2334 vxlan_fdb_destroy(vxlan, f);
d342894c 2335 }
2336 }
2337 spin_unlock_bh(&vxlan->hash_lock);
2338}
2339
2340/* Cleanup timer and forwarding table on shutdown */
2341static int vxlan_stop(struct net_device *dev)
2342{
2343 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2344 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
56ef9c90 2345 int ret = 0;
d342894c 2346
24c0e683 2347 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
f13b1689 2348 !vxlan_group_used(vn, vxlan))
56ef9c90 2349 ret = vxlan_igmp_leave(vxlan);
d342894c 2350
2351 del_timer_sync(&vxlan->age_timer);
2352
2353 vxlan_flush(vxlan);
205f356d 2354 vxlan_sock_release(vxlan);
d342894c 2355
56ef9c90 2356 return ret;
d342894c 2357}
2358
d342894c 2359/* Stub, nothing needs to be done. */
2360static void vxlan_set_multicast_list(struct net_device *dev)
2361{
2362}
2363
345010b5
DB		 2364static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2365{
2366 struct vxlan_dev *vxlan = netdev_priv(dev);
2367 struct vxlan_rdst *dst = &vxlan->default_dst;
2368 struct net_device *lowerdev;
2369 int max_mtu;
2370
f01ec1c0 2371 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
345010b5
DB		 2372 if (lowerdev == NULL)
2373 return eth_change_mtu(dev, new_mtu);
2374
2375 if (dst->remote_ip.sa.sa_family == AF_INET6)
2376 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2377 else
2378 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2379
2380 if (new_mtu < 68 || new_mtu > max_mtu)
2381 return -EINVAL;
2382
2383 dev->mtu = new_mtu;
2384 return 0;
2385}
2386
fc4099f1
PS		 2387static int egress_ipv4_tun_info(struct net_device *dev, struct sk_buff *skb,
2388 struct ip_tunnel_info *info,
2389 __be16 sport, __be16 dport)
2390{
2391 struct vxlan_dev *vxlan = netdev_priv(dev);
2392 struct rtable *rt;
2393 struct flowi4 fl4;
2394
2395 memset(&fl4, 0, sizeof(fl4));
2396 fl4.flowi4_tos = RT_TOS(info->key.tos);
2397 fl4.flowi4_mark = skb->mark;
2398 fl4.flowi4_proto = IPPROTO_UDP;
2399 fl4.daddr = info->key.u.ipv4.dst;
2400
2401 rt = ip_route_output_key(vxlan->net, &fl4);
2402 if (IS_ERR(rt))
2403 return PTR_ERR(rt);
2404 ip_rt_put(rt);
2405
2406 info->key.u.ipv4.src = fl4.saddr;
2407 info->key.tp_src = sport;
2408 info->key.tp_dst = dport;
2409 return 0;
2410}
2411
2412static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2413{
2414 struct vxlan_dev *vxlan = netdev_priv(dev);
2415 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2416 __be16 sport, dport;
2417
2418 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2419 vxlan->cfg.port_max, true);
2420 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2421
2422 if (ip_tunnel_info_af(info) == AF_INET)
2423 return egress_ipv4_tun_info(dev, skb, info, sport, dport);
2424 return -EINVAL;
2425}
2426
d342894c 2427static const struct net_device_ops vxlan_netdev_ops = {
2428 .ndo_init = vxlan_init,
ebf4063e 2429 .ndo_uninit = vxlan_uninit,
d342894c 2430 .ndo_open = vxlan_open,
2431 .ndo_stop = vxlan_stop,
2432 .ndo_start_xmit = vxlan_xmit,
e8171045 2433 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 2434 .ndo_set_rx_mode = vxlan_set_multicast_list,
345010b5 2435 .ndo_change_mtu = vxlan_change_mtu,
d342894c 2436 .ndo_validate_addr = eth_validate_addr,
2437 .ndo_set_mac_address = eth_mac_addr,
2438 .ndo_fdb_add = vxlan_fdb_add,
2439 .ndo_fdb_del = vxlan_fdb_delete,
2440 .ndo_fdb_dump = vxlan_fdb_dump,
fc4099f1 2441 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
d342894c 2442};
2443
2444/* Info for udev, that this is a virtual tunnel endpoint */
2445static struct device_type vxlan_type = {
2446 .name = "vxlan",
2447};
2448
53cf5275 2449/* Calls the ndo_add_vxlan_port of the caller in order to
35e42379
JG		 2450 * supply the listening VXLAN udp ports. Callers are expected
2451 * to implement the ndo_add_vxlan_port.
53cf5275
JG		 2452 */
2453void vxlan_get_rx_port(struct net_device *dev)
2454{
2455 struct vxlan_sock *vs;
2456 struct net *net = dev_net(dev);
2457 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2458 sa_family_t sa_family;
35e42379
JG		 2459 __be16 port;
2460 unsigned int i;
53cf5275
JG		 2461
2462 spin_lock(&vn->sock_lock);
2463 for (i = 0; i < PORT_HASH_SIZE; ++i) {
35e42379
JG		 2464 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2465 port = inet_sk(vs->sock->sk)->inet_sport;
705cc62f 2466 sa_family = vxlan_get_sk_family(vs);
53cf5275
JG		 2467 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2468 port);
2469 }
2470 }
2471 spin_unlock(&vn->sock_lock);
2472}
2473EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2474
d342894c 2475/* Initialize the device structure. */
2476static void vxlan_setup(struct net_device *dev)
2477{
2478 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 2479 unsigned int h;
d342894c 2480
2481 eth_hw_addr_random(dev);
2482 ether_setup(dev);
2483
2484 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 2485 dev->destructor = free_netdev;
d342894c 2486 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2487
d342894c 2488 dev->features |= NETIF_F_LLTX;
d6727fe3 2489 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 2490 dev->features |= NETIF_F_RXCSUM;
05c0db08 2491 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666 2492
1eaa8178
PS		 2493 dev->vlan_features = dev->features;
2494 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
0afb1666 2495 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 2496 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1eaa8178 2497 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
02875878 2498 netif_keep_dst(dev);
22dba393 2499 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
d342894c 2500
553675fb 2501 INIT_LIST_HEAD(&vxlan->next);
d342894c 2502 spin_lock_init(&vxlan->hash_lock);
2503
2504 init_timer_deferrable(&vxlan->age_timer);
2505 vxlan->age_timer.function = vxlan_cleanup;
2506 vxlan->age_timer.data = (unsigned long) vxlan;
2507
0dfbdf41 2508 vxlan->cfg.dst_port = htons(vxlan_port);
05f47d69 2509
d342894c 2510 vxlan->dev = dev;
2511
58ce31cc
TH		 2512 gro_cells_init(&vxlan->gro_cells, dev);
2513
d342894c 2514 for (h = 0; h < FDB_HASH_SIZE; ++h)
2515 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2516}
2517
2518static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2519 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 2520 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
e4c7ed41 2521 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
d342894c 2522 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2523 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
e4c7ed41 2524 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
d342894c 2525 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2526 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2527 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2528 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2529 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 2530 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 2531 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2532 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2533 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2534 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
f8a9b1bc 2535 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
823aa873 2536 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
5c91ae08
TH		 2537 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2538 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2539 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
dfd8645e
TH		 2540 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2541 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3511494c 2542 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
0ace2ca8 2543 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
d342894c 2544};
2545
2546static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2547{
2548 if (tb[IFLA_ADDRESS]) {
2549 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2550 pr_debug("invalid link address (not ethernet)\n");
2551 return -EINVAL;
2552 }
2553
2554 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2555 pr_debug("invalid all zero ethernet address\n");
2556 return -EADDRNOTAVAIL;
2557 }
2558 }
2559
2560 if (!data)
2561 return -EINVAL;
2562
2563 if (data[IFLA_VXLAN_ID]) {
2564 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2565 if (id >= VXLAN_VID_MASK)
2566 return -ERANGE;
2567 }
2568
05f47d69 2569 if (data[IFLA_VXLAN_PORT_RANGE]) {
2570 const struct ifla_vxlan_port_range *p
2571 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2572
2573 if (ntohs(p->high) < ntohs(p->low)) {
2574 pr_debug("port range %u .. %u not valid\n",
2575 ntohs(p->low), ntohs(p->high));
2576 return -EINVAL;
2577 }
2578 }
2579
d342894c 2580 return 0;
2581}
2582
1b13c97f
YB		 2583static void vxlan_get_drvinfo(struct net_device *netdev,
2584 struct ethtool_drvinfo *drvinfo)
2585{
2586 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2587 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2588}
2589
2590static const struct ethtool_ops vxlan_ethtool_ops = {
2591 .get_drvinfo = vxlan_get_drvinfo,
2592 .get_link = ethtool_op_get_link,
2593};
2594
553675fb 2595static void vxlan_del_work(struct work_struct *work)
2596{
2597 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
acbf74a7 2598 udp_tunnel_sock_release(vs->sock);
553675fb 2599 kfree_rcu(vs, rcu);
2600}
2601
3ee64f39
TH		 2602static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2603 __be16 port, u32 flags)
553675fb 2604{
e4c7ed41 2605 struct socket *sock;
3ee64f39
TH		 2606 struct udp_port_cfg udp_conf;
2607 int err;
e4c7ed41 2608
3ee64f39 2609 memset(&udp_conf, 0, sizeof(udp_conf));
e4c7ed41 2610
3ee64f39
TH		 2611 if (ipv6) {
2612 udp_conf.family = AF_INET6;
3ee64f39 2613 udp_conf.use_udp6_rx_checksums =
3dc2b6a8 2614 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
a43a9ef6 2615 udp_conf.ipv6_v6only = 1;
3ee64f39
TH		 2616 } else {
2617 udp_conf.family = AF_INET;
e4c7ed41 2618 }
e4c7ed41 2619
3ee64f39 2620 udp_conf.local_udp_port = port;
553675fb 2621
3ee64f39
TH		 2622 /* Open UDP socket */
2623 err = udp_sock_create(net, &udp_conf, &sock);
2624 if (err < 0)
2625 return ERR_PTR(err);
e4c7ed41 2626
39deb2c7 2627 return sock;
e4c7ed41
CW		 2628}
2629
2630/* Create new listen socket if needed */
b1be00a6
JB		 2631static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2632 __be16 port, u32 flags)
e4c7ed41
CW		 2633{
2634 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2635 struct vxlan_sock *vs;
2636 struct socket *sock;
e4c7ed41 2637 unsigned int h;
acbf74a7 2638 struct udp_tunnel_sock_cfg tunnel_cfg;
e4c7ed41 2639
dc01e7d3 2640 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
e4c7ed41
CW		 2641 if (!vs)
2642 return ERR_PTR(-ENOMEM);
2643
2644 for (h = 0; h < VNI_HASH_SIZE; ++h)
2645 INIT_HLIST_HEAD(&vs->vni_list[h]);
2646
2647 INIT_WORK(&vs->del_work, vxlan_del_work);
2648
3ee64f39 2649 sock = vxlan_create_sock(net, ipv6, port, flags);
39deb2c7 2650 if (IS_ERR(sock)) {
56ef9c90
MRL		 2651 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2652 PTR_ERR(sock));
553675fb 2653 kfree(vs);
e50fddc8 2654 return ERR_CAST(sock);
553675fb 2655 }
e4c7ed41
CW		 2656
2657 vs->sock = sock;
9c2e24e1 2658 atomic_set(&vs->refcnt, 1);
af33c1ad 2659 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
553675fb 2660
dc01e7d3
OG		 2661 /* Initialize the vxlan udp offloads structure */
2662 vs->udp_offloads.port = port;
2663 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2664 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2665
9c2e24e1
PS		 2666 spin_lock(&vn->sock_lock);
2667 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
dc01e7d3 2668 vxlan_notify_add_rx_port(vs);
9c2e24e1 2669 spin_unlock(&vn->sock_lock);
553675fb 2670
2671 /* Mark socket as an encapsulation socket. */
acbf74a7
AZ		 2672 tunnel_cfg.sk_user_data = vs;
2673 tunnel_cfg.encap_type = 1;
2674 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2675 tunnel_cfg.encap_destroy = NULL;
2676
2677 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
e4c7ed41 2678
9c2e24e1
PS		 2679 return vs;
2680}
2681
b1be00a6 2682static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
9c2e24e1 2683{
205f356d
JB		 2684 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2685 struct vxlan_sock *vs = NULL;
9c2e24e1 2686
205f356d 2687 if (!vxlan->cfg.no_share) {
56ef9c90 2688 spin_lock(&vn->sock_lock);
205f356d
JB		 2689 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2690 vxlan->cfg.dst_port, vxlan->flags);
2691 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
56ef9c90 2692 spin_unlock(&vn->sock_lock);
205f356d 2693 return -EBUSY;
56ef9c90
MRL		 2694 }
2695 spin_unlock(&vn->sock_lock);
2696 }
205f356d 2697 if (!vs)
b1be00a6
JB		 2698 vs = vxlan_socket_create(vxlan->net, ipv6,
2699 vxlan->cfg.dst_port, vxlan->flags);
205f356d
JB		 2700 if (IS_ERR(vs))
2701 return PTR_ERR(vs);
b1be00a6
JB		 2702#if IS_ENABLED(CONFIG_IPV6)
2703 if (ipv6)
2704 vxlan->vn6_sock = vs;
2705 else
2706#endif
2707 vxlan->vn4_sock = vs;
205f356d
JB		 2708 vxlan_vs_add_dev(vs, vxlan);
2709 return 0;
553675fb 2710}
2711
b1be00a6
JB		 2712static int vxlan_sock_add(struct vxlan_dev *vxlan)
2713{
2714 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2715 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2716 int ret = 0;
2717
2718 vxlan->vn4_sock = NULL;
2719#if IS_ENABLED(CONFIG_IPV6)
2720 vxlan->vn6_sock = NULL;
2721 if (ipv6 || metadata)
2722 ret = __vxlan_sock_add(vxlan, true);
2723#endif
2724 if (!ret && (!ipv6 || metadata))
2725 ret = __vxlan_sock_add(vxlan, false);
2726 if (ret < 0)
2727 vxlan_sock_release(vxlan);
2728 return ret;
2729}
2730
0dfbdf41
TG		 2731static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2732 struct vxlan_config *conf)
d342894c 2733{
33564bbb 2734 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
d342894c 2735 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 2736 struct vxlan_rdst *dst = &vxlan->default_dst;
b1be00a6 2737 unsigned short needed_headroom = ETH_HLEN;
d342894c 2738 int err;
e4c7ed41 2739 bool use_ipv6 = false;
0dfbdf41 2740 __be16 default_port = vxlan->cfg.dst_port;
d342894c 2741
33564bbb 2742 vxlan->net = src_net;
f01ec1c0 2743
0dfbdf41 2744 dst->remote_vni = conf->vni;
d342894c 2745
0dfbdf41 2746 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
d342894c 2747
0dfbdf41
TG		 2748 /* Unless IPv6 is explicitly requested, assume IPv4 */
2749 if (!dst->remote_ip.sa.sa_family)
2750 dst->remote_ip.sa.sa_family = AF_INET;
e4c7ed41 2751
0dfbdf41 2752 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
057ba29b
JB		 2753 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2754 if (!IS_ENABLED(CONFIG_IPV6))
2755 return -EPFNOSUPPORT;
e4c7ed41 2756 use_ipv6 = true;
b1be00a6 2757 vxlan->flags |= VXLAN_F_IPV6;
057ba29b 2758 }
d342894c 2759
0dfbdf41 2760 if (conf->remote_ifindex) {
34e02aa1 2761 struct net_device *lowerdev
0dfbdf41
TG		 2762 = __dev_get_by_index(src_net, conf->remote_ifindex);
2763
2764 dst->remote_ifindex = conf->remote_ifindex;
34e02aa1 2765
2766 if (!lowerdev) {
c7995c43 2767 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 2768 return -ENODEV;
2769 }
d342894c 2770
e4c7ed41
CW		 2771#if IS_ENABLED(CONFIG_IPV6)
2772 if (use_ipv6) {
2773 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2774 if (idev && idev->cnf.disable_ipv6) {
2775 pr_info("IPv6 is disabled via sysctl\n");
2776 return -EPERM;
2777 }
e4c7ed41
CW		 2778 }
2779#endif
2780
0dfbdf41 2781 if (!conf->mtu)
e4c7ed41 2782 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1ba56fb4 2783
b1be00a6 2784 needed_headroom = lowerdev->hard_header_len;
9dc2ad10 2785 }
d342894c 2786
b1be00a6
JB		 2787 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2788 needed_headroom += VXLAN6_HEADROOM;
2789 else
2790 needed_headroom += VXLAN_HEADROOM;
2791 dev->needed_headroom = needed_headroom;
2792
0dfbdf41
TG		 2793 memcpy(&vxlan->cfg, conf, sizeof(*conf));
2794 if (!vxlan->cfg.dst_port)
2795 vxlan->cfg.dst_port = default_port;
2796 vxlan->flags |= conf->flags;
2797
2798 if (!vxlan->cfg.age_interval)
2799 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2800
2801 if (vxlan_find_vni(src_net, conf->vni, use_ipv6 ? AF_INET6 : AF_INET,
2802 vxlan->cfg.dst_port, vxlan->flags))
2803 return -EEXIST;
2804
2805 dev->ethtool_ops = &vxlan_ethtool_ops;
2806
2807 /* create an fdb entry for a valid default destination */
2808 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2809 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2810 &vxlan->default_dst.remote_ip,
2811 NUD_REACHABLE|NUD_PERMANENT,
2812 NLM_F_EXCL|NLM_F_CREATE,
2813 vxlan->cfg.dst_port,
2814 vxlan->default_dst.remote_vni,
2815 vxlan->default_dst.remote_ifindex,
2816 NTF_SELF);
2817 if (err)
2818 return err;
2819 }
2820
2821 err = register_netdevice(dev);
2822 if (err) {
2823 vxlan_fdb_delete_default(vxlan);
2824 return err;
2825 }
2826
2827 list_add(&vxlan->next, &vn->vxlan_list);
2828
2829 return 0;
2830}
2831
2832struct net_device *vxlan_dev_create(struct net *net, const char *name,
2833 u8 name_assign_type, struct vxlan_config *conf)
2834{
2835 struct nlattr *tb[IFLA_MAX+1];
2836 struct net_device *dev;
2837 int err;
2838
2839 memset(&tb, 0, sizeof(tb));
2840
2841 dev = rtnl_create_link(net, name, name_assign_type,
2842 &vxlan_link_ops, tb);
2843 if (IS_ERR(dev))
2844 return dev;
2845
2846 err = vxlan_dev_configure(net, dev, conf);
2847 if (err < 0) {
2848 free_netdev(dev);
2849 return ERR_PTR(err);
2850 }
2851
2852 return dev;
2853}
2854EXPORT_SYMBOL_GPL(vxlan_dev_create);
2855
2856static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2857 struct nlattr *tb[], struct nlattr *data[])
2858{
2859 struct vxlan_config conf;
2860 int err;
2861
0dfbdf41 2862 memset(&conf, 0, sizeof(conf));
e277de5f
JG		 2863
2864 if (data[IFLA_VXLAN_ID])
2865 conf.vni = nla_get_u32(data[IFLA_VXLAN_ID]);
0dfbdf41
TG		 2866
2867 if (data[IFLA_VXLAN_GROUP]) {
2868 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2869 } else if (data[IFLA_VXLAN_GROUP6]) {
2870 if (!IS_ENABLED(CONFIG_IPV6))
2871 return -EPFNOSUPPORT;
2872
2873 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2874 conf.remote_ip.sa.sa_family = AF_INET6;
2875 }
2876
2877 if (data[IFLA_VXLAN_LOCAL]) {
2878 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2879 conf.saddr.sa.sa_family = AF_INET;
2880 } else if (data[IFLA_VXLAN_LOCAL6]) {
2881 if (!IS_ENABLED(CONFIG_IPV6))
2882 return -EPFNOSUPPORT;
2883
2884 /* TODO: respect scope id */
2885 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2886 conf.saddr.sa.sa_family = AF_INET6;
2887 }
2888
2889 if (data[IFLA_VXLAN_LINK])
2890 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2891
d342894c 2892 if (data[IFLA_VXLAN_TOS])
0dfbdf41 2893 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
d342894c 2894
afb97186 2895 if (data[IFLA_VXLAN_TTL])
0dfbdf41 2896 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
afb97186 2897
d342894c 2898 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
0dfbdf41 2899 conf.flags |= VXLAN_F_LEARN;
d342894c 2900
2901 if (data[IFLA_VXLAN_AGEING])
0dfbdf41 2902 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
d342894c 2903
e4f67add 2904 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
0dfbdf41 2905 conf.flags |= VXLAN_F_PROXY;
e4f67add
DS		 2906
2907 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
0dfbdf41 2908 conf.flags |= VXLAN_F_RSC;
e4f67add
DS		 2909
2910 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
0dfbdf41 2911 conf.flags |= VXLAN_F_L2MISS;
e4f67add
DS		 2912
2913 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
0dfbdf41 2914 conf.flags |= VXLAN_F_L3MISS;
e4f67add 2915
d342894c 2916 if (data[IFLA_VXLAN_LIMIT])
0dfbdf41 2917 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
d342894c 2918
f8a9b1bc
AS		 2919 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
2920 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
2921 conf.flags |= VXLAN_F_COLLECT_METADATA;
2922
05f47d69 2923 if (data[IFLA_VXLAN_PORT_RANGE]) {
2924 const struct ifla_vxlan_port_range *p
2925 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
0dfbdf41
TG		 2926 conf.port_min = ntohs(p->low);
2927 conf.port_max = ntohs(p->high);
05f47d69 2928 }
2929
823aa873 2930 if (data[IFLA_VXLAN_PORT])
0dfbdf41 2931 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
823aa873 2932
359a0ea9 2933 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
0dfbdf41 2934 conf.flags |= VXLAN_F_UDP_CSUM;
359a0ea9
TH		 2935
2936 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2937 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
0dfbdf41 2938 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
359a0ea9
TH		 2939
2940 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2941 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
0dfbdf41 2942 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
359a0ea9 2943
dfd8645e
TH		 2944 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2945 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
0dfbdf41 2946 conf.flags |= VXLAN_F_REMCSUM_TX;
dfd8645e
TH		 2947
2948 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2949 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
0dfbdf41 2950 conf.flags |= VXLAN_F_REMCSUM_RX;
dfd8645e 2951
3511494c 2952 if (data[IFLA_VXLAN_GBP])
0dfbdf41 2953 conf.flags |= VXLAN_F_GBP;
3511494c 2954
0ace2ca8 2955 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
0dfbdf41 2956 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
0ace2ca8 2957
0dfbdf41
TG		 2958 err = vxlan_dev_configure(src_net, dev, &conf);
2959 switch (err) {
2960 case -ENODEV:
2961 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
2962 break;
1b13c97f 2963
0dfbdf41
TG		 2964 case -EPERM:
2965 pr_info("IPv6 is disabled via sysctl\n");
2966 break;
d342894c 2967
0dfbdf41
TG		 2968 case -EEXIST:
2969 pr_info("duplicate VNI %u\n", conf.vni);
2970 break;
afbd8bae
MR		 2971 }
2972
0dfbdf41 2973 return err;
d342894c 2974}
2975
2976static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2977{
2978 struct vxlan_dev *vxlan = netdev_priv(dev);
f01ec1c0 2979 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
d342894c 2980
fe5c3561 2981 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 2982 if (!hlist_unhashed(&vxlan->hlist))
2983 hlist_del_rcu(&vxlan->hlist);
fe5c3561 2984 spin_unlock(&vn->sock_lock);
2985
58ce31cc 2986 gro_cells_destroy(&vxlan->gro_cells);
553675fb 2987 list_del(&vxlan->next);
d342894c 2988 unregister_netdevice_queue(dev, head);
2989}
2990
2991static size_t vxlan_get_size(const struct net_device *dev)
2992{
2993
2994 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
e4c7ed41 2995 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
d342894c 2996 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
e4c7ed41 2997 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
d342894c 2998 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2999 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
3000 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 3001 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3002 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3003 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3004 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
da8b43c0 3005 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
d342894c 3006 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3007 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 3008 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
359a0ea9
TH		 3009 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3010 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3011 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3012 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
dfd8645e
TH		 3013 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3014 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
d342894c 3015 0;
3016}
3017
3018static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3019{
3020 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 3021 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 3022 struct ifla_vxlan_port_range ports = {
0dfbdf41
TG		 3023 .low = htons(vxlan->cfg.port_min),
3024 .high = htons(vxlan->cfg.port_max),
05f47d69 3025 };
d342894c 3026
c7995c43 3027 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 3028 goto nla_put_failure;
3029
e4c7ed41
CW		 3030 if (!vxlan_addr_any(&dst->remote_ip)) {
3031 if (dst->remote_ip.sa.sa_family == AF_INET) {
930345ea
JB		 3032 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3033 dst->remote_ip.sin.sin_addr.s_addr))
e4c7ed41
CW		 3034 goto nla_put_failure;
3035#if IS_ENABLED(CONFIG_IPV6)
3036 } else {
930345ea
JB		 3037 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3038 &dst->remote_ip.sin6.sin6_addr))
e4c7ed41
CW		 3039 goto nla_put_failure;
3040#endif
3041 }
3042 }
d342894c 3043
c7995c43 3044 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 3045 goto nla_put_failure;
3046
0dfbdf41
TG		 3047 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3048 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
930345ea 3049 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
0dfbdf41 3050 vxlan->cfg.saddr.sin.sin_addr.s_addr))
e4c7ed41
CW		 3051 goto nla_put_failure;
3052#if IS_ENABLED(CONFIG_IPV6)
3053 } else {
930345ea 3054 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
0dfbdf41 3055 &vxlan->cfg.saddr.sin6.sin6_addr))
e4c7ed41
CW		 3056 goto nla_put_failure;
3057#endif
3058 }
3059 }
d342894c 3060
0dfbdf41
TG		 3061 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3062 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
e4f67add
DS		 3063 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3064 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3065 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3066 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3067 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3068 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3069 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3070 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3071 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
da8b43c0
AS		 3072 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3073 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
0dfbdf41
TG		 3074 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3075 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3076 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
359a0ea9
TH		 3077 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
3078 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
3079 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3080 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3081 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
dfd8645e
TH		 3082 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3083 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3084 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3085 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3086 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
d342894c 3087 goto nla_put_failure;
3088
05f47d69 3089 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3090 goto nla_put_failure;
3091
3511494c
TG		 3092 if (vxlan->flags & VXLAN_F_GBP &&
3093 nla_put_flag(skb, IFLA_VXLAN_GBP))
3094 goto nla_put_failure;
3095
0ace2ca8
TH		 3096 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3097 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3098 goto nla_put_failure;
3099
d342894c 3100 return 0;
3101
3102nla_put_failure:
3103 return -EMSGSIZE;
3104}
3105
1728d4fa
ND		 3106static struct net *vxlan_get_link_net(const struct net_device *dev)
3107{
3108 struct vxlan_dev *vxlan = netdev_priv(dev);
3109
3110 return vxlan->net;
3111}
3112
d342894c 3113static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3114 .kind = "vxlan",
3115 .maxtype = IFLA_VXLAN_MAX,
3116 .policy = vxlan_policy,
3117 .priv_size = sizeof(struct vxlan_dev),
3118 .setup = vxlan_setup,
3119 .validate = vxlan_validate,
3120 .newlink = vxlan_newlink,
3121 .dellink = vxlan_dellink,
3122 .get_size = vxlan_get_size,
3123 .fill_info = vxlan_fill_info,
1728d4fa 3124 .get_link_net = vxlan_get_link_net,
d342894c 3125};
3126
acaf4e70
DB		 3127static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3128 struct net_device *dev)
3129{
3130 struct vxlan_dev *vxlan, *next;
3131 LIST_HEAD(list_kill);
3132
3133 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3134 struct vxlan_rdst *dst = &vxlan->default_dst;
3135
3136 /* In case we created vxlan device with carrier
3137 * and we loose the carrier due to module unload
3138 * we also need to remove vxlan device. In other
3139 * cases, it's not necessary and remote_ifindex
3140 * is 0 here, so no matches.
3141 */
3142 if (dst->remote_ifindex == dev->ifindex)
3143 vxlan_dellink(vxlan->dev, &list_kill);
3144 }
3145
3146 unregister_netdevice_many(&list_kill);
3147}
3148
3149static int vxlan_lowerdev_event(struct notifier_block *unused,
3150 unsigned long event, void *ptr)
3151{
3152 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
783c1463 3153 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
acaf4e70 3154
783c1463 3155 if (event == NETDEV_UNREGISTER)
acaf4e70
DB		 3156 vxlan_handle_lowerdev_unregister(vn, dev);
3157
3158 return NOTIFY_DONE;
3159}
3160
3161static struct notifier_block vxlan_notifier_block __read_mostly = {
3162 .notifier_call = vxlan_lowerdev_event,
3163};
3164
d342894c 3165static __net_init int vxlan_init_net(struct net *net)
3166{
3167 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 3168 unsigned int h;
d342894c 3169
553675fb 3170 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 3171 spin_lock_init(&vn->sock_lock);
d342894c 3172
553675fb 3173 for (h = 0; h < PORT_HASH_SIZE; ++h)
3174 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 3175
3176 return 0;
3177}
3178
f01ec1c0
ND		 3179static void __net_exit vxlan_exit_net(struct net *net)
3180{
3181 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3182 struct vxlan_dev *vxlan, *next;
3183 struct net_device *dev, *aux;
3184 LIST_HEAD(list);
3185
3186 rtnl_lock();
3187 for_each_netdev_safe(net, dev, aux)
3188 if (dev->rtnl_link_ops == &vxlan_link_ops)
3189 unregister_netdevice_queue(dev, &list);
3190
3191 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3192 /* If vxlan->dev is in the same netns, it has already been added
3193 * to the list by the previous loop.
3194 */
58ce31cc
TH		 3195 if (!net_eq(dev_net(vxlan->dev), net)) {
3196 gro_cells_destroy(&vxlan->gro_cells);
13c3ed6a 3197 unregister_netdevice_queue(vxlan->dev, &list);
58ce31cc 3198 }
f01ec1c0
ND		 3199 }
3200
3201 unregister_netdevice_many(&list);
3202 rtnl_unlock();
3203}
3204
d342894c 3205static struct pernet_operations vxlan_net_ops = {
3206 .init = vxlan_init_net,
f01ec1c0 3207 .exit = vxlan_exit_net,
d342894c 3208 .id = &vxlan_net_id,
3209 .size = sizeof(struct vxlan_net),
3210};
3211
3212static int __init vxlan_init_module(void)
3213{
3214 int rc;
3215
758c57d1
SH		 3216 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3217 if (!vxlan_wq)
3218 return -ENOMEM;
3219
d342894c 3220 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3221
783c1463 3222 rc = register_pernet_subsys(&vxlan_net_ops);
d342894c 3223 if (rc)
3224 goto out1;
3225
acaf4e70 3226 rc = register_netdevice_notifier(&vxlan_notifier_block);
d342894c 3227 if (rc)
3228 goto out2;
3229
acaf4e70
DB		 3230 rc = rtnl_link_register(&vxlan_link_ops);
3231 if (rc)
3232 goto out3;
d342894c 3233
acaf4e70
DB		 3234 return 0;
3235out3:
3236 unregister_netdevice_notifier(&vxlan_notifier_block);
d342894c 3237out2:
783c1463 3238 unregister_pernet_subsys(&vxlan_net_ops);
d342894c 3239out1:
758c57d1 3240 destroy_workqueue(vxlan_wq);
d342894c 3241 return rc;
3242}
7332a13b 3243late_initcall(vxlan_init_module);
d342894c 3244
3245static void __exit vxlan_cleanup_module(void)
3246{
b7153984 3247 rtnl_link_unregister(&vxlan_link_ops);
acaf4e70 3248 unregister_netdevice_notifier(&vxlan_notifier_block);
758c57d1 3249 destroy_workqueue(vxlan_wq);
783c1463
DB		 3250 unregister_pernet_subsys(&vxlan_net_ops);
3251 /* rcu_barrier() is called by netns */
d342894c 3252}
3253module_exit(vxlan_cleanup_module);
3254
3255MODULE_LICENSE("GPL");
3256MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 3257MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
ead5139a 3258MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
d342894c 3259MODULE_ALIAS_RTNL_LINK("vxlan");
Linux kernel - Deliverables
This page took 0.519221 seconds and 5 git commands to generate.